Further search for supersymmetry at √s=7TeV in final states with jets, missing transverse momentum, and isolated leptons with the ATLAS detector

This work presents a new inclusive search for supersymmetry (SUSY) by the ATLAS experiment at the LHC in proton-proton collisions at a center-of-mass energy ﬃﬃﬃ s p ¼ 7 TeV in ﬁnal states with jets, missing transverse momentum and one or more isolated electrons and/or muons. The search is based on data from the full 2011 data-taking period, corresponding to an integrated luminosity of 4 : 7 fb (cid:1) 1 . Single-lepton and multilepton channels are treated together in one analysis. An increase in sensitivity is obtained by simultaneously ﬁtting the number of events in statistically independent signal regions, and the shapes of distributions within those regions. A dedicated signal region is introduced to be sensitive to decay cascades of SUSY particles with small mass differences (‘‘compressed SUSY’’). Background uncertainties are constrained by ﬁtting to the jet-multiplicity distribution in background control regions. Observations are consistent with Standard Model expectations, and limits are set or extended on a number of SUSY models. show the ratio between data and the total Standard Model expectation. expectation for multijets is derived from the data. The remaining Standard Model expectation is entirely derived from simulation, normalized to the theoretical cross sections. The uncertainty band around the Standard Model expectation combines the statistical uncertainty on the simulated event samples with the systematic uncertainties on the jet energy scale, b tagging, data-driven multijet background, and luminosity. systematic uncertainties are largely correlated


I. INTRODUCTION
Supersymmetry (SUSY) [1][2][3][4][5][6][7][8][9] is a candidate for physics beyond the Standard Model (SM). If strongly interacting supersymmetric particles are present at the TeV scale, they may be copiously produced in 7 TeV proton-proton collisions at the Large Hadron Collider [10]. In the minimal supersymmetric extension of the Standard Model (MSSM) [11][12][13][14][15] such particles decay into jets, leptons and the lightest supersymmetric particle (LSP). Jets arise in the decays of squarks and gluinos, while leptons can arise in decays involving charginos or neutralinos. A long-lived, weakly interacting LSP will escape detection, leading to missing transverse momentum (p miss T and its magnitude E miss T ) in the final state. Significant E miss T can also arise in scenarios where neutrinos are created somewhere in the SUSY decay cascade. This paper presents a new inclusive search with the ATLAS detector for SUSY in final states containing jets, one or more isolated leptons (electrons or muons) and E miss T . Previous searches in these channels have been conducted by both the ATLAS [16][17][18] and CMS [19][20][21][22] collaborations. In this paper, the analysis is extended to 4:7 fb À1 , and single-lepton and multilepton channels (with jets and E miss T ) are treated simultaneously. A signal region with a soft lepton and soft jets is introduced in order to probe SUSY decays involving small mass differences between the particles in the decay chain. A new, simultaneous fit to the yield in multiple signal regions and to the shapes of distributions within those signal regions is employed. Background uncertainties are constrained by fitting to the jet-multiplicity distribution in background control regions.

II. THE ATLAS DETECTOR
The ATLAS detector [23,24] consists of a tracking system (inner detector, ID) surrounded by a thin superconducting solenoid providing a 2 T magnetic field, electromagnetic and hadronic calorimeters and a muon spectrometer (MS). The ID consists of pixel and silicon microstrip detectors, surrounded by a straw-tube tracker with transition radiation detection (transition radiation tracker, TRT). The electromagnetic calorimeter is a lead liquid-argon (LAr) detector. Hadronic calorimetry is based on two different detector technologies, with scintillator tiles or LAr as active media, and with either steel, copper, or tungsten as the absorber material. The MS is based on three large superconducting toroid systems arranged with an eight-fold azimuthal coil symmetry around the calorimeters, and three stations of chambers for the trigger and for precise position measurements. The nominal pp interaction point at the center of the detector is defined as the origin of a right-handed coordinate system. The positive x axis is defined by the direction from the interaction point to the center of the LHC ring, with the positive y axis pointing upwards, while the beam direction defines the z axis. The azimuthal angle is measured around the beam axis and the polar angle is the angle from the z axis. The pseudorapidity is defined as ¼ À ln tanð=2Þ. Transverse coordinates, such as the transverse momentum, p T , are defined in the (x-y) plane.

III. SUSY SIGNAL MODELING AND SIMULATED EVENT SAMPLES
The SUSY models considered are minimal supergravity (MSUGRA) or constrained MSSM (CMSSM) [25,26], minimal gauge-mediated SUSY breaking (GMSB) [27][28][29][30][31] and a number of simplified models [32,33]. The MSUGRA/CMSSM model is characterized by five parameters: the universal scalar and gaugino mass parameters m 0 and m 1=2 , a universal trilinear coupling parameter A 0 , the ratio of the vacuum expectation values of the two Higgs doublets tan, and the sign of the Higgsino mass parameter . In this analysis, the values of m 0 and m 1=2 are scanned, and the other parameters are fixed as follows: tan ¼ 10, A 0 ¼ 0 and is taken to be positive. A diagram showing the decay of the associated production of a squark and a gluino is depicted in Fig. 1(a). Other diagrams representative for the SUSY models discussed in the following are shown in Figs. 1 The minimal GMSB model has six parameters: the SUSY breaking scale Ã, the mass scale of the messenger fields M mes , the number of messenger fields N 5 , the scale of the gravitino coupling C grav , the ratio of the vacuum expectation values of the two Higgs doublets tan, and the sign of the Higgsino mass parameter . For the minimal GMSB model, the parameters tan and Ã are scanned and the other parameters are assigned fixed values: M mes ¼ 250 TeV, N 5 ¼ 3, C grav ¼ 1 and the sign of is taken to be positive. The mass scale of the colored superpartners is set by the parameter Ã, while the nextto-lightest SUSY particle (NLSP) is determined by a combination of Ã and tan. At low values of Ã, the NLSP is the lightest neutralino ( 0 1 ) while at the higher values of Ã where this search provides new sensitivity, the NLSP is a stau for tan * 10 and a slepton of the first and second generation otherwise. The NLSP decays into its SM partner and a nearly massless gravitino. The gaugino and sfermion masses are proportional to N 5 and ffiffiffiffiffiffi N 5 p , respectively. The parameter C grav determines the NLSP lifetime, set here such that all NLSPs decay promptly.
Several simplified models are considered in this paper. In the ''one-step'' models, SUSY production proceeds via either pp !gg or pp !q Lq Ã L , where only left-handed squarks of the first and second generation are considered. The gluino decays to the neutralino LSP viag ! q " q 0AE 1 ! q " q 0 W AE0 1 , and the squark viaq L ! q 0AE 1 ! q 0 W AE0 1 , where the W boson can be real or virtual. The gluino and LSP masses are varied while the chargino mass is set to be halfway between them. In a variant of the one-step model, the LSP mass is held fixed at 60 GeV while the gluino (squark) and chargino masses are scanned.
In the ''two-step'' models, SUSY production proceeds via either pp !gg or pp !q Lq Ã L , again where squarks of the first and second generation are considered. In the first class of two-step models all squarks and gluinos decay via a chargino:g ! q " q 0AE 1 andq L ! q 0AE 1 . The charginos decay via AE 1 ! ' L or AE 1 ! ' L ; in the case of third generation sleptons, the decay to the stau is via AE 1 ! 1 . All three generations of sleptons and sneutrinos are allowed with equal probability, resulting in an equal branching ratio to sleptons and to sneutrinos. In the second class of two-step models, the gluinos or left-handed squarks decay either via a chargino (g ! q " q 0AE 1 orq L ! q 0AE 1 ) or via a neutralino (g ! q " q 0 2 orq L ! q 0 2 ). The events are generated such that one chargino and one neutralino are always present in the decays of the pair produced gluinos or left-handed squarks. Neutralino decays proceed via either 0 2 ! '' L or 0 2 ! . As in the first two-step model, all three generations of sleptons and sneutrinos are allowed with equal probability, resulting in a 50% branching ratio to sleptons and to sneutrinos. Finally, in the third class of two-step models without intermediate sleptons, the gluino and squark decay viag ! q " q 0AE 1 or q L ! q 0AE 1 ; the decay of the chargino then proceeds viã AE 1 ! W ðÃÞAE0 2 ! W ðÃÞAE Z ðÃÞ0 1 . This signature is realized in the MSSM in a parameter region where additional decay modes, not contained in the simplified model, may lead to a significant reduction of the cross section times branching fraction of the WZ signature.
In the first two types of two-step models, the chargino and neutralino have equal masses (again set to be halfway between the gluino/squark and LSP mass); the slepton and sneutrino masses are set to be equal and halfway between FIG. 1. Representative diagrams for the different SUSY models considered in this analysis: (a) MSUGRA/CMSSM model with pp !qg and subsequent decay of the squark via a chargino; (b) GMSB model with pp !qq Ã and subsequent decay via sleptons and staus; (c) one-step simplified model with pp !q Lq the chargino/neutralino and LSP masses. In the third twostep model, the AE 1 mass is set halfway between the gluino/ squark and LSP while the 0 2 mass is set halfway between the chargino and LSP. In all the simplified models, the superpartners that have not been mentioned are decoupled by setting their masses to multi-TeV values.
Simulated event samples are used for estimating the signal acceptance, the detector efficiency, and for estimating many of the backgrounds (in most cases in association with data-driven techniques). The MSUGRA/CMSSM and minimal GMSB signal samples are generated with HERWIG++ 2.5.2 [34] and MRST2007LO* [35] parton distribution functions (PDFs); ISAJET 7.80 [36] is used to generate the physical particle masses. The simplified models are generated with one extra jet in the matrix element using MADGRAPH5 [37], interfaced to PYTHIA [38], with the CTEQ6L1 [39] PDF set; MLM matching [40] is done with a scale parameter that is set to onefourth of the mass of the lightest sparticle in the hardscattering matrix element. Signal cross sections are calculated in the MSSM at next-to-leading order in the strong coupling constant, including the resummation of soft gluon emission at next-to-leading-logarithmic accuracy (NLO þ NLL) [41][42][43][44][45].
The simulated event samples for the SM backgrounds are summarized in Table I. The ALPGEN and MADGRAPH samples are produced with the MLM matching scheme. The ALPGEN samples are generated with a number of partons 0 N parton 5 in the matrix element, except for W þ light-flavored jets which are generated with up to six partons. The Wb " b, Wc " c and Wc cross sections shown are the leading-order values from ALPGEN multiplied by a K factor of 1.2, based on the K factor for light-flavored jets. For the final result, measured cross sections are used for the W=Z þ heavy-flavor-jets samples [55]. The overlap between the heavy-flavored and light-flavored W=Z þ jets samples is removed. The cross section for Z þ jets with 10 GeV < m '' < 40 GeV is obtained by assuming the same K factor as for m '' > 40 GeV. The single-top cross sections are taken from MC@NLO; for the s and t channels, they are listed for a single lepton flavor.
The theoretical cross sections for W þ jets and Z þ jets are calculated with FEWZ [48] with the MSTW2008NNLO [56] PDF set. For the diboson cross sections, MCFM [52] with the MSTW2008NLO PDFs is used. The t " t cross section is calculated with HATHOR 1.2 [47] using MSTW2008NNLO PDFs. The t " t þ W cross section is taken from Ref. [53]. The t " t þ Z cross section is the leading-order value multiplied by a K factor deduced from the NLO calculation at ffiffi ffi s p ¼ 14 TeV [54].
Parton shower and fragmentation processes are simulated for the ALPGEN and MC@NLO samples using HERWIG [51] with JIMMY [57] for underlying event modeling; PYTHIA is used for the ACERMC single-top sample and t " t þ W=Z. The PDFs used in this analysis are CTEQ6L1 for the ALPGEN and MADGRAPH samples, CT10 [58] for MC@NLO, and MRSTMCal (LO**) [59] for HERWIG. The underlying event tunes are the ATLAS AUET2B_LO** tunes [60].
The detector simulation [61] is performed using GEANT4 [62]. All samples are produced with a range of simulated minimum-bias interactions overlaid on the hard-scattering event to account for multiple pp interactions in the same beam crossing (pileup). The overlay also treats the impact TABLE I. Simulated background event samples used in this analysis, with the corresponding production cross sections. The notation LO Â K indicates that the process is calculated at leading order and corrected by a factor derived from the ratio of NLO to LO cross sections for a closely related process. The t " t, W þ light-jets and Z þ light-jets samples are normalized using the inclusive cross sections; the values shown for the W þ light-jets and Z þ light-jets samples are for a single lepton flavor. The single-top cross sections are listed for a single lepton flavor in the s and t channels. Further details are given in the text.

IV. OBJECT RECONSTRUCTION
This analysis is based on three broad classes of event selection: (i) a hard single-lepton channel that is an extension to higher masses of the previous search [16], (ii) a soft single-lepton channel geared towards SUSY models with small mass differences in the decay cascade, and (iii) a multilepton channel aimed at decay chains with higher lepton multiplicities. The event selection requirements are described in detail in Sec. VI. Here the final-state object reconstruction and selection are discussed.

A. Object preselection
The primary vertex [63] is required to be consistent with the beam spot envelope and to have at least five associated tracks; when more than one such vertex is found, the vertex with the largest summed jp T j 2 of the associated tracks is chosen.
Electrons are reconstructed from energy clusters in the electromagnetic calorimeter matched to a track in the ID [64]. Preselected electrons are required to have jj < 2:47 and pass a variant of the ''medium'' selection defined in Ref. [64] that differs mainly in having a tighter trackcluster matching in , stricter pixel hit requirements, additional requirements in the TRT, and tighter shower-shape requirements for jj > 2:0. These requirements provide background rejection close to the ''tight'' selection of Ref. [64] with only a few percent loss in efficiency with respect to medium. Preselected electrons are further required to pass a p T requirement depending on the analysis channel: 10 GeV for the hard-lepton and multilepton channels, and 7 GeV in the soft-lepton channel.
Muons are identified either as a combined track in the MS and ID systems, or as an ID track matched with a MS segment [65,66]. Requirements on the quality of the ID track are identical to those in Ref. [16]. Preselected muons are required to have jj < 2:4 and a p T requirement that depends on the analysis channel: 10 GeV for the hardlepton and multilepton channels, and 6 GeV in the softlepton channel.
Jets are reconstructed using the anti-k t algorithm [67,68] with a radius parameter R ¼ 0:4. Jets arising from detector noise, cosmic rays or other noncollision sources are rejected [69]. To account for the differences between the calorimeter response to electrons and hadrons, p T -and -dependent factors, derived from simulated events and validated with test beam and collision data, are applied to each jet to provide an average energy scale correction [69] back to particle level. Preselected jets are required to have p T > 20 GeV and jj < 4:5. Since electrons are also reconstructed as jets, preselected jets which overlap with preselected electrons within a distance ÁR ¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi ðÁÞ 2 þ ðÁÞ 2 p ¼ 0:2 are discarded.

B. Signal object selection
For the final selection of signal events, ''signal'' electrons are required to pass a variant of the ''tight'' selection of Ref. [64], providing 1%-2% gain in efficiency and slightly better background rejection. Signal electrons must have jj < 2:47 and a distance to the closest jet ÁR > 0:4. They are also required to satisfy isolation criteria: the scalar sum of the p T of tracks within a cone of radius ÁR ¼ 0:2 around the electron (excluding the electron itself) is required to be less than 10% of the electron p T .
Muons in the final selection (signal muons) are required to have jj < 2:4 and ÁR > 0:4 with respect to the closest jet. Further isolation criteria are imposed: the scalar sum of the p T of tracks within a cone of radius ÁR ¼ 0:2 around the muon candidate (excluding the muon itself) is required to be less than 1.8 GeV. The p T requirements for signal electrons and muons depend on the signal regions and are described in Sec. VI.
Signal jets are required to have p T > 25 GeV and jj < 2:5. In addition, they are required to be associated with the hard-scattering process, by demanding that at least 75% of the scalar sum of the p T of all tracks associated with the jet come from tracks associated with the primary vertex of the event. Jets with no associated tracks are rejected. The above requirements are applied to cope with the high pileup conditions of the 2011 data-taking, in particular the later part of the run.
The missing transverse momentum is computed as the negative of the vector sum of the p T of all preselected electrons, preselected muons and preselected jets (after removing those overlapping with preselected electrons), and all calorimeter clusters with jj < 4:9 that are not associated with any of the above-mentioned objects.
For approximately 20% of the 2011 data-taking period, an electronics failure created a region in the electromagnetic calorimeter, located at 0 < < 1:4 and À0:8 < < À0:6, where no signals could be read out. Events with an electron in this region are vetoed for the entire data set, leading to an acceptance loss of less than 1% for signal events in the signal region. For jets, the amount of transverse energy (E T ) lost in the dead region can be estimated from the energy deposited in the neighboring calorimeter cells. If this lost E T projected along the E miss T direction amounts to more than 10 GeV and constitutes more than 10% of the E miss T , the event is rejected. The effect of the electronics failure is described in the detector simulation, and the loss of signal acceptance from this requirement is negligible.
Jets arising from b quarks are identified using information about track impact parameters and reconstructed secondary vertices [70]; the b-tagging algorithm is based on a neural network using the output weights of the JetFitter þ IP3D, IP3D, and SV1 algorithms (defined in Ref. [70]) as input. The b-tagging requirements are set at an operating point corresponding to an average efficiency of 60% for b jets in simulated t " t events, for which the algorithm provides a rejection factor of approximately 200-400 for light-quark and gluon jets (depending on the p T of the jet) and a rejection of approximately 7-10 for charm jets.

V. TRIGGER AND DATA COLLECTION
The data used in this analysis were collected from March through October 2011, during which the instantaneous luminosity of the LHC reached 3:65 Â 10 33 cm À2 s À1 . The average number of interactions per beam crossing ranged from approximately 4 to 16 during the run, with an average of 10. After the application of beam, detector, and data-quality requirements, the total integrated luminosity is 4:7 fb À1 . The uncertainty on the luminosity is determined to be 3.9% [71,72].
Three types of triggers were used to collect the data: electron, muon and E miss T . The electron trigger selects events containing one or more electron candidates, based on the presence of a cluster in the electromagnetic calorimeter, with a shower shape consistent with that of an electron. The transverse energy threshold at the trigger level was either 20 GeV or 22 GeV, depending on the instantaneous luminosity. For signal electrons satisfying p T > 25 GeV, the trigger efficiency is in the plateau region and ranges between 95% and 97%. In order to recover some of the efficiency for high-p T electrons during running periods with the highest instantaneous luminosities, events were also collected with an electron trigger with looser shower-shape requirements but with a p T threshold of 45 GeV.
The muon trigger selects events containing one or more muon candidates based on tracks identified in the MS and ID. The muon trigger p T threshold was 18 GeV. During running periods with the highest instantaneous luminosities, the trigger requirements on the number of MS hits were tightened; in order to recover some of the resulting loss in efficiency, events were also collected with a muon trigger that maintained the looser requirement on the number of chamber hits but that required in addition a jet with p T greater than 10 GeV. This jet requirement is fully efficient for jets with offline calibrated p T greater than approximately 50 GeV. The muon triggers reach their efficiency plateaus below a signal muon p T threshold of 20 GeV. The plateau efficiency ranges from about 70% for jj < 1:05 to 88% for 1:05 < jj < 2:4.
The E miss T trigger bases the bulk of its rejection on the vector sum of transverse energies deposited in projective trigger towers (each with a size of approximately Á Â Á $ 0:1 Â 0:1 for jj < 2:5 and larger and less regular in the more forward regions). A more refined calculation based on the vector sum of all calorimeter cells above threshold is made at a later stage in the trigger processing. The trigger required E miss T > 60 GeV, reaching its efficiency plateau for offline calibrated E miss T > 180 GeV. The efficiency on the plateau is close to 100%.

VI. EVENT SELECTION
Two variables, derived from the kinematic properties of the reconstructed objects, are used in the event selection. The transverse mass (m T ) computed from the momentum of the lepton (') and the missing transverse momentum (p miss T ), defined as is useful in rejecting events containing a single W boson. The inclusive effective mass (m inc eff ) is the scalar sum of the p T of the leptons, the jets and E miss T : where the index i runs over all the signal leptons and j runs over all the signal jets in the event. The inclusive effective mass is correlated with the overall mass scale of the hardscattering process and provides good discrimination against the SM background, without being too sensitive to the details of the SUSY decay cascade. The analysis in Ref. [16] used the three or four leading-p T jets in the calculation of the effective mass; the additional jets used here improve the discrimination between signal and background. A second definition for the effective mass, denoted by m eff , is based on the sum over the two, three, or four leading p T jets, depending on the minimum number of jets required in a given signal region. This variable is used to compute the ratio E miss T =m eff which reflects the fluctuations in the E miss T as a function of the calorimeter activity in the event; the definition used here improves the rejection of the background from mismeasured jets.
This analysis is based on five signal regions, each tailored to maximize the sensitivity to different SUSY event topologies: (1,2) Signal regions requiring a hard lepton plus three or four jets are extensions of the previous analysis [16] to higher SUSY mass scales; these signal regions have been optimized for the MSUGRA/CMSSM model as well as for the bulk of the one-step simplified models with large mass difference (Ám) between the gluino and the LSP; (3) a soft-lepton signal region targets the simplified models with small Ám, where the hard leading jet comes from initial-state radiation (ISR); (4) a multilepton signal region with ! 2 jets is tailored to GMSB models; (5) a multilepton signal region with ! 4 jets is geared towards the two-step simplified models with  Table II.
(1) Hard lepton plus three jets. Events are selected with the electron and muon triggers. The number of signal leptons with p T > 25ð20Þ GeV for electrons (muons) is required to be exactly 1. Events containing additional signal leptons with p T > 10 GeV are rejected. The number of signal jets is required to be ! 3, with a leading jet satisfying p T > 100 GeV and the other jets having p T > 25 GeV. Events with four or more jets are rejected if the fourth jet has p T > 80 GeV; this requirement keeps this signal region disjoint from the four-jet signal region. In addition, the following conditions are imposed: m T > 100 GeV, E miss T > 250 GeV, E miss T =m eff > 0:3, and m inc eff > 1200 GeV. (2) Hard lepton plus four jets. The lepton requirements are the same as in the previous signal region. The number of signal jets is required to be ! 4, with the four leading jets satisfying p T > 80 GeV. In addition, the following requirements are applied: trigger. The number of signal leptons (electron or muon) is required to be exactly 1. Electrons are required to have 7 GeV < p T < 25 GeV, and muons are required to be in the range 6 GeV < p T < 20 GeV. Events containing an additional signal electron (muon) with p T > 7ð6Þ GeV are rejected. The number of signal jets is required to be ! 2, with the leading jet satisfying p T > 130 GeV and the second jet having p T > 25 GeV. In addition, the following conditions are required: m T > 100 GeV, E miss T > 250 GeV, and E miss T =m eff > 0:3. No explicit requirement on m inc eff is applied.
(4) Multilepton plus two jets. Events are selected with the electron and muon triggers. Two or more signal leptons are required, with a leading electron (muon) with p T > 25ð20Þ GeV and subleading leptons with p T > 10 GeV. The two leading leptons must have opposite charge. At least two signal jets with p T > 200 GeV are required. Events with four or more signal jets are rejected if the fourth leading jet has p T > 50 GeV; this requirement keeps this signal region disjoint from the multilepton plus four-jet signal region. In addition the E miss T is required to be >300 GeV. No explicit requirements are made on E miss T =m eff or m inc eff . (5) Multilepton plus four jets. The lepton requirements are the same as in the multilepton plus two jets signal region. At least four signal jets with p T > 50 GeV are required. In addition, the following requirements are imposed: E miss T > 100 GeV, E miss T =m eff > 0:2, and m inc eff > 650 GeV. In contrast to the previous analysis [16], no requirement on the azimuthal angle between the E miss T vector and any of the jets is imposed as the background from multijet events is already low. This adds sensitivity to SUSY decay chains where the LSP is boosted along the jet direction.

VII. BACKGROUND ESTIMATION
The dominant sources of background in the singlelepton channels are the production of semileptonic and fully leptonic t " t events, and W þ jets where the W decays leptonically. For the multilepton channels, the main background sources are Z þ jets and t " t. Other background processes which are considered are multijets, single-top, dibosons and t " t plus vector bosons. The major backgrounds are estimated by isolating each of them in a dedicated control region, normalizing the simulation to data in that control region, and then using the simulation to extrapolate the background expectations  (20) 25 (20) into the signal region. The multijet background is determined from the data by a matrix method described below. All other (smaller) backgrounds are estimated entirely from the simulation, using the most accurate theoretical cross sections available (Table I). To account for the cross contamination of physics processes across control regions, the final estimate of the background is obtained with a simultaneous, combined fit to all control regions, as described in Sec. IX. Several correction factors are applied to the simulation. The p T of the Z boson is reweighted based on a comparison of data with simulation in an event sample enriched in Z þ jets events. The same correction factor is applied to W boson production and improves the agreement between data and simulation in the E miss T distribution. Other correction factors are derived during the combined fit. The relative normalization of the ALPGEN samples (W þ jets, Z þ jets and t " t) with different numbers of partons (N parton ) in the matrix element is adjusted by comparing the jet-multiplicity distributions in data and simulation in all control regions. A common set of corrections is obtained for the W þ jets and Z þ jets samples, and a separate set of Top panels: m inc eff distribution in the W þ jets (left) and t " t (right) control regions for data and simulation for the single hard-lepton channels. Bottom panels: Distribution of the number of jets in the W þ jets (left) and t " t (right) control regions. In all plots, the last bin includes all overflows. The electron and muon channels are combined for ease of presentation. The ''Data/SM'' plots show the ratio between data and the total Standard Model expectation. The expectation for multijets is derived from the data. The remaining Standard Model expectation is entirely derived from simulation, normalized to the theoretical cross sections. The uncertainty band around the Standard Model expectation combines the statistical uncertainty on the simulated event samples with the systematic uncertainties on the jet energy scale, b tagging, data-driven multijet background, and luminosity. The systematic uncertainties are largely correlated from bin to bin. An example of the distribution for a simulated signal is also shown (not stacked); the signal point is chosen to be near the exclusion limit of the analysis in Ref. [16]. A. W=Z þ jets and t t control regions The W þ jets and t " t processes are isolated in control regions defined by the following requirements. For the hard single-lepton channel, ! 3 jets are required, with a leading jet p T > 80 GeV and the other jets above 25 GeV. The lepton requirements are the same as in the signal region. The E miss T is required to be between 40 and 150 GeV while the transverse mass is required to be between 40 and 80 GeV. Furthermore, the m inc eff requirement is relaxed to be >500 GeV. The W þ jets and t " t control regions are distinguished by requirements on the number of b-tagged jets. For the W þ jets control region, events are rejected if any of the three highest p T jets is b tagged; the rejected events then define the t " t control region. Table III summarizes Top panels: E miss T =m eff distribution in the W þ jets (left) and t " t (right) control regions for data and simulation for the soft-lepton channel. Bottom panels: Jet-multiplicity distribution in the W þ jets (left) and t " t (right) control regions. In all distributions, electron and muon channels are combined. The ''Data/SM'' plots show the ratio between data and the total Standard Model expectation. The expectation for multijets is derived from the data. The remaining Standard Model expectation is entirely derived from simulation, normalized to the theoretical cross sections. The uncertainty band around the Standard Model expectation combines the statistical uncertainty on the simulated event samples with the systematic uncertainties on the jet energy scale, b tagging, data-driven multijet background, and luminosity. The systematic uncertainties are largely correlated from bin to bin. An example of the distribution for a simulated signal is also shown (not stacked); the signal point is near the exclusion limit of this analysis.
G. AAD et al.
PHYSICAL REVIEW D 86, 092002 (2012) 092002-8 the control region definitions; Fig. 2 shows the composition of the W þ jets and t " t control regions as a function of m inc eff and of the jet multiplicity. A discrepancy between simulation and data can be seen in the m inc eff distribution. This effect is significantly reduced by the reweighting procedure discussed in Sec. VII B.
For the soft-lepton channel, the control region requirements on the leptons and jets are the same as in the signal region. However, the E miss T is required to be between 180 GeV and 250 GeV and the transverse mass to be between 40 GeV and 80 GeV. The tighter E miss T requirement, compared to the hard single-lepton control regions, is dictated by the trigger selection for this channel. Again, the W þ jets and t " t control regions are distinguished by the presence of b-tagged jets. For W þ jets, events are rejected if any of the two highest p T jets is b tagged; the rejected events form the t " t control region. Figure 3 shows the composition of the W þ jets and t " t control regions for the soft-lepton channel as a function of E miss T =m eff and the jet multiplicity. The significant contamination of t " t background in the W þ jets control sample at high jet multiplicity is taken into account in the combined fit to all backgrounds; uncertainties in the W þ jets background at high jet multiplicity are suppressed Top panels: m inc eff distribution in the Z þ jets (left) and t " t (right) control regions for data and simulation for the multilepton channels. Bottom panels: Distribution of the number of jets in the Z þ jets (left) and t " t (right) control regions; the last bin includes all overflows. The ee and channels are combined for Z þ jets and ee, and e channels are combined for the t " t distributions for ease of presentation. The ''Data/SM'' plots show the ratio between data and the total Standard Model expectation. The expectation for multijets is derived from the data. The remaining Standard Model expectation is entirely derived from simulation, normalized to the theoretical cross sections. The uncertainty band around the Standard Model expectation combines the statistical uncertainty on the simulated event samples with the systematic uncertainties on the jet energy scale, b tagging, data-driven multijet background, and luminosity. The systematic uncertainties are largely correlated from bin to bin. An example of the distribution for a simulated signal is also shown (not stacked); the signal point is chosen to be near the exclusion limit of the analysis in Refs. [82,83].
FURTHER SEARCH FOR SUPERSYMMETRY AT . . . PHYSICAL REVIEW D 86, 092002 (2012) 092002-9 by the fact that t " t becomes the dominant background in this region.
For the multilepton channels, the Z þ jets control region is defined by requiring ! 2 jets with the two leading jets having p T > 80 GeV and 50 GeV, respectively, or with four leading jets having p T > 50 GeV. In addition, E miss T < 50 GeV and an opposite-sign, same-flavor dilepton pair with invariant mass between 81 GeV and 101 GeV are required. The lepton selection requirements are the same as in the signal region. The t " t control region is defined with the same jet requirements as the Z þ jets control region; at least one jet is required to be b tagged. In addition, E miss T between 30 GeV and 80 GeV and a dilepton invariant mass outside the window [81,101] GeV are required. Figure 4 (top panels) shows the composition of the Z þ jets and t " t control regions for the multilepton channel as a function of m inc eff .
B. Reweighting of W þ jets and Z þ jets simulated samples The samples of simulated W þ jets and Z þ jets events are reweighted as a function of the generated p T of the vector boson. A common set of corrections to the p T of the vector boson, applied to both W þ jets and Z þ jets samples, is found to improve the agreement between data and simulation for a number of variables (E miss T , m inc eff , and jet p T ).
The p Z T distribution is measured in data by selecting a sample with two oppositely charged, same-flavor leptons with an invariant mass between 80 GeV and 100 GeV, ! 3 signal jets with p T > 25 GeV, and m inc eff > 400 GeV. The p Z T distribution in five bins of reconstructed p T is compared to the ALPGEN simulation in five bins of generated p T , with the first four bins ranging from 0 to 200 GeV and the last bin integrated above 200 GeV; the ratio of the two distributions is taken as the p Z;gen T -dependent weighting factor. The simulation employed here uses the cross sections listed in Table I. Only the systematic uncertainty from the jet energy scale is considered (in addition to statistical uncertainties) when computing the uncertainty on the weighting factors.

C. Multijet background
Multijet events become a background when a jet is misidentified as an isolated lepton or when a real lepton appears as a decay product of hadrons in jets, for example, from b or c jets, and is sufficiently isolated. In the following, such leptonlike objects are collectively referred to as misidentified leptons. The multijet background in each signal region, and in the W þ jets and t " t control regions, where it is more significant, is estimated from the data following a matrix method similar to that employed in Ref. [16].
The multijet background from all sources (but separated by lepton flavor) is determined collectively. In the single-lepton channels, the multijet process is enhanced in control samples with all the signal region criteria applied but where the lepton isolation criteria are not imposed and the shower-shape requirements on electrons are relaxed. Defining N pass and N fail as the number of events in such a loose sample passing or failing the final lepton selection criteria, and defining N real and N misid as the number of real and the number of misidentified leptons, the following equations hold: where real is the relative identification efficiency for real leptons, and misid is the misidentification efficiency for misidentified leptons. Solving the equations leads to  Overview of the selection criteria for the W þ jets, Z þ jets and t " t control regions (CR). Only the criteria that are different from the signal selection criteria listed in Table II are shown.

Hard lepton
Soft lepton Multilepton 25, 25 >80, 25, 25 >130, 25 >130, 25 >80, 50 or >50, 50, 50, 50 >80, 50 or >50, 50, 50, 50 N jet   The efficiency real is measured from data samples of Z ! '' decays. The lepton misidentification efficiency is obtained as follows. For electrons (muons) with p T > 25ð20Þ GeV misid is estimated with events containing at least one electron (muon) satisfying the relaxed criteria, and at least one signal jet with p T > 30ð60Þ GeV. In addition, for the electron case, E miss T < 30 GeV is required. For the muon case, the event is required to contain exactly one muon with jd 0 j= d 0 > 5, where d 0 and d 0 are the transverse impact parameter and its uncertainty, respectively, measured with respect to the primary vertex. For the softlepton channel, the sample for deriving misid consists of events containing a same-sign and same-flavor lepton pair where the leptons satisfy the relaxed isolation criteria. The selection of a lepton pair allows the low-p T region to be studied with a large data sample. The same-sign requirement reduces the dominance of b hadrons in the sample, providing a better mix of the different mechanisms by which leptons can be misidentified. One of the leptons is required to fail the signal lepton criteria to further enhance the background; the misidentification efficiency is measured with the other lepton. An additional veto around the Z boson mass is applied. In all channels, the electron misidentification efficiency is evaluated separately for samples enhanced (depleted) in heavy-flavor contributions by requiring (vetoing) a b-tagged jet in the event.
Events / 50 GeV For the multilepton channels, the misidentification probabilities as determined above are applied to the number of events where two leptons pass the loose selection criteria. The contribution from processes where one lepton is real and the other misidentified has been studied in both simulation and data, using a generalization of the above matrix method to two leptons. Both methods give similar results; the final estimate is taken from the simulation.

D. Other backgrounds
The backgrounds from single-top, diboson and t " t þ vector boson production are estimated almost purely from simulation, as is the Z þ jets background for the single-lepton channels. The background from cosmic-ray muons overlapping a hard-scattering event is estimated from a control sample with large z 0 , defined as the distance in the z direction with respect to the primary vertex, evaluated at the point of closest approach of the muon to the primary vertex in the transverse plane. The extrapolated contribution to the signal region, jz 0 j < 5 mm, is found to be negligible.

VIII. SYSTEMATIC UNCERTAINTIES
Systematic uncertainties have an impact on the expected background and signal event yields in the control and signal regions. These uncertainties are treated as nuisance parameters in a profile likelihood fit described in Sec. IX. The following systematic uncertainties on the reconstructed objects are taken into account. The jet energy scale (JES) uncertainty has been determined from a combination of test beam, simulation and in situ measurements from 2010 pp collision data [69]. Additional contributions from the higher luminosity and pileup in 2011 are taken into account. Uncertainties on the lepton identification, momentum/energy scale and resolution are estimated from samples of Z ! ' þ ' À , J=c ! ' þ ' À and W AE ! ' AE decays in data [64][65][66]. The uncertainties on the jet and lepton energies are propagated to the E miss T ; an additional E miss T uncertainty arising from energy deposits not associated with reconstructed objects is also included [73]. Uncertainties on the b-tagging efficiency are derived from dedicated data samples [74,75], e.g. containing muons associated with jets. Uncertainties on the lightflavor mistag rate are derived by examining tracks with negative impact parameter [76] while charm mistag uncertainties are obtained from data samples tagged by reconstructing D Ã mesons [77].
Uncertainties in the matrix method for the determination of the multijet background include the statistical uncertainty in the number of events available in the various control samples, the difference in misidentification efficiency for electrons from heavy-versus light-flavored jets, the dependence of the misidentification efficiency on the jet multiplicity, and the uncertainty in the subtraction of other backgrounds from the samples used to estimate the misidentification efficiency.
Uncertainties from the identification efficiency for jets associated with the primary vertex and from the overlay of pileup in simulated events are both found to be negligible.
Theoretical modeling uncertainties in the simulation include the following contributions. In previous versions of the analysis, renormalization and factorization scale uncertainties were estimated by varying the corresponding parameters in the ALPGEN generator by a factor of 2, up and down from their nominal settings. Since these variations affect mostly the overall normalization of the cross sections for the samples with different values of N parton , they are replaced here by a normalization of the individual lightparton bins to the data (see Sec. IX). Additional generator uncertainties arise from the parameter that describes the jet p T threshold used in the matching (p T;min ). This uncertainty is assessed by changing its default value from 15 GeV to 30 GeV; the difference is assigned as both a positive and negative uncertainty. Uncertainties arising from initial-and final-state radiation are taken into account by the variation of the MLM-matching parameter in multileg generators as well as by studying dedicated PYTHIA tunes with increased or decreased radiation [78]. Fragmentation/hadronization uncertainties are estimated by comparing HERWIG with PYTHIA. In order to vary the heavy-flavor fraction, the cross sections for Wb " b þ jets and Wc " c þ jets in Table I are scaled by 1:63 AE 0:76, while Wc þ jets is scaled by 1:11 AE 0:35, based on correction factors derived from data [55]. The uncertainty on Zb " b þ jets is taken to be AE100%. The uncertainties on the cross sections for t " t þ W and t " t þ Z are taken from the NLO calculations in Refs. [53,54].
The uncertainty in the signal cross section is taken from an envelope of cross section predictions using different PDF sets (including the S uncertainty) and factorization and renormalization scales, as described in Ref. [79]. For the simplified models, uncertainties in the modeling of initial-state radiation play a significant role for low gluino masses and for small mass differences in the decay cascade. These uncertainties are estimated by varying generator tunes in the simulation as well as by generator-level studies ofgg andqq production with an additional ISR jet generated in the matrix element with MADGRAPH5.
The impact of these systematic uncertainties on the background yields and signal estimates is evaluated via an overall fit, described in Secs. IX and X.

IX. BACKGROUND FIT
The background in the signal region is estimated with a fit based on the profile likelihood method [80]. The inputs to the fit are as follows: (1) The observed numbers of events in the W þ jets (or Z þ jets in the multilepton channels) and t " t control regions, and the numbers expected from simulation. These are separated into seven jet-multiplicity bins, ranging from three to nine jets for the hard-lepton channel, eight jet-multiplicity bins ranging from two to nine jets for the multilepton channels, and six bins ranging from two to seven jets for the soft-lepton channel. This information is shown in the bottom half of Figs. 2-4. (2) Transfer factors (TF), derived from simulation, which are multiplicative factors that propagate the event counts for W þ jets, Z þ jets and t " t backgrounds from one control region to another, or from one control region to the signal region. Typical values of the TFs from the control to the signal region are 10 À2 and 10 À4 for the soft-and hard-lepton channels, respectively. For each analysis channel (hard-lepton, soft-lepton, multilepton) the event count in each bin of the control region is treated with a Poisson probability density function. The statistical and systematic uncertainties on the expected yields are included in the probability density function as nuisance parameters, constrained to be Gaussian with a width given by the size of the uncertainty. Approximately 150 nuisance parameters are included in the fit.
Correlations in the nuisance parameters from bin to bin are taken into account where necessary. The Poisson probability density functions also include free parameters, for example to scale the expected contributions from the major backgrounds; these are described in more detail below. A likelihood is formed as the product of these probability density functions and the constraints on the nuisance parameters. Each lepton flavor (in the multilepton channel, each combination of flavors of the two leading leptons) is treated separately in the likelihood function.
The free parameters and nuisance parameters are adjusted to maximize the likelihood. An important difference with respect to the analysis in Ref. [16] is the increase in the number of measurements, allowing the fit to be constrained. This has been used in this analysis to constrain the nuisance parameters for the jet energy scale and the uncertainty in the ALPGEN scale parameters from the shape information provided in the control regions. The free parameters considered in the fit are as follows: (1) t " t background: Each t " t sample, divided into N parton bins (from 0 to 3, with the last being inclusive), is scaled by a free parameter. For each N parton bin, a common parameter is used for semileptonic and dileptonic t " t samples. (2) W=Z background: Each W þ jets and Z þ jets sample, again divided into N parton bins from 2 to 5, is scaled by a free parameter. The N parton ¼ 6 bin for W þ light-flavored jets shares its fit parameter with  25, 25 >80, 25, 25 >80, 25, 25 >130, 25 N jet [150,250] [150, 250] [40,250] [180, 250] m T (GeV) [40,80] [ 40,80] >100 [80,100]

Hard lepton Soft lepton
Overview of the selection criteria for the background VR for the multilepton channels. Only the criteria that are different from the signal selection criteria listed in Table II are shown. For the two-jet validation regions, the fourth leading jet (if present) is required to have p T < 50 GeV.
p jet T (GeV) >120, 120 >120, 120 >80, 50, 50, 50 >80, 50, 50, 50 N jet    allowed to float in the fit within their respective uncertainties. Notable nuisance parameters in the fit are as follows: (1) The uncertainty in the ALPGEN MLM-matching parameter p T;min manifests itself in the relative normalization of the ALPGEN N parton samples and in the jet p T spectra within each sample. The change in the event counts in the array of all control regions, resulting from this shift in the relative normalization, is mapped to one parameter for both W þ jets and Z þ jets and a separate parameter for t " t.
(2) The uncertainty in the normalization of the N parton ¼ 0, 1 bins for W þ jets and Z þ jets, due to uncertainties in renormalization and factorization scales, is treated by one nuisance parameter. (3) The overall normalization of the vector boson plus heavy-flavor samples is assigned a nuisance parameter reflecting the uncertainty in the cross section. (4) The uncertainty from the fit of the p Z T distribution is treated by assigning one nuisance parameter for each bin in true p T . Four equal-width bins are used from 0 to 200 GeV, and one bin for p T > 200 GeV. (5) The uncertainty due to the jet energy scale is considered in three jet p T bins (25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35)(36)(37)(38)(39)(40) GeV and >100 GeV). The resulting uncertainty in the event counts in the array of all control regions is mapped to one nuisance parameter for each of the three jet p T bins. The usage of three jet p T bins prevents the fit from artificially overconstraining the jet energy scale.

A. Background fit validation
The background fit is cross-checked in a number of validation regions, situated between the control and signal regions, where the results of the background fit can be compared to observation. These validation regions are not used to constrain the fit. For the single hard-lepton channels, one common set of validation regions, which receives contributions from both three-and four-jet channels, is defined as follows: (1) The W þ jets validation region is identical to the W þ jets control region for the three-jet channel except that the E miss T requirement is changed to 150 GeV < E miss T < 250 GeV (from [40,150] GeV). (2) Similarly, the t " t validation region is identical to the t " t control region for the three-jet channel except for the change in the E miss T requirement to 150 GeV < E miss T < 250 GeV (from [40,150] GeV). TABLE VII. The observed numbers of events in the multilepton signal regions, and the background expectations from the fit. The inputs to the fit are also shown; these consist of the data-driven multijet background estimate and the nominal expectations from simulation (MC), normalized to theoretical cross sections. The errors shown are the statistical plus systematic uncertainties.   The E miss T =m eff distribution in the soft-lepton signal region after all selection requirements except for that on E miss T =m eff . In all plots the different lepton flavors have been combined for ease of presentation. The ''Data/SM'' plots show the ratio between data and the total Standard Model expectation. The Standard Model expectation shown here is the input to the final fit, and is derived from simulation only, normalized to the theoretical cross sections. The uncertainty band around the Standard Model expectation combines the statistical uncertainty on the simulated event samples with the systematic uncertainties on the jet energy scale, b tagging, data-driven multijet background, and luminosity. The systematic uncertainties are largely correlated from bin to bin. An example of the distribution for a simulated signal is also shown (not stacked); the signal point is chosen to be near the exclusion limit of the analysis in Ref. [16].
G. AAD et al.
PHYSICAL REVIEW D 86, 092002 (2012) 092002-16 (3) The high transverse mass validation region is defined by m T > 100 GeV and 40 GeV < E miss T < 250 GeV. This region tests the validity of the background yields from dileptonic t " t events.
For the soft-lepton channel, the validation region is based on the sum of the W þ jets and t " t control regions but with the transverse mass selection changed to 80 GeV < m T < 100 GeV (from [40,80] GeV).
For the multilepton channels, two Z þ jets validation regions and two t " t validation regions are defined: (1) The two-jet Z þ jets validation region is similar to the Z þ jets control region with ! 2 jets, but the leading two jets are required to have p T > 120 GeV (instead of 80 GeV and 50 GeV); the fourth leading jet (if present) is required to have p T < 50 GeV.
(2) The four-jet Z þ jets validation region is similar to the control region with ! 4 jets but the leading jet p T requirement is tightened to p T > 80 GeV (instead of 50 GeV). (3) The two-jet t " t validation region is similar to the t " t control region with ! 2 jets but the leading two jets are required to have p T > 120 GeV (instead of 80 GeVand 50 GeV); the fourth leading jet (if present) is required to have p T < 50 GeV. The E miss T requirement is changed to 100 GeV < E miss T < 300 GeV.
(4) The four-jet t " t validation region is similar to the t " t control region with ! 4 jets but the leading jet p T requirement is tightened to p T > 80 GeV (instead of 50 GeV). The E miss T requirement is tightened to 80 GeV < E miss T < 100 GeV.
In both Z þ jets validation regions the E miss T requirement is tightened to 50 GeV < E miss T < 100 GeV, and the number of b-tagged jets is required to be zero in order to suppress the t " t contamination. The selection requirements for the validation regions are summarized in Tables IV and V for the single-lepton and multilepton channels, respectively.
The results of the fit to the control regions, as well as the comparison of observed versus predicted event counts in the validation regions, are summarized in Fig. 6. The difference between the observed and predicted event counts is normalized by the total (statistical and systematic) uncertainty on the prediction. The agreement between predicted and observed yields is good.

X. RESULTS AND INTERPRETATION
The predicted background in the signal regions and the observed numbers of events are shown in Tables VI and VII. The data are consistent with SM expectations in all signal regions.
The dominant background uncertainty comes from the limited number of events in the background  8 (color online). Expected and observed 95% CL exclusion limits in the MSUGRA/CMSSM model with tan ¼ 10, A 0 ¼ 0 and the sign of taken to be positive. The results are obtained by combining ten signal regions from the hard singlelepton and multilepton channels. The band around the median expected limit shows the AE1 variations, including all uncertainties except theoretical uncertainties on the signal. The dotted lines around the observed limit indicate the sensitivity to AE1 variations on these theoretical uncertainties. The dashed grid shows contours of constant squark (curved lines) and gluino (nearly horizontal lines) masses. The previous limit from ATLAS [16] and the results from the LEP experiments [84] are also shown. . Expected and observed 95% CL exclusion limits in the minimal GMSB model, combining six signal regions from the multilepton channels. The band around the median expected limit shows the AE1 variations, including all uncertainties except theoretical uncertainties on the signal. The dotted lines around the observed limit indicate the sensitivity to AE1 variations on these theoretical uncertainties. The different NLSP regions are indicated. The coNLSP region denotes the region where 1 and' R are nearly mass degenerate. Previous OPAL and ATLAS limits in this model can be found in Refs. [85] and [82,83], respectively. Limits derived from the LEP slepton mass limits [86] are also shown.
FURTHER SEARCH FOR SUPERSYMMETRY AT . . . PHYSICAL REVIEW D 86, 092002 (2012) 092002-17 simulation samples in the signal region. Uncertainties on the jet energy scale and the scale uncertainties for the t " t background at high jet multiplicity are also significant. In the soft-lepton channel, an important contribution comes from the evaluation of the multijet background.
For the signal prediction, the dominant uncertainties at the highest excluded SUSY masses arise from the PDFs (30%-40%) and the JES (10%-20%); the former reflect the uncertainty in the gluon distribution at high values of x. In the simplified models with small mass differences, typical uncertainties from ISR variations are approximately 30%.
Model-independent limits on the visible cross section (i.e. the cross section evaluated inside a given signal region) are derived by including the number of events . Excluded regions at 95% confidence level in the parameter space of one-step simplified models. Top row: gluino pair production withg ! q " q 0AE 1 ! q " q 0 W AE0 1 . Bottom row: squark pair production withq L ! q 0AE 1 ! q 0 W AE0 1 . In the left column, the chargino mass is set to be halfway between the gluino (top) or squark (bottom) and LSP masses. In the right column, the LSP mass is fixed at 60 GeV and the masses of the chargino and gluino (top) or squark (bottom) are varied. The band around the median expected limit shows the AE1 variations, including all uncertainties except theoretical uncertainties on the signal. The dotted lines around the observed limit indicate the sensitivity to AE1 variations on these theoretical uncertainties. The plots are from the combination of the hard and soft single-lepton channels. The numbers indicate the excluded cross section in fb. A smaller excluded cross section implies a more stringent limit.
G. AAD et al.
PHYSICAL REVIEW D 86, 092002 (2012) 092002-18 observed in that region as an input to the fit and deriving an additional parameter, representing the non-SM signal strength (constrained to be non-negative), as the output of the fit. Potential signal contamination in the control regions is ignored. Limits on the number of non-SM events in the signal region, derived using the CL s [81] prescription, are divided by the integrated luminosity to obtain the constraints on the visible cross section. The limits at 95% confidence level (CL) are shown in Table VIII. For excluding specific models of new physics, the fit in the signal regions proceeds in the same way except that in this case the signal contamination in control regions is treated by providing transfer factors from the signal regions to the control regions as further input to the fit. In addition, the likelihood fit makes use of the m inc eff shape information (E miss T =m eff for the soft-lepton channel) in the signal region as a further discriminant. Examples of these distributions are shown in Fig. 7 (the figure shows the distributions summed over lepton flavors, while the fit treats each lepton flavor channel independently). The likelihood is extended to include bin-by-bin m inc eff or E miss T =m eff information by dividing the signal region into several bins of m inc eff or E miss T =m eff . The ten statistically independent hard-lepton and multilepton channels are combined to set limits in the MSUGRA/CMSSM model. For the minimal GMSB model, only the multilepton channels are used. The softlepton channels are used together with the hard-lepton and multilepton channels to set limits in the one-and two-step simplified models.
The limit in the plane of m 1=2 versus m 0 in the MSUGRA/CMSSM model is shown in Fig. 8. The band around the expected limit includes all uncertainties except theoretical uncertainties on the signal prediction while the band on the observed limit indicates the sensitivity to the theoretical uncertainties on the signal. A large improvement in exclusion coverage over the previous analysis [16] can be seen. The simultaneous fit to the ten signal regions and the inclusion of the shapes of the m inc eff distributions increase the expected reach in m 1=2 and m 0 by about 100 GeV, approximately uniformly across the plane. Along the line of equal masses between squarks and gluinos in the MSUGRA/ CMSSM model, masses below approximately 1200 GeV are excluded at 95% CL.
For the minimal GMSB model, the limit in the plane of tan versus Ã is shown in Fig. 9. The exclusion reach is dominated by the dilepton plus two jets channel. Values of Ã below about 50 TeV are excluded at 95% CL for tan < 45, improving on previous constraints.
Exclusion limits in the one-step simplified models are shown in Fig. 10. The figures also show the cross section excluded at 95% CL. The exclusion limits in the two-step simplified models are shown in Fig. 11 for gluino pair production and Fig. 12 for squark pair production. Simplified models with varying chargino mass and twostep simplified models are considered here for the first time The band around the median expected limit shows the AE1 variations, including all uncertainties except theoretical uncertainties on the signal. The dotted lines around the observed limit indicate the sensitivity to AE1 variations on these theoretical uncertainties. The plots are dominated by the multilepton channels. The numbers indicate the excluded cross section in fb. A smaller excluded cross section implies a more stringent limit.
FURTHER SEARCH FOR SUPERSYMMETRY AT . . . PHYSICAL REVIEW D 86, 092002 (2012) in leptonic SUSY searches. For both one-and two-step models, for the case of low LSP masses, gluinos with masses below approximately 900-1000 GeV and squarks with masses below approximately 500-600 GeV are excluded. Squark limits are considerably weaker, primarily due to the lower production cross section. Furthermore in the one-step model, gluinos with mass below 550 GeV are excluded for essentially all values of the LSP mass if the latter is more than 30 GeV smaller than the gluino mass. Care has to be taken when interpreting the simplified model limit in the context of a MSSM scenario, where the mass of the sneutrino is lighter than the mass of the lefthanded slepton, as this can lead to modification of the lepton momenta. . Excluded regions at 95% confidence level in the parameter space of two-step simplified models with squark pair production. Top left panel: both squarks decay viaq L ! q 0AE 1 ! q 0 ' AE L ! q 0 ' AE 0 1 orq L ! q 0AE 1 ! q 0'AE ! q 0 ' AE 0 1 . Top right panel: one squark decays viaq L ! q 0AE 1 ! q 0 ' AE L ! q 0 ' AE 0 1 orq L ! q 0AE 1 ! q 0'AE ! q 0 ' AE 0 1 and the other squark decays viaq L ! q 0 2 ! q' AE'Ç L ! q' AE ' Ç0 1 orq L ! q 0 2 ! q L ! q 0 1 . Bottom panel: both squarks decay viaq L ! q 0AE 1 ! W ðÃÞAE0 2 ! W ðÃÞAE Z ðÃÞ0 1 . The band around the median expected limit shows the AE1 variations, including all uncertainties except theoretical uncertainties on the signal. The dotted lines around the observed limit indicate the sensitivity to AE1 variations on these theoretical uncertainties. The plots are dominated by the multilepton channels. The numbers indicate the excluded cross section in fb. A smaller excluded cross section implies a more stringent limit.
G. AAD et al.

XI. CONCLUSION
A new search with the ATLAS detector for SUSY in final states containing jets, one or more isolated leptons (electron or muon) and E miss T has been presented. Data from the full 2011 data-taking period, corresponding to an integrated luminosity of 4:7 fb À1 , have been analyzed. Singlelepton and multilepton channels are treated in one analysis. A signal region with a soft lepton and soft jets has been introduced to increase the sensitivity to SUSY decay spectra involving small mass differences (''compressed SUSY''), where the sensitivity is improved by a factor of 10-30 compared to the hard-lepton channel. A simultaneous fit is performed to the event yield in multiple signal and control regions and to the shapes of distributions in those regions.
Observations are in good agreement with SM expectations, and constraints have been set on the visible cross section for new physics processes. Exclusion limits have also been extended for the MSUGRA/CMSSM and minimal GMSB models as well as for a number of simplified models. In MSUGRA/CMSSM, squark and gluino masses below approximately 1200 GeV are excluded at 95% CL (for equal squark and gluino masses). In minimal GMSB, values of Ã below about 50 TeV are excluded for tan < 45.
In one-step simplified models (with the chargino mass halfway between the masses of the gluino/squark and LSP) gluinos are excluded for masses below approximately 900 GeV for low values of the LSP mass. Gluinos with mass below 550 GeV are excluded for essentially all values of the LSP mass if the latter is more than 30 GeV smaller than the mass of the gluino. In the one-step simplified model with a fixed LSP mass and varying chargino and gluino (squark) masses, gluinos below approximately 950 GeV are excluded for a wide range of chargino masses; squarks are excluded below 500 GeV, albeit for a narrower range of chargino masses.
A variety of two-step simplified models have been considered. Limits on gluino masses range from about 900 GeV to 1000 GeV, while squark mass limits range from about 500 GeV to 600 GeV, all for low LSP masses.
These results improve significantly on previous constraints.

ACKNOWLEDGMENTS
We thank CERN for the very successful operation of the LHC, as well as the support staff from our institutions without whom ATLAS could not be operated efficiently. We acknowledge the support of ANPCyT, Argentina; YerPhI, Armenia; ARC, Australia; BMWF, Austria;