Search for electroweak production of charginos in final states with two tau leptons in pp collisions at sqrt(s) = 8 TeV

Results are presented from a search for the electroweak production of supersymmetric particles in pp collisions in final states with two tau leptons. The data sample corresponds to an integrated luminosity between 18.1 inverse femtobarns and 19.6 inverse femtobarns depending on the final state of tau lepton decays, at sqrt(s) = 8 TeV, collected by the CMS experiment at the LHC. The observed event yields in the signal regions are consistent with the expected standard model backgrounds. The results are interpreted using simplified models describing the pair production and decays of charginos or tau sleptons. For models describing the pair production of the lightest chargino, exclusion limits are obtained in the plane of chargino mass vs. neutralino mass under the following assumptions: the chargino decays into third-generation sleptons, which are taken as the lightest sleptons, and the slepton masses lie midway between those of the chargino and the neutralino. Chargino masses below 420 GeV are excluded at a 95% confidence level in the limit of a massless neutralino, and for neutralino masses up to 100 GeV, chargino masses up to 325 GeV are excluded at 95% confidence level. Constraints are also placed on the cross section for pair production of tau sleptons as a function of mass, assuming a massless neutralino.


Introduction
Supersymmetry (SUSY) [1][2][3][4][5] is one of the most promising extensions of the standard model (SM) of elementary particles. Certain classes of SUSY models can lead to the unification of gauge couplings at high energy, provide a solution to the gauge hierarchy problem without fine tuning by stabilizing the mass of the Higgs boson against large radiative corrections, and provide a stable dark matter candidate in models with conservation of R-parity. A key prediction of SUSY is the existence of new particles with the same gauge quantum numbers as SM particles but differing by a half-unit in spin (sparticles). Extensive searches at the LHC have excluded the existence of strongly produced (colored) sparticles in a broad range of scenarios, with lower limits on sparticle masses ranging up to 1.8 TeV for gluino pair production [6][7][8][9][10][11][12][13]. While the limits do depend on the details -1 -JHEP04(2017)018 Figure 1. Schematic production of τ lepton pairs from chargino (left) or τ slepton (right) pair production. of the assumed SUSY particle mass spectrum, constraints on the colorless sparticles are generally much less stringent. This motivates the electroweak SUSY search described in this paper.
Searches for charginos ( χ ± ), neutralinos ( χ 0 ), and sleptons ( ) by the ATLAS and CMS Collaborations are described in refs. [14][15][16][17][18][19][20]. In various SUSY models, the lightest SUSY partners of the SM leptons are those of the third generation, resulting in enhanced branching fractions for final states with τ leptons [21]. The previous searches for charginos, neutralinos, and sleptons by the CMS Collaboration either did not include the possibility that the scalar τ lepton and its neutral partner ( τ and ν τ ) are the lightest sleptons [16], or that the initial charginos and neutralinos are produced in vector-boson fusion processes [18]. An ATLAS search for SUSY in the di-τ channel is reported in ref. [19], excluding chargino masses up to 345 GeV for a massless neutralino ( χ 0 1 ). The ATLAS results on direct τ production is improved and updated in ref. [20].
In this paper, a search for the electroweak production of the lightest charginos ( χ ± 1 ) and scalar τ leptons ( τ ) is reported using events with two opposite-sign τ leptons and a modest requirement on the magnitude of the missing transverse momentum vector, assuming the masses of the third-generation sleptons are between those of the chargino and the lightest neutralino. Two τ leptons can be generated in the decay chain of χ ± 1 and τ , as shown in figure 1. The results of the search are interpreted in the context of SUSY simplified model spectra (SMS) [22,23] for both production mechanisms.
The results are based on a data set of proton-proton (pp) collisions at √ s = 8 TeV collected with the CMS detector at the LHC during 2012, corresponding to integrated luminosities of 18.1 and 19.6 fb −1 in different channels. This search makes use of the stransverse mass variable (M T2 ) [24,25], which is the extension of transverse mass (M T ) to the case where two massive particles with equal mass are created in pairs and decay to two invisible and two visible particles. In the case of this search, the visible particles are both τ leptons. The distribution of M T2 reflects the scale of the produced particles and has a longer tail for heavy sparticles compared to lighter SM particles. Hence, SUSY can manifest itself as an excess of events in the high-side tail of the M T2 distribution. Final states are considered where two τ leptons are each reconstructed via hadronic decays (τ h τ h ), or where only one τ lepton decays hadronically and the other decays leptonically ( τ h , where is an electron or muon).

JHEP04(2017)018
The paper is organized as follows. The CMS detector, the event reconstruction, and the data sets are described in sections 2 and 3. The M T2 variable is introduced in section 4. The selection criteria for the τ h τ h and τ h channels are described in section 5 and 6, respectively. A detailed study of the SM backgrounds is presented in section 7, while section 8 is devoted to the description of the systematic uncertainties. The results of the search with its statistical interpretation are presented in section 9. Section 10 presents the summaries. The efficiencies for the important selection criteria are summarized in appendix A and can be used to interpret these results within other phenomenological models.

The CMS detector and event reconstruction
The central feature of the CMS apparatus is a superconducting solenoid of 6 m internal diameter that provides a magnetic field of 3.8 T. Within the solenoid volume are a silicon pixel and strip tracker, a lead tungstate crystal electromagnetic calorimeter, and a brass and scintillator hadron calorimeter, each composed 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 more detailed description of the CMS detector, together with a definition of the coordinate system used and the relevant kinematic variables, can be found in ref. [26].
To be recorded for further study, events from pp interactions must satisfy criteria imposed by a two-level trigger system. The first level of the CMS trigger system, composed of custom hardware processors, uses information from the calorimeters and muon detectors to select the most interesting events in a fixed time interval of less than 4 µs. The high-level trigger processor farm further decreases the event rate from around 100 kHz to less than 1 kHz before data storage [27].
The particle-flow (PF) algorithm [28,29] reconstructs and identifies each individual particle with an optimized combination of information from the various elements of the CMS detector. Jets are reconstructed from the PF candidates with the anti-k t clustering algorithm [30] using a distance parameter of 0.5. We apply corrections dependent on transverse momentum (p T ) and pseudorapidity (η) to account for residual effects of nonuniform detector response [31]. A correction to account for multiple pp collisions within the same or nearby bunch crossings (pileup interactions) is estimated on an event-by-event basis using the jet area method described in ref. [32], and is applied to the reconstructed jet p T . The combined secondary vertex algorithm [33] is used to identify ("b tag") jets originating from b quarks. This algorithm is based on the reconstruction of secondary vertices, together with track-based lifetime information. In this analysis a working point is chosen such that, for jets with a p T value greater than 60 GeV the efficiency for tagging a jet containing a b quark is 70% with a light-parton jet misidentification rate of 1.5%, and c quark jet misidentification rate of 20%. Scale factors are applied to the simulated events to reproduce the tagging efficiencies measured in data, separately for jets originating from b or c quarks, and from light-flavor partons. Jets with p T > 40 GeV and |η| < 5.0 and b-tagged jets with p T > 20 GeV and |η| < 2.4 are considered in this analysis.

JHEP04(2017)018
The PF candidates are used to reconstruct the missing transverse momentum vector p miss T , defined as the negative of the vector sum of the transverse momenta of all PF candidates. For each event, p miss T is defined as the magnitude of p miss T . Hadronically decaying τ leptons are reconstructed using the hadron-plus-strips algorithm [34]. The constituents of the reconstructed jets are used to identify individual τ lepton decay modes with one charged hadron and up to two neutral pions, or three charged hadrons. Additional discriminators are used to separate τ h from electrons and muons. Prompt τ leptons are expected to be isolated in the detector. To discriminate them from quantum chromodynamics (QCD) jets, an isolation variable [35] is defined by the scalar sum of the transverse momenta of the charged hadrons and photons falling within a cone around the τ lepton momentum direction after correcting for the effect of pileup. The "loose", "medium", and "tight" working points are defined by requiring the value of the isolation variable not to exceed 2.0, 1.0, and 0.8 GeV, respectively. A similar measure of isolation is computed for charged leptons (e or µ), where the isolation variable is divided by the p T of the lepton. This quantity is used to suppress the contribution from leptons produced in hadron decays in jets.

The Monte Carlo samples
The SUSY signal processes and SM samples, which are used to evaluate potential background contributions, are simulated using CTEQ6L1 [36] parton distribution functions. To model the parton shower and fragmentation, all generators are interfaced with pythia 6.426 [37]. The SM processes of Z+jets, W+jets, tt, and dibosons are generated using the MadGraph 5.1 [38] generator. Single top quark and Higgs boson events are generated with powheg 1.0 [39][40][41][42]. In the following, the events from Higgs boson production via gluon fusion, vector-boson fusion, or in association with a W or Z boson or a tt pair are referred to as "hX." Later on, the events containing at least one top quark or one Z boson are referred to as "tX" and "ZX," respectively. The masses of the top quark and Higgs boson are set to be 172.5 GeV [43] and 125 GeV [44], respectively. Since the final state arising from the pair production of W bosons decaying into τ leptons is very similar to our signal, in the following figures its contribution is shown as an independent sample labeled as "WW." In one of the signal samples, pairs of charginos are produced with pythia 6.426 and decayed exclusively to the final states that contain two τ leptons, two τ neutrinos, and two neutralinos, as shown in figure 1 (left). The daughter sparticle in the two-body decay of the χ ± 1 can be either a τ or ν τ . In this scenario, no decay modes are considered other than those shown in figure 1 (left), so for m( τ ) = m( ν τ ), the two decay chains (via the τ or ν τ ) have 50% branching fraction. The masses of the τ and ν τ are set to be equal to the mean value of the χ ± 1 and χ 0 1 masses and consequently are produced on mass shell. If the τ ( ν τ ) mass is close to the χ 0 1 mass, the τ lepton from the τ ( χ ± 1 ) decay will have a low (high) momentum, resulting in a lower (higher) overall event selection efficiency, producing a weaker (stronger) limit on the chargino mass. In the case where the τ ( ν τ ) mass is close to the χ ± 1 mass, the situations are opposite. Of the scenarios in which the τ slepton and the τ sneutrino have the same mass, the scenario with the highest efficiency overall corresponds to the one in -4 -JHEP04(2017)018 which these masses are half-way between the masses of the χ ± 1 and χ 0 1 . In the other signal sample, pairs of staus are also produced with pythia 6.426, that decay always to two τ leptons and two neutralinos, figure 1 (right). To improve the modeling of the τ lepton decays, the tauola 1.1.1a [45] package is used for both signal and background events.
In the data set considered in this paper, there are on average 21 pp interactions in each bunch crossing. Such additional interactions are generated with pythia and superimposed on simulated events in a manner consistent with the instantaneous luminosity profile of the data set. The detector response in the Monte Carlo (MC) background event samples is modeled by a detailed simulation of the CMS detector based on Geant4 [46]. For the simulation of signal events, many samples of events, corresponding to a grid of χ ± 1 and χ 0 1 mass values, must be generated. To reduce computational requirements, signal events are processed by the CMS fast simulation [47] instead of Geant4. It is verified that the CMS fast simulation is in reasonable agreement with the detailed simulation for our signal which has hadronic decays of tau leptons in the final state. The simulated events are reconstructed with similar algorithms used for collision data.
The yields for the simulated SM background samples are normalized to the cross sections available in the literature. These cross sections correspond to next-to-next-toleading-order (NNLO) accuracy for Z+jets [48] and W+jets [49] events. For the tt simulated samples, the cross section used is calculated to full NNLO accuracy including the resummation of next-to-next-to-leading-logarithmic (NNLL) terms [50]. The event yields from diboson production are normalized to the next-to-leading-order (NLO) cross section taken from ref. [51]. The Resummino [52][53][54] program is used to calculate the signal cross sections at NLO+NLL level where NLL refers to next-to-leading-logarithmic precision.

Definition of M T2
The M T2 variable [24,25] is used in this analysis to discriminate between the SUSY signal and the SM backgrounds as proposed in ref. [55]. This variable has been used extensively by both CMS and ATLAS in searches for supersymmetry [10,19]. The variable was introduced to measure the mass of primary pair-produced particles that eventually decay to undetected particles (e.g. neutralinos). Assuming the two primary SUSY particles undergo the same decay chain with visible and undetectable particles in the final state, the system can be described by the visible mass (m vis(i) ), transverse energy (E vis(i) T ), and transverse momentum ( p vis(i) T ) of each decay branch (i = 1, 2), together with the p miss T , which is shared between the two decay chains. The quantity p miss T is interpreted as the sum of the transverse momenta of the neutralinos, p . In decay chains with neutrinos, p miss T also includes contributions from the p T of the neutrinos.
The transverse mass of each branch can be defined as For a given m χ 0 1 , the M T2 variable is defined as M T2 (m χ 0 1 ) = min . (4.2) -5 -

JHEP04(2017)018
For the correct value of m χ 0 1 , the kinematic endpoint of the M T2 distribution is at the mass of the primary particle [56,57], and it shifts accordingly when the assumed m χ 0 1 is lower or higher than the correct value. In this analysis, the visible part of the decay chain consists of either the two τ h (τ h τ h channel) or a combination of a muon or an electron with a τ h candidate ( τ h channel), so m vis(i) is the mass of a lepton and can be set to zero. We also set m χ 0 1 to zero. The background processes with a back-to-back topology of τ h τ h or τ h are expected from Drell-Yan (DY) or dijet events where two jets are misidentified as τ h τ h or τ h . The resulting M T2 value is close to zero with our choices of m χ 0 1 and m vis(i) , regardless of the values of p miss T and the p T of the τ candidates. This is not the case for signal events, where the leptons are not in a back-to-back topology because of the presence of two undetected neutralinos.

Event selection for the τ h τ h channel
In this channel data of pp collisions, corresponding to an integrated luminosity of 18.1 fb −1 , are used. The events are first selected with a trigger [58] that requires the presence of two isolated τ h candidates with p T > 35 GeV and |η| < 2.1, passing loose identification requirements. Offline, the two τ h candidates must pass the medium τ isolation discriminator, p T > 45 GeV and |η| < 2.1, and have opposite sign (OS). In events with more than one τ h τ h pair, only the pair with the most isolated τ h objects is considered.
Events with extra isolated electrons or muons of p T > 10 GeV and |η| < 2.4 are rejected to suppress backgrounds from diboson decays. Inspired from the MC studies, the contribution from the Z → τ h τ h background is reduced by rejecting events where the visible di-τ h invariant mass is between 55 and 85 GeV (Z boson veto). Furthermore, contributions from low-mass DY and QCD multijet production are reduced by requiring the invariant mass to be greater than 15 GeV . To further reduce Z → τ h τ h and QCD multijet events, p miss T > 30 GeV and M T2 > 40 GeV are also required. The minimum angle ∆φ in the transverse plane between the p miss T and any of the τ h and jets, including b-tagged jets, must be greater than 1.0 radians. This requirement reduces backgrounds from QCD multijet events and W+jets events. After applying the preselection described above, additional requirements are introduced to define two search regions. The first search region (SR1) targets models with a large mass difference (∆m) between charginos and neutralinos. In this case, the M T2 signal distribution can have a long tail beyond the distribution of SM backgrounds. The second search region (SR2) is dedicated to models with small values of ∆m. In this case, the sum of the two transverse mass values, , provides additional discrimination between signal and SM background processes.

JHEP04(2017)018
The veto on events containing b-tagged jets in SR2 reduces the number of tt events, which are expected in the low-M T2 region. Table 1 summarizes the selection requirements for the different signal regions.
6 Event selection for the τ h channel Events in the τ h final states (eτ h and µτ h ) are collected with triggers that require a loosely isolated τ h with p T > 20 GeV and |η| < 2.3, as well as an isolated electron or muon with |η| < 2.1 [58][59][60]. The minimum p T requirement for the electron (muon) was increased during the data taking from 20 to 22 GeV (17 to 18 GeV) due to the increase in instantaneous luminosity. An integrated luminosity of 19.6 fb −1 is used to study these channels.
In the offline analysis, the electron, muon, and τ h objects are required to have p T > 25, 20, and 25 GeV, respectively, and the corresponding identification and isolation requirements are tightened. The |η| requirements are the same as those in the online selections. In events with more than one opposite-sign τ h pair, only the pair that maximizes the scalar p T sum of τ h and electron or muon is considered. Events with additional loosely isolated leptons with p T > 10 GeV are rejected to suppress backgrounds from Z boson decays.
Just as for the τ h τ h channel, preselection requirements to suppress QCD multijet, tt, Z → τ τ , and low-mass resonance events are applied. These requirements are τ h invariant mass between 15 and 45 GeV or > 75 GeV (Z boson veto), p miss T > 30 GeV, M T2 > 40 GeV, and ∆φ > 1.0 radians. The events with b-tagged jets are also rejected to reduce the tt background. The final signal region requirements are M T2 > 90 GeV and M τ h T > 200 GeV . The latter requirement provides discrimination against the W+jets background. Unlike in the τ h τ h channel, events with M T2 < 90 GeV are not used because of the higher level of background.
The summary of the selection requirements is shown in table 1. Figure 2 shows the M T2 distribution after the preselection requirements are imposed. The data are in good agreement with the SM expectations, evaluated from MC simulation, within the statistical uncertainties. A SUSY signal corresponding to high ∆m (m χ ± 1 = 380 GeV, m χ 0 1 = 1 GeV) is used to show the expected signal distribution.

Backgrounds
The backgrounds are studied in two categories: those with "misidentified" τ h , i.e., events where a quark or gluon jet has been misidentified as a τ h , and those with genuine τ h candidates. The QCD multijet and W+jets events are the dominant sources in the first category, while a mixture of tt, Z+jets, diboson, and Higgs boson events dominate the second category. Background estimates are performed using control samples in data whenever possible. Those backgrounds that are taken from simulation are either validated in dedicated control regions or corrected using data-to-simulation scale factors. The estimates of the main backgrounds are discussed below, while the remaining contributions are small and are taken from simulation.

The QCD multijet background estimation in the τ h τ h channel
Events from QCD multijet production can appear in the signal regions if two hadronic jets are misidentified as a τ h τ h pair. The isolation variable is a powerful discriminant between misidentified and genuine τ h candidates. To estimate the QCD multijet contribution, an ABCD method is used, where three τ h τ h control regions (CRs) are defined using the loose τ h isolation requirement, together with lower thresholds on M T2 or ΣM τ i T variables for the corresponding signal region. The former is changed from M T2 > 90 to >40 GeV, whereas the latter is reduced from ΣM τ i T > 250 to >100 GeV . In addition, the requirement on Figure 3. Schematic illustration of three control regions and the signal region used to estimate the QCD multijet background.
∆φ is removed to increase the number of events in the CRs. To reduce contamination from genuine τ h τ h events in CRs with at least one loose τ h candidate, same-sign (SS) τ h τ h pairs are selected. Residual contributions from genuine τ h τ h and W+jets events (non-QCD events) are subtracted based on MC expectations. The CR and signal region are illustrated in figure 3. In the samples dominated by QCD multijet events (CR1 and CR2), the isolation of misidentified τ h candidates is found to be uncorrelated with the search variables M T2 and ΣM τ i T . The QCD multijet background in the signal regions is therefore estimated by scaling the number of QCD multijet events with high M T2 or high ΣM τ i T and loosely isolated SS τ h τ h (CR3) by a transfer factor, which is the y-intercept of a horizontal line fitted to the ratio of the numbers of events in CR1 and CR2 in different bins of the low values of the search variables. The final estimate of the background is corrected for the efficiency of the ∆φ requirement for QCD multijet events. This efficiency is measured in CR1 and CR2, in which the contribution of QCD multijet events is more than 80%. It is checked that the efficiency versus the search variable is same in both CR1 and CR2 and to gain in statistics, two CRs are combined before measuring the efficiency. The efficiency is a falling distribution as a function of the search variable (M T2 or ΣM τ i T ) and the value of the last bin (65 < M T2 < 90 GeV or 200 < ΣM τ i T < 250 GeV) is used conservatively as the value of the efficiency in the signal regions.
The systematic uncertainty in the background estimates includes the uncertainty in the validity of the assumption that isolation and M T2 or ΣM τ i T are not correlated, the -9 -

JHEP04(2017)018
Signal region QCD multijet background estimate Table 2. The estimated QCD multijet background event yields in the τ h τ h channel. The first two uncertainties are the statistical and systematic uncertainties of the method, and the last uncertainty is the extra systematic uncertainty due to the correlation assumptions.
∆φ efficiency is extrapolated correctly to the signal regions, and the uncertainties in the residual non-QCD SM backgrounds which are subtracted based on MC expectations for different components of the background estimation. The latter includes both the statistical uncertainty of the simulated events and also a 22% systematic uncertainty that will be discussed in section 8, assigned uniformly to all simulated events. Table 2 summarizes the estimation of the QCD multijet background contribution in the two signal regions after extrapolation from the control regions and correcting for the ∆φ efficiency. To evaluate the uncertainties in the transfer factor and ∆φ efficiency due to the correlation assumptions, different fit models are examined: (i) a horizontal line or a line with a constant slope is fitted in the distributions of the transfer factor or ∆φ efficiency for 40 < M T2 < 90 GeV in the SR1 case (100 < ΣM τ i T < 250 GeV in the SR2 case); or (ii) the value of the last bin adjacent to the signal region is used. The weighted average of the estimates is compared with the reported values in table 2 to extract the "fit" uncertainty.

W+jets background estimation in the τ h τ h channel
In the τ h τ h channel, the number of remaining events for W+jets from MC is zero, but it has a large statistical uncertainty due to the lack of the statistics in the simulated sample. To have a better estimation, the contribution of the W+jets background in the τ h τ h channel is taken from simulated events, using the formula: Here N SR is the estimation of W+jets events in the signal region, N BFS is the number of W+jets events before applying the final selection criterion (M T2 > 90 GeV for SR1 and ΣM The FS is evaluated in a simulated W+jets sample with a pair of opposite-sign τ h candidates, where the τ h candidates are selected with the same identification requirements as in the signal region, but with looser kinematic selection criteria to improve statistical precision. Additional signal selection requirements on ∆φ or the lepton veto are applied one by one such that two orthogonal subsamples (passing and failing) are obtained. The Signal Region W+jets background estimate τ h τ h SR1 0.70 ± 0.21 (stat) ± 0.09 (syst) ± 0.54 (shape) τ h τ h SR2 4.36 ± 1.05 (stat) ± 1.14 (syst) ± 1.16 (shape) Table 3. The W+jets background estimate in the two search regions. The systematic uncertainty "syst" comes from the maximum variation of the estimation found from varying the τ h energy scale within its uncertainty. The "shape" uncertainty takes into account the difference between the shape of the search variable distribution in data and simulation.
measured FS values from the looser-selection samples are 0.028 ± 0.010 and 0.098 ± 0.032 for SR1 and SR2, respectively. The uncertainty in the τ h energy scale is also taken into account in the uncertainty in FS . The W+jets simulated sample is validated in data using a same-sign µτ h control sample, where both the normalization and FS are checked. The ratio of data to MC expectation is found to be 1.05±0.13 (1.02±0.09) for SR1 (SR2), which is compatible with unity within the uncertainties. For FS , to take into account the difference between the data and MC values, the MC prediction in each of the two signal regions is corrected by the ratio of FS(data) to FS(MC) , which is 0.73 ± 0.57 (1.49 ± 0.38) for SR1 (SR2), and its uncertainty is also taken to be the "shape" systematic uncertainty. Table 3 summarizes the estimated results for different signal regions for the τ h τ h channel.

The Drell-Yan background estimation
The DY background yield is obtained from the MC simulation. The simulated sample includes production of different lepton pairs (ee, µµ, and τ τ ). The contribution from Z → and Z → τ τ → events is found to be very small, because the misidentification probabilities for → τ h are sufficiently low. The dominant background events are Z → τ τ → τ h and Z → τ τ → τ h τ h decays. The misidentification probability for τ h → is also low, so the probability to have DY background contribution from Z → τ τ → τ h τ h events in the τ h channels is negligible. The simulation is validated in a µτ h control region obtained by removing the ∆φ requirement and by inverting the Z boson veto and also by requiring M T2 < 20 GeV, 40 < M τ h T < 100 GeV . The distributions of the invariant mass of the µτ h system for data and simulated events are in good agreement. The p T of the Z boson system, which is correlated with M T2 , is also well reproduced in simulation. Table 4 summarizes the DY background contribution in the different signal regions. For τ h channels, only the contributions from the genuine lepton+τ h are reported. A separate method is developed in section 7.4 to estimate the misidentified lepton contamination in these channels. The systematic uncertainties of the DY background are discussed in detail in section 8.

Misidentified τ h in the τ h channels
The contribution from misidentified τ h in the τ h channels is estimated using a method which takes into account the probability that a loosely isolated misidentified or genuine τ h passes the tight isolation requirements. If the signal selection is done using the τ h JHEP04(2017)018 candidates that pass the loose isolation, the number of loose τ h candidates (N l ) is: where N g is the number of genuine τ h candidates and N m is the number of misidentified τ h candidates. If the selection is tightened, the number of tight τ h candidates (N t ) is where r g (r m ) is the genuine (misidentified τ h ) rate, i.e., the probability that a loosely selected genuine (misidentified) τ h candidate passes the tight selection. One can obtain the following expression by eliminating N g : Here, the product r m N m is the contamination of misidentified τ h candidates in the signal region. This is determined by measuring r m and r g along with the number of loose τ h candidates (N l ) and the number of tight τ h candidates (N t ).
The misidentification rate (r m ) is measured as the ratio of tightly selected τ h candidates to loosely selected τ h candidates in a sample dominated by misidentified τ h candidates. This is done in a data sample with the same selection as τ h , except with an inverted p miss T requirement, i.e., p miss T < 30 GeV . The misidentification rate is measured to be 0.54 ± 0.01. The genuine τ h candidate rate (r g ) is estimated in simulated DY events; it is found to be r g = 0.766 ± 0.003 and almost independent of M T2 . A relative systematic uncertainty of 5% is assigned to the central value of r g to cover its variations for different values of M T2 . The method is validated in the simulated W+jets sample using the misidentification rate which is evaluated with the same method as used for data. This misidentification rate is r m = 0.51. This difference is taken as the systematic uncertainty of 5% in the central value of the misidentification rate (r m = 0.54). The method predicts the number of τ h background events in this sample within the uncertainties. These include statistical uncertainties due to the number of events in the sidebands (loosely selected τ h candidates), as well as systematic uncertainties. The uncertainties in the misidentification rate and the genuine τ h candidate rate are negligible compared to the statistical uncertainties associated to the control regions.
The estimates of the misidentified τ h contamination in the two τ h channels are summarized in table 5. The relative statistical and systematic uncertainties are reported separately. Since the same misidentified and genuine τ h candidate rates are used to estimate  Table 5. Estimation of the misidentified τ h contribution in the signal region of the τ h channels. The total systematic uncertainty is the quadratic sum of the individual components. All uncertainties are relative. The r m (r g ) is shorthand for misidentified (genuine) τ h candidate rate.
the backgrounds for both the eτ h and µτ h channels, the total systematic uncertainties are considered fully correlated between the two channels. The numbers of misidentified events (3.30 for the eτ h channel and 8.15 for the µτ h channel) are consistent within the statistical uncertainties in our control samples.

Systematic uncertainties
Systematic uncertainties can affect the shape or normalization of the backgrounds estimated from simulation (tt, Z+jets, diboson, and Higgs boson events), as well as the signal acceptance. Systematic uncertainties of other background contributions are described in sections 7.1, 7.2 and 7.4. The uncertainties are listed below, and summarized in table 6.
• The energy scales for electron, muon, and τ h objects affect the shape of the kinematic distributions. The systematic uncertainties in the muon and electron energy scales are negligible. The visible energy of τ h object in the MC simulation is scaled up and down by 3%, and all τ h -related variables are recalculated. The resulting variations in final yields are taken as the systematic uncertainties. They are evaluated to be 10-15% for backgrounds and 2-15% in different parts of the signal phase space.
• The uncertainty in the τ h identification efficiency is 6%. The uncertainty in the trigger efficiency of the τ h part of the eτ h and µτ h (τ h τ h ) triggers amounts to 3.0% (4.5%) per τ h candidate. A "tag-and-probe" technique [61] on Z → τ τ data events is used to estimate these uncertainties [35].
• The uncertainty in electron and muon trigger, identification, and isolation efficiencies is 2% [35].
• The uncertainty due to the scale factor for the b-tagging efficiency and misidentification rate is evaluated by varying the factors within their uncertainties. The yields of signal and background events are changed by 8% and 4%, respectively [33].
• To evaluate the uncertainty due to pileup, the measured inelastic pp cross section is varied by 5% [62], resulting in a change in the number of simulated pileup interactions. The relevant efficiencies for signal and background events are changed by 4%.
• The uncertainty in the signal acceptance due to parton distribution function (PDF) uncertainties is taken to be 2% from a similar analysis [16] which follows the PDF4LHC recommendations [63].

JHEP04(2017)018
• The uncertainty in the integrated luminosity is 2.6% [64]. This affects only the normalization of the signal MC samples. Because for the backgrounds either control samples in data are used or the normalization is measured from data.
• The uncertainty in the signal acceptance associated with initial-state radiation (ISR) is evaluated by comparing the efficiencies of jet-related requirements in the Mad-Graph+pythia program. Using the SM WW process, which is expected to be similar to chargino pair production in terms of parton content and process, a 3% uncertainty in the efficiency of b-tagged jets veto and a 6% uncertainty in the ∆φ requirement are assigned.
• The uncertainties related to p miss T can arise from different sources, e.g. the energy scales of lepton, τ h , and jet objects, and unclustered energy. The unclustered energy is the energy of the reconstructed objects which do not belong to any jet or lepton with p T > 10 GeV . The effect of lepton and τ h energy scales is discussed above. The contribution from the uncertainty in the jet energy scale (2-10% depending on η and p T ) and unclustered energy (10%) is found to be negligible. A conservative value of 5% uncertainty is assigned to both signal and background processes based on MC simulation studies [16, 18].
• The performance of the fast detector simulation has some differences compared to the full detector simulation, especially in track reconstruction [18] that can affect the τ h isolation. A 5% systematic uncertainty per τ h candidate is assigned by comparing the τ h isolation and identification efficiency in the fast and full simulations.
• The statistical uncertainties due to limited numbers of simulated events also contributes to the overall uncertainties. This uncertainty amounts to 3-15% for the different parts of the signal phase space and 13-70% for the backgrounds in different signal regions.
• For less important backgrounds like tt, dibosons, and Higgs boson production, the number of simulated events remaining after event selection is very small. A 50% uncertainty is considered for these backgrounds to account for the possible theoretical uncertainty in the cross section calculation as well as the shape mismodeling.
The systematic uncertainties that can alter the shapes are added in quadrature and treated as correlated when two signal regions of the τ h τ h channel are combined. Other systematic uncertainties of these two channels and all of the systematic uncertainties of the τ h channels are treated as uncorrelated.

Results and interpretation
The observed data and predicted background yields for the four signal regions are summarized in table 7. There is no evidence for an excess of events with respect to the predicted SM values in any of the signal regions. In SR2, two events are observed while 7.07 events -14 -

JHEP04(2017)018
Systematic uncertainty source Background (%)  Table 6. Summary of the systematic uncertainties that affect the signal event selection efficiency, DY and rare backgrounds normalization and their shapes. The sources that affect the shape are indicated by (*) next to their names. These sources are considered correlated between two signal regions of the τ h τ h analysis in the final statistical combination.
are expected. The dominant background source is W+jets events. As a cross-check, data and the prediction in the sideband (200 < ΣM τ i T < 250 GeV) are studied: 13 events are observed with an expectation of 17.1 ± 5.0 (stat+syst) events. This result indicates that the difference between the observed and predicted event yields in SR2 can be attributed to a downward fluctuation in the data. Figure 4 compares the data and the SM expectation in four search regions. The top row shows the M T2 distributions in the τ h channels. In these plots, the QCD multijet, W+jets, and misidentified lepton contribution from other channels are based on the estimate described in section 7.4 and labeled as W+jets. The bottom row shows the M T2 and ΣM τ i T distributions in the two different signal regions of the τ h τ h channel. The QCD multijet contribution in these plots is obtained using control samples in data, as described in section 7.1. The W+jets contribution in the last bin of the bottom plots is described in section 7.2, while the contribution to other bins is based on simulated events. The uncertainty band in these four plots includes both the statistical and systematic uncertainties.
There is no excess of events over the SM expectation. These results are interpreted in the context of a simplified model of chargino pair production and decay, which is described in section 3 and corresponds to the left diagram in figure 1. 2.14 ± 0.08 ± 0.38 2.16 ± 0.08 ± 0.39 4.10 ± 0.10 ± 0.90 1.10 ± 0.05 ± 0.27 SUSY(240, 40) 1.43 ± 0.19 ± 0.21 0.96 ± 0.14 ± 0.14 4.35 ± 0.27 ± 0.91 3.60 ± 0.25 ± 0.83 SUSY(180, 60) 0.12 ± 0.04 ± 0.02 0.04 ± 0.02 ± 0.01 0.73 ± 0.11 ± 0.17 2.36 ± 0.17 ± 0.54 Table 7. Data yields and background predictions with uncertainties in the four signal regions of the search. The uncertainties are reported in two parts, the statistical and systematic uncertainties, respectively. The W+jets and QCD multijet main backgrounds are derived from data as described in section 7; the abbreviation "VV" refers to diboson events. The yields for three signal points representing the low, medium, and high ∆m are also shown. SUSY(X, Y) stands for a SUSY signal with m χ ± 1 = X GeV and m χ 0 1 = Y GeV.
A modified frequentist approach, known as the LHC-style CL s criterion [65][66][67], is used to set limits on cross sections at a 95% confidence level (CL). The results on the excluded regions are shown in figure 5. Combining all four signal regions, the observed limits rule out χ ± 1 masses up to 420 GeV for a massless χ 0 1 . This can be compared to the ATLAS limit of 345 GeV for a massless χ 0 1 [19]. It should be noted that the ATLAS results are based on the τ h τ h channel alone. Figure 6 shows the results in the τ h τ h channel, where the χ ± 1 masses are excluded up to 400 GeV for a massless χ 0 1 . In the whole region, the observed limits are within one standard deviation of the expected limits.
The results are also interpreted to set limits on τ τ production, which corresponds to the right diagram in figure 1. In this simplified model, two τ particles are directly produced from the pp collision and decay promptly to two τ leptons and two neutralinos. The effect of the two τ h channels are found to be negligible and therefore are not considered. To calculate the production cross section, τ is defined as the left-handed τ gauge eigenstates [54]. Since the cross section for direct production of sleptons is lower, no point is excluded and a 95% CL upper limit is set on the cross section as a function of the τ mass. Figure 7 displays the ratio of the obtained upper limit on the cross section and the cross section expected from SUSY (signal strength) versus the mass of the τ particle, with the χ 0 1 mass set to 1 GeV . The observed limit is within one standard deviation of the expected limit. The best limit, which corresponds to the lowest signal strength, is obtained for m τ = 150 GeV. The observed (expected) upper limit on the cross section at this mass is 43 (56) fb which is almost two times larger than the theoretical NLO prediction.

Summary
A search for SUSY in the τ τ final state has been performed where the τ pair is produced in a cascade decay from the electroweak production of a chargino pair. The data analyzed were from pp collisions at √ s = 8 TeV collected by the CMS detector at the LHC corresponding  to integrated luminosities between 18.1 and 19.6 fb −1 . To maximize the sensitivity, the selection criteria are optimized for τ h τ h (small ∆m), τ h τ h (large ∆m), and τ h channels using the variables M T2 , M τ h T , and ΣM τ i T . The observed number of events is consistent with the SM expectations. In the context of simplified models, assuming that the third generation sleptons are the lightest sleptons and that their masses lie midway between that of the chargino and the neutralino, charginos lighter than 420 GeV for a massless neutralino are excluded at a 95% confidence level. For neutralino masses up to 100 GeV, chargino masses up to 325 GeV are excluded at a 95% confidence level. Upper limits on 95% CL upper limit on cross section (pb) Figure 5. Expected and observed exclusion regions in terms of simplified models of chargino pair production with the total data set of 2012. The triangle in the bottom-left corner corresponds to τ masses below 96 GeV, which has been excluded by the LEP experiments [68]. The expected limits and the contours corresponding to ±1 standard deviation from experimental uncertainties are shown as red lines. The observed limits are shown with a black solid line, while the ±1 standard deviation based on the signal cross section uncertainties are shown with narrower black lines. the direct τ τ production cross section are also provided, and the best limit obtained is for the massless neutralino scenario, which is two times larger than the theoretical NLO cross sections.

Acknowledgments
We congratulate our colleagues in the CERN accelerator departments for the excellent performance of the LHC and thank the technical and administrative staffs at CERN and at other CMS institutes for their contributions to the success of the CMS effort. In addition, we gratefully acknowledge the computing centres and personnel of the Worldwide LHC Computing Grid for delivering so effectively the computing infrastructure essential to our analyses. Finally, we acknowledge the enduring support for the construction and operation of the LHC and the CMS detector provided by the following funding agencies: the Austrian

A Additional information for new model testing
In the previous sections, a simplified SUSY model is used to optimize the selection criteria and interpret the results. Here, the main efficiencies versus generated values are reported, so that these results can be used in an approximate manner to examine new models in a MC generator-level study. The number of the passed signal events and its uncertainty that can be evaluated by a generator-level study should be combined statistically with the results in table 7 to find the upper limit on the number of signal events and decide if a model is excluded or still allowed according to the analysis presented in this paper.
Efficiencies are provided as a function of the kinematic properties (e.g., p T ) of visible τ lepton decay products at the generator level. The visible τ lepton (τ vis ), if it decays leptonically, is defined as the 4-vector of the light charged lepton. In hadronic decays, τ vis is the difference between the 4-vector of the τ lepton and neutrino in the hadronic decay. The visible τ objects are required to pass the offline kinematic selection criteria (η and p T requirements). The p gen T / variable is defined as the magnitude of the negative vector sum of the τ vis pairs in the transverse plane. The 4-vector of the τ vis objects and p gen T / are used to calculate the M T of the τ vis objects and also the generator-level M T2 . All efficiencies are derived using the SUSY chargino pair production sample. The chargino mass is varied from 120 to 500 GeV and the neutralino mass from 1 to 500 GeV . Table 8 shows the efficiencies for selecting a lepton or τ h for different channels versus p T (τ vis ). These efficiencies include the scale factors, and efficiencies of object identification, isolation, and trigger. Table 9 shows the efficiencies in all channels to pass the p miss T > 30 GeV requirement as a function of the p gen T / . Table 10 shows the efficiencies in different channels to pass the requirement of the reconstructed invariant mass versus the invariant mass of the τ vis pair (generated mass). The requirements on the invariant mass of the reconstructed pair are (>15 GeV) and (<45 or >75 GeV) for the τ h channels and (<55 or >85 GeV) for the τ h τ h channel. The efficiencies of the (M T2 > 90 GeV) requirement in τ h signal region and τ h τ h SR1 are listed in table 11. In the τ h τ h SR2, the reconstructed M T2 is constrained to lie between 40 and 90 GeV. Table 13 shows the efficiencies in τ h τ h SR2 to pass the 40 < M T2 < 90 GeV requirement versus generated M T2 . The last selection in this channel is the requirement on ΣM τ i T , which is calculated using the 4-vector of the two τ vis and p gen T / . Table 14 shows the efficiencies in τ h τ h SR2 to pass the ΣM τ i T > 250 GeV requirement versus generated ΣM τ i T . To take into account the inefficiencies and misidentifications for charge reconstruction of the objects, identification of the b-tagged jets, identification of the extra leptons and the minimum angle between the jets and E miss T in the transverse plane, the final yields in τ h and τ h τ h channels must be multiplied by 0.8 and 0.7, respectively.
To use these efficiencies, one needs to multiply the values one after another and combine statistically the final value with the values reported in table 7 statistically, to decide if a signal point is excluded. At the generator level, a pair of τ h or τ h τ h is selected, when the τ vis objects pass the corresponding offline kinematic selection criteria. . Zero for the efficiency shows the region that the generated M T2 is much greater than the selection cut.

JHEP04(2017)018
The efficiencies are used to reproduce the yields in the SMS plane. The results are in agreement with the yields from the full chain of simulation and reconstruction within ∼30%. A user of these efficiencies should be aware that some assumptions can be broken close to the diagonal (very low mass difference between chargino and neutralino) and these efficiencies cannot be used. This compressed region requires a separate analysis, because the mass difference of the parent particle and its decay products is comparable to the energy threshold used in this analysis to select the objects.  Open Access. This article is distributed under the terms of the Creative Commons Attribution License (CC-BY 4.0), which permits any use, distribution and reproduction in any medium, provided the original author(s) and source are credited.

JHEP04(2017)018
[7] CMS collaboration, Search for supersymmetry in hadronic final states with missing transverse energy using the variables α T and b-quark multiplicity in pp collisions at √ s = 8