Search for supersymmetry in events with a photon, a lepton, and missing transverse momentum in pp collisions at √ s = 8 TeV

A search for supersymmetry involving events with at least one photon, one electron or muon, and large missing transverse momentum has been performed by the CMS experiment. The data sample corresponds to an integrated luminosity of 19 . 7 fb − 1 of pp collisions at √ s = 8 TeV, produced at the CERN LHC. No excess of events is observed beyond expectations from standard model processes. The result of the search is interpreted in the context of a general model of gauge-mediated supersymmetry breaking, where the charged and neutral winos are the next-to-lightest supersymmetric particles. Within this model, winos with a mass up to 360 GeV are excluded at the 95% conﬁdence level. Two simpliﬁed models inspired by gauge-mediated supersymmetry breaking are also examined, and used to derive upper limits on the production cross sections of speciﬁc supersymmetric processes.


Introduction
The extension of the standard model (SM) of particle physics through the concept of supersymmetry (SUSY) [1], which imposes a symmetry between fermions and bosons, can offer a solution to some of the issues not accommodated in the SM, such as the existence of dark matter in the universe or the extreme fine tuning required to control radiative corrections to the Higgs boson mass (hierarchy problem) [2][3][4]. The minimal supersymmetric standard model (MSSM) [5][6][7] provides a calculable framework with a fully known particle content, introducing a superpartner for each SM particle. For example, squarks, gluinos, and gravitinos are the SUSY partners of quarks, gluons, and gravitons, respectively. The MSSM has guided the search program for physics beyond the SM at facilities such as the Fermilab Tevatron and CERN LHC. Existing searches have not yet found evidence for SUSY, but a large parameter space of the MSSM remains to be explored.
Within the MSSM, scenarios based on gauge-mediated SUSY breaking (GMSB) [8][9][10][11][12][13][14][15][16][17][18] are of particular interest because of their ability to naturally circumvent the so-called SUSY flavour problem [19]. The framework of general gauge mediation (GGM) [20] offers a clear definition of GMSB and establishes its key aspects. For example, GMSB predicts the gravitino ( G) to be the lightest supersymmetric particle (LSP). The combination of this feature and the weakness of the coupling of G to other MSSM particles E-mail address: cms-publication-committee-chair@cern.ch. has specific consequences in collider phenomenology. Under the assumption that R-parity [6] is conserved, SUSY particles are pairproduced at the LHC. Except for direct LSP pair production, each SUSY particle initiates a decay chain that yields the next-to-lightest supersymmetric particle (NLSP). Branching fraction for the SUSY particle decay involving G is negligible except for the NLSP, leaving the decay of the NLSP to its SM partner and the G as effectively the only gravitino production mechanism. The gravitino escapes detection, leading to missing momentum in the event. The signature of a GMSB signal is thus strongly dependent on the identity of the NLSP. In most GMSB models, the NLSP is taken to be a bino-or wino-like lightest neutralino, where a bino and wino are the superpartners of the SM U(1) and SU(2) gauge fields, respectively. Previous searches for a GMSB signal typically exploited the diphoton signature [21][22][23][24][25][26][27][28][29], in which each of the two bino-like neutralinos decays promptly into a photon and a gravitino. Similar scenarios with nonprompt NLSP decays have also been considered [30,31]. Thus far, no evidence for GMSB SUSY has been observed, resulting in upper limits on the production cross sections given as a function of the SUSY particle masses, the NLSP lifetime, or other model parameters.
This paper presents a search for SUSY with the CMS experiment at the LHC, and targets GGM models with wino-like NLSPs. The data sample corresponds to an integrated luminosity of 19.7 fb −1 of pp collision data collected in 2012 at √ s = 8 TeV. In particular, we study the wino co-NLSP model [32], in which nearly massdegenerate charged and neutral winos are significantly lighter than the other electroweakinos and constitute the lightest SUSY parti-  cles aside from the gravitino. Although the lifetime of the NLSP is effectively a free parameter in GGM phenomenology, a prompt decay of winos is assumed in this analysis. A signature of at least one photon (γ ), one electron or muon ( ), and large missing transverse momentum ( p miss T ) is used in this search. The photon is assumed to be emitted by a neutralino NLSP, and the leptons by either a charged or neutral NLSP decaying to a W or Z boson, respectively. This signature suppresses many SM backgrounds, obviating the need for additional requirements such as associated jet activity. The diagrams in Fig. 1 provide examples of the decay chains studied in this analysis. The present search is sensitive to the direct electroweakino production mode of Fig. 1(a), where the winos are produced without involving coloured SUSY particles, but also to strong production modes such as the gluino ( g) pair-production process shown in Fig. 1(b). Similar searches were conducted by the ATLAS [33] and CMS [34,35]  analyses sees an excess of events over the respective SM predictions. The wino co-NLSP model has also been probed through the signatures of three leptons or two leptons and two jets [37,38], which target the decay of the neutralino NLSP to a gravitino and a Z boson rather than to a gravitino and a photon. None of these analyses observed a significant excess of events over their respective SM predictions.

CMS detector
The central feature of the CMS apparatus is a superconducting solenoid of 6 m internal diameter, providing a magnetic field of 3.8 T. Within the solenoid volume are a silicon pixel and strip tracker, a lead tungstate crystal electromagnetic calorimeter (ECAL), and a brass and scintillator hadron calorimeter (HCAL), each consisting of a barrel and two endcap sections. Muons are measured in gas-ionization detectors embedded in the steel flux-return yoke outside the solenoid. Extensive forward calorimetry complements the coverage provided by the barrel and endcap detectors. A detailed description of the CMS detector, together with a definition of the coordinate system and the relevant kinematic variables, can be found in Ref. [39].
In the barrel section of the ECAL, an energy resolution of about 1% is achieved for unconverted and late-converting photons with transverse energy E T ≈ 10 GeV. The remaining barrel photons have a resolution of about 1.3% up to a pseudorapidity |η| < 1.0, rising to about 2.5% for |η| = 1.4 [40].
The electron momentum is determined by combining the energy measurement in the ECAL with the momentum measurement in the tracker. The momentum resolution for electrons with transverse momentum p T ≈ 45 GeV from Z → e + e − decays ranges from 1.7% for non-showering electrons in the barrel region to 4.5% for showering electrons in the endcaps [41].
Muons are measured in the range |η| < 2.4, with detector elements based on three technologies: drift tubes, cathode strip chambers, and resistive plate chambers. Through the matching of track segments measured in the muon detectors with tracks measured in the tracker, a transverse momentum resolution of 1.3-2.0% is achieved for barrel muons with 20 < p T < 100 GeV. In the endcaps, the resolution increases up to around 6%. The p T resolution in the barrel is better than 10% for muons with transverse momentum up to 1 TeV [42].
Physics objects are defined using the particle-flow (PF) algorithm [43,44], which reconstructs and identifies individual particles through an optimized combination of information from different elements of the CMS detector. The PF candidates are classified as photons, charged hadrons, neutral hadrons, electrons, or muons. Finally, the CMS detector is nearly hermetic, permitting accurate measurements of p miss T .

Data collection and event selection
The search is conducted in the electron-photon (eγ ) and muon-photon (μγ ) channels. The data samples are collected using a dedicated trigger for each channel, as described below. An event is considered to be in the eγ (μγ ) channel if it contains at least one high-energy photon and an electron (muon). Events that simultaneously satisfy the criteria for the two search channels, representing about 0.1% of the selected events, are classified as μγ candidates because muon objects are less often the result of hadron misidentification than are electron objects.
The trigger for the eγ channel requires at least two isolated photon-like objects, with E T thresholds of 36 and 22 GeV for the highest and second-highest E T photon, respectively. The trigger does not veto photon objects that can be matched to a track, allowing events with a photon and an electron to also satisfy the trigger. The μγ channel uses a muon-photon trigger with a p T threshold of 22 GeV for both the photon and muon objects. To ensure a fully efficient trigger and a similar selection efficiency for the two channels, the subsequent analysis requires E T > 40 GeV for the photon and p T > 25 GeV for the electron or muon. With these requirements, the trigger efficiency for the signal models described in Section 7 is found to be 93-98% for both channels, depending on the model and SUSY mass values.
Photon candidates are reconstructed from clusters of energy in the ECAL [40]. The momentum vector of the photon points from the primary pp interaction vertex to the center of the ECAL energy cluster, under the assumption that the photon originates from the primary vertex, which is defined as the vertex with the highest p 2 T of associated tracks. Only photons from clusters in the pseudorapidity range |η| < 1.44 are included in this analysis. These clusters were selected as photon candidates by a set of criteria that are designed to achieve a 90% identification efficiency for true photons. For a cluster to be identified as a photon, its shape must be consistent with that expected from a photon, and the en-ergy detected in the HCAL behind the cluster cannot exceed 5% of the ECAL energy. To further suppress the misidentification of hadrons as photons, a PF-based isolation requirement is imposed. , targeting a 90% efficiency for true muons, utilizes the quality of the track fit, the number of detector hits used in the tracking, and the isolation. The isolation requirement for muons is similar to that for electrons, but uses a larger cone size R = 0.4. Electrons and muons must originate from a primary vertex, with respective distances of closest approach for electrons (muons) of less than 0.2 mm (2 mm) in the transverse plane and < 1 mm (< 5 mm) along the beam direction.
The reconstruction of jets and p miss T is also based on the PF objects. All reconstructed PF candidates are clustered into jets using the anti-k T clustering algorithm [46,47], with a distance parameter of 0.5. Jet objects are used to calculate the H T variable, defined as the scalar p T sum of jets. To be considered in the H T sum, a jet must have a calibrated and pileup-corrected [48] p T value greater than 30 GeV, |η| < 2.5, and be consistent with an origin at the primary vertex [49]. In addition, it must be no closer than R = 0.5 to the photon or lepton candidates. The missing momentum p miss T is given by the negative of the vector p T sum of all PF objects, with jet-energy corrections applied. The magnitude of p miss T is referred to as E miss T .
To suppress the background from final-state radiation events with an on-shell W (Z) boson that decays to νγ ( γ ), the highest-E T photon in an event must be separated by R > 0.8 from the highest-p T electron or muon. Additionally, for the eγ channel, the invariant mass of the electron-photon system is required to differ by more than 10 GeV from the nominal Z boson mass [50], to reduce background from electrons misidentified as photons.
After applying the selection requirements described above, the obtained event yields are compared to expectations from SM background processes. The signal region of interest is defined by ment reduces backgrounds from processes that produce W bosons. Table 1 shows the observed number of events at different stages of the selection process. Because of a higher selection efficiency for muons, after implementing the selection requirements, the number of observed events in the μγ channel is larger than in the eγ channel.

Background estimation
Three sources of SM background are considered: misidentified photons, misidentified leptons, and electroweak backgrounds.

Misidentified-photon background
The background from misidentified photons arises from events in which a photon object does not correspond to a genuine prompt photon. The dominant background processes in this category are Drell-Yan dielectron (qq → γ * → e + e − ) and W (→ ν) + jets production, in which an electron or jet, respectively, is misidentified as a photon. Minor contributions arise from tt events with leptonic top quark decays, for both the eγ and μγ channels. Events with tt production also contribute to the background if a jet is misidentified as a photon. An electron can be misidentified as a photon if it fails to register track seeds due to detector inefficiencies such as non-operational sensors in the tracker. A jet can be misidentified as a photon if a large fraction of its energy is carried by mesons decaying to photons, such as π 0 → γ γ . These two types of background are estimated from data using weighted control samples. The method proceeds in two steps. First, a control sample enriched in particles that are prone to be misidentified as photon candidates, i.e., electrons or neutral hadrons, is selected by inverting certain criteria in the photon identification, while keeping the other selection requirements identical to those for signal candidates. This control sample is called the proxy sample. The second step is to determine the transfer factor N misid /N proxy , where N misid is the estimate of the number of misidentified events in the signal candidate sample and N proxy is the number of events in the proxy sample. The proxy sample is then scaled by the transfer factor. The definition of the proxy sample is tuned to make its kinematic properties similar to those of events with misidentified photons. Thus, this two-step procedure takes the set of misidentified events in a control region where the SUSY signal contribution is expected to be negligible, e.g. in events with small E miss T , and utilizes it to model the distribution of the misidentified background for a given kinematic variable. In particular, from the extrapolation of the observed events in the control region, the method predicts an expectation for the number of events and corresponding kinematic distribution in the signal region. A detailed validation of this background estimation is performed by applying the method to Monte Carlo (MC) simulation samples, and comparing the outcome of this procedure to the known generated MC content. Good agreement is found in all such tests.
The proxy sample for events with an electron-to-photon misidentification is constructed by inverting the electron-seed veto in the photon identification, which turns the photon candidate into an electron proxy. The transfer factor for this proxy sample is determined by counting Z → e + e − decays in a separate control sample, defined by E miss T < 70 GeV. The ratio N misid Z /N proxy Z constitutes the transfer factor, where N misid Z is the number of Z → e + e − events in the control sample with an e + or e − misidentified as a photon, while N proxy Z is the number of Z → e + e − events in the control sample with the proxy condition applied. The control sample for the eγ channel is taken from the data set collected with the same diphoton trigger as the signal candidates, while the sample for the μγ channel is from a data set based on a trigger that requires at least one isolated electron. To ensure that the samples dominantly consist of Z → e + e − decays, events with one high-purity electron object ("tag") are selected from the respective data sets. The photon candidate and the electron-proxy object are called "probes". For each sample of probe candidates, a fit is performed to the invariant mass distribution of the tag-probe system to extract the The "tag-and-probe" method described above [51] is executed in bins of three variables: the transverse momentum of the probe object (p probe T ), the track multiplicity of the primary vertex (N track ), and the number of reconstructed interaction vertices in the event (N vtx ). To account for the correlations in the distributions of the three variables, the dependence of the transfer factor on these quantities is modelled by a three-dimensional parametric function, which is then used to assign an event-by-event weight to the proxy sample. The transfer factor is a decreasing function of p probe T and N track , and an increasing function of N vtx . For a median value of N vtx , its value varies from 0.04 for events with low p probe T and low N track , to 0.007 for high p probe T and high N track . The relative uncertainty in the transfer factor is typically of order 10%, which is dominated by systematic uncertainties such as those arising from the tag-and-probe fitting procedure and the parametrization of the transfer factor. The dependence of the transfer factor on N vtx is approximately linear, with a value that changes from about 0.005 to 0.012 for N vtx values between 5 (low pileup) and 25 (high pileup).
The estimation of the jet-to-photon misidentification background follows the same procedure of defining a proxy sample and scaling it with the transfer factor. The proxy sample for events with a jet-to-photon misidentification is constructed by inverting the requirements on the variable describing the ECAL cluster shape (σ ηη in Ref. [41]) and on one of the isolation variables in the photon selection. The transfer factor for the hadronic-proxy sample is determined through an assessment of the fraction of events with jet-to-photon misidentification among the photon candidates. This fraction is denoted as the "hadron fraction". This measurement is performed in a low-E miss T control sample from a fit to the distribution of σ ηη based on two templates, one representing pure photons and one modelling the events with jet-to-photon misidentification. The fit is performed with photon candidates in muon-photon events, where a very small contamination of misidentified electrons is expected in the photon sample. The pure photon template is obtained from Z → μμγ data by tagging two muons and requiring the three-body μμγ invariant mass to be consistent with the Z boson mass. The template that models events with jet-to-photon misidentification is obtained by inverting the isolation requirement on the signal-photon candidates. The hadron fraction is measured in p T bins of the photon candidate and decreases, in general, as a function of p T . In the eγ channel, its value varies from 0.25 ± 0.03 at p T = 40 GeV to 0.08 ± 0.02 at p T = 120 GeV. In the μγ channel, the corresponding values are 0.30 ± 0.03 and 0.09 ± 0.02. The uncertainties are dominantly due to possible mismodelling of the fit templates. The small difference in the hadron fraction of the photon candidates between the eγ and μγ channels originates from small differences in trigger requirements on the photon object between the diphoton and muon-photon triggers used to select the eγ and μγ data sets.
The p T distribution of the photon objects is multiplied by the hadron fractions determined as described above. In the eγ channel, the estimated p T distribution of misidentified electrons is subtracted first. The resulting distribution provides the estimate of the p T shape for the events with jet-to-photon misidentification. Rather than forming the ratio of this distribution with the p T distribution of the proxy sample, both distributions are parameterized individually by simple analytic functions. The ratio of these two parameterizations constitutes the transfer factor for the jet-tophoton misidentification.

Misidentified-lepton and electroweak backgrounds
The misidentified-lepton and electroweak (EWK) backgrounds are evaluated together, as described below. A misidentified lepton is defined as a reconstructed lepton that does not arise directly from W or Z boson decays, nor from τ decays that originate from a W or Z boson. Misidentified-lepton events arise primarily from the decay of heavy-flavour quarks and from hadrons misidentified as leptons, with other sources such as decays-in-flight constituting a much smaller contribution. Events where both the lepton and photon are misidentified, which constitute up to 30% of the total misidentified-photon background, are already accounted for by the procedure described in Section 4.1. The SM electroweak background is dominated by events with Vγ (V = W, Z) production. In particular, Wγ events have the same signature as signal events: an energetic photon, a lepton, and significant E miss T . The EWK background includes rare multiboson events and events with ttγ production, which we collectively refer to as the "rare EWK" background. Rare EWK events provide only a minor contribution to the overall background but are relevant in the high-E miss T signal region.
Similar to the determination of the misidentified-photon background, proxy samples are formed and scaled by transfer factors to estimate the contribution of misidentified leptons to the signal region. Each event in the proxy sample contains at least one candidate photon and at least one misidentified-lepton proxy, but no candidate lepton. Proxy objects that model misidentified leptons are selected by inverting the isolation condition in the lepton selection. For electrons, the track-cluster matching requirements are also inverted to further enrich the proxy sample with hadronic objects. The calculation of the transfer factor used to evaluate the misidentified-lepton background is described below.
The modelling of the EWK background is based on MC simulation. Samples of Wγ , Zγ , ttγ , and WWγ events, listed in the order of decreasing overall background contributions, are generated with up to two additional partons using the MadGraph 5 1.3 [52] event generator. The pythia 6.4 [53] program is used to describe the parton shower and hadronization. The pythia program is further used to generate samples of WZγ events, which produce an even smaller background contribution than WWγ events. All samples use the CTEQ6L1 [54] parton distribution functions (PDF). Simulated minimum-bias events are overlaid on the main hard-scattering events to simulate pileup. The generated particles are processed through the full CMS detector simulation framework based on the Geant4 [55] package, and are subjected to the same event selection procedure as the data, including the trigger requirements. Differences between simulation and data in the pileup profile, trigger efficiency, and object identification efficiency are corrected by reweighting the MC events by factors that lie within a few percent of unity. The ttγ , WWγ , and WZγ samples are normalized to the integrated luminosity of the data using cross sections calculated with the event generators, which are valid to leading order (LO) in quantum chromodynamics. For the ttγ sample, a next-to-leading order (NLO)-to-LO scale factor of 2.0 [56] is applied to the cross section to account for higher-order contributions.
For the Vγ background, calculated cross sections are used to fix the ratio between the Wγ and Zγ components, but the overall normalization of the combined sample is derived from data to mitigate potential uncertainties in the theoretical calculation. This is accomplished through a two-component template fit describing the Vγ and misidentified-lepton backgrounds. The templates originate from two background samples obtained using the event selection criteria for the Vγ MC sample and for the misidentified-lepton proxy sample. Distributions of the variable φ( , p miss T ) from the two background samples in the control region 40 < E miss T < 70 GeV are employed as templates. The lower bound E miss T = 40 GeV is applied to reduce the contribution of Zγ events. Expected contributions from the misidentified-photon background and rare EWK backgrounds are subtracted from the data distribution before the fit. The fit provides scale factors for the template histograms that are used directly as transfer factors for the Vγ and the misidentified-lepton proxy samples. Besides avoiding a reliance on the value of the theoretical Vγ cross section, which is observed to underestimate the measured production rate of Wγ events [57,58], this method has the benefit that it does not double count the contributions of background events with both a misidentified photon and lepton. This class of events is already accounted for in the misidentified-photon background sample, as mentioned above.  Table 2 summarizes the sources of systematic uncertainty for the background predictions and the signal yields. For each source, the size of the uncertainty is given (in percentage) relative to the number of events in the corresponding background or signal sample. For the background, the size of the uncertainty relative to the total number of background events is also shown. If the relative uncertainties differ significantly among background sam- Table 2 Summary of systematic uncertainties. The third column gives the uncertainty relative to the number of events in the corresponding background or signal sample. The fourth column shows, for the background terms, the uncertainty relative to the total number of background events in the signal region. ples because of statistical fluctuations due to the limited number of events available for the evaluation of the systematic uncertainties, the range from the minimum to the maximum relative uncertainty is shown. The dominant experimental uncertainty for the background prediction is due to the normalization scale factors applied to the rare EWK and Vγ samples. For the rare EWK sample, a 50% uncertainty is assigned as a conservative approximation of the uncertainty in the NLO-to-LO cross section ratio of ttγ production, which is the dominant component in this sample. Also, for the rare EWK sample, we evaluate the uncertainty due to the luminosity determination [59].  Table 2 also lists the systematic uncertainties considered for the signal MC samples that are used for the interpretation of the result of this search. The uncertainties due to the JES and JER are evaluated using the procedure described above. In addition, for the signal samples, uncertainties in the description of initial-state radiation as well as the renormalization scale and PDF [60] are considered.  [200,300], and >300 GeV, respectively). The data are found to be consistent with the background prediction in all regions. Thus no significant excess of events beyond the SM expectation is observed.

Interpretation
The results of the search are interpreted in terms of cross section upper limits on a GMSB model and two distinct simplified models. For each parameter point of the three models, a large number of hard-scattering simulation events are generated. These events are processed with a detailed fast simulation of the CMS detector response [61]. A large number of minimum-bias interactions are superimposed on the hard-scattering process in order to reproduce the pileup profile observed in data. The event selection applied to the simulated signal events is identical to that applied to data, including the trigger requirements. The resulting event yields are weighted by correction factors to account for selection efficiency differences between data and simulation.
For each mass point of the signal models, a 95% confidence level (CL) cross section upper limit is obtained utilizing the "LHCstyle" CL s prescription [62][63][64], which calculates frequentist CL s limits using the one-sided profiled likelihood as a test statistic. The SM background prediction, signal expectation, and observed number of events in each signal-region bin of the eγ and μγ channels as shown in Fig. 4 are combined into one statistical interpretation, turning the analysis into a multichannel counting experiment.

Interpretation in a GMSB model
A GMSB model with wino co-NLSPs [32], which contains both electroweak and strong production as the primary SUSY production process, is examined. All SUSY particles except for the gluino and winos are considered in the limit of very large mass values such that they do not participate in the interactions. In this limit, the lightest chargino and neutralino become purely wino-like. There is no restriction on the decays, but the gluino always undergoes at least a three-body decay and the charged wino decays to a W boson and the gravitino. The neutral wino decays to a gravitino and either a photon or a Z boson, with branching fractions dictated by the weak mixing angle and the wino mass. In the generated scans, the gluino mass (M g ) ranges from 715 to 1415 GeV in 50 GeV steps, and the wino mass (M W ) from 205 GeV to [M g − 10 GeV], also in 50 GeV steps.
The SUSY particle spectra and branching fractions are determined using the SuSpect 2.41 [65] and sdecay 1.3 [66] programs. pythia 6.4 is employed for the SUSY particle generation, decays, and the subsequent parton showers. The cross section for each mass point is determined to NLO accuracy in quantum chromodynamics using the prospino 2.1 [67] program. This cross section result, along with its uncertainty, is used to derive 95% CL exclusion limits on the SUSY particle masses. In this inclusive GMSB model, electroweak production will always take place when the wino is light, independent of the gluino mass. Thus the exclusion curve becomes horizontal for M W ≈ 360 GeV. Note that for this and for all other instances in this paper where a numerical result is quoted for a mass limit, the result is based on the theoretical prediction for the cross section minus its 1σ uncertainty. The expected distributions for signal in Fig. 3 correspond to the GMSB model for the mass point (M g , M W ) = (915, 405) GeV. This mass point has competing contributions from strong and electroweak SUSY production and exhibits a non-trivial behavior in H T as can be seen in Fig. 3(g) and (h).

Interpretation in simplified models
In a simplified model [68][69][70], a limited set of hypothetical particles and decay chains are introduced to describe a given topological signature such as the γ final state studied in this analysis.
The production and decay amplitudes of these particles are parameterized in terms of the particle masses.
The two simplified models considered are denoted the TChiWg and T5Wg models. The TChiWg model is initiated by the direct production of hypothetical particles χ ± and χ 0 , whose decays are restricted to W ± G and γ G, respectively. The gravitino G is nearly massless as in GMSB models. Thus, this model can be identified with electroweak production in the GMSB wino co-NLSP model, depicted by the diagram in Fig. 1(a), differing only in the decay branching fractions. The particles χ ± and χ 0 are therefore identified with gauginos in the remainder of this paper. A mass range of 100 ≤ M χ ≤ 800 GeV is considered, where M χ is the degenerate mass of the gauginos. The generation of events for the T5Wg model, corresponding to the diagram in Fig. 1(b), starts with the pair production of gluinos. The two gluinos undergo three-body decays g → qq χ ± and g → qq χ 0 , followed by the decays of the χ ± and χ 0 as discussed above. The T5Wg samples are generated in a mass region 700 ≤ M g ≤ 1400 GeV and 25 GeV ≤ M χ ≤ [M g − 25 GeV]. No other non-SM particle is involved in either model.
Events for both models are generated with the MadGraph 5 1.3 program, with up to two final-state partons in addition to the hard interaction. The events are then interfaced to pythia 6.4, which is used to describe the SUSY decay chains and parton showers. The neutralino-chargino and the gluino pair production cross sections are calculated to NLO and NLO + NLL (next-to-leading logarithm) accuracy [60], respectively, and used to derive 95% exclusion limits. Fig. 6(a) shows the computed 95% CL cross section upper limit for the TChiWg model as a function of M χ , together with the theoretical cross section. Assuming a 100% branching fraction for χ 0 → γ G, the mass range 100 < M χ < 540 GeV is excluded, where the lower limit corresponds to the lowest M χ included in the scan. Assuming a more physically motivated branching fraction through a rescaling of the theoretical cross section by the weak mixing angle, the exclusion range is 100 < M χ < 340 GeV. The latter result is similar to the limit M W < 360 GeV obtained from the GMSB model with a wino-like NLSP. The production cross section of the T5Wg model is determined solely by M g . Nevertheless, the M g − M χ mass difference affects the H T and E miss T spectra, resulting in nontrivial exclusion-limit contours in the M χ -M g plane. The 95% CL cross section upper limits and exclusion contours for the T5Wg model are shown in Fig. 6(b). For M χ > 200 GeV, pair production of gluinos is excluded for gluino masses below 1 TeV. For 500 < M χ < 700 GeV, gluinos below approximately 1.1 TeV are excluded.

Summary
This paper presents a search for the anomalous production of events with a photon, an electron or muon, and large missing transverse momentum produced in proton-proton collisions at