Standard Model Higgs Discovery Potential of CMS in the H → WW → ‘ν‘ν Channel

The discovery potential of the CMS detector for the Standard Model Higgs boson in the H → WW → ‘ν‘ν channel is assessed using a full detector simulation. Sources of systematic uncertainties as well as methods to determine backgrounds from data are discussed. If the Standard Model Higgs boson has a mass between 150 GeV and 180 GeV, it should be observed with a signiﬁcance of more than 5 σ with a luminosity of about 10 fb − 1 .


Introduction
The Higgs boson decay into two W bosons and subsequently into two leptons (H → WW → ν ν) is expected to be the main discovery channel for the intermediate Higgs-boson mass range, between 2m W and 2m Z [1]. The signature of this decay is characterized by two leptons and high missing energy. In this mass range, the H → WW branching ratio is close to one.
As no narrow mass peak can be reconstructed in this channel, a good background control together with a high signal to background ratio is needed. The most important backgrounds, which give a similar signature as the signal (i.e. two leptons and high missing energy) are continuum WW production and tt production. To reduce these backgrounds, one has to require a small opening angle between the leptons in the plane transverse to the beam and apply a jet veto.
In this note, a study of the CMS discovery potential for this channel is presented. It is based on a full detector simulation. This note concentrates on the Higgs discovery range for Higgs masses between 150 GeV and 180 GeV. The selection cuts are chosen to be mass independent. Three final states are reconstructed: ee, µµ and eµ. In the first part of this note, the signal and background simulation are detailed. Then, the event reconstruction and selection is discussed. Finally, the expected numbers of signal and background events for an integrated luminosity of 1 fb −1 are given and a method to normalize each of the background components is proposed.

Signal and Background Simulation and Reconstruction
The signal samples were generated using the PYTHIA Monte Carlo generator. The two major Higgs-boson production modes relevant for the mass range under investigation were generated: gluon fusion and vector-boson fusion. Table 1 lists the cross section times H → WW branching ratio and the sum of branching ratios for W decaying into e, µ and τ 1) for different Higgs-boson masses. The W boson decay into τ was also simulated but no specific τ reconstruction was applied: τ are selected through their decays into e and µ. The cross sections are shown at leading order (LO), using the PYTHIA predictions, and at next-to-leading order (NLO), using a calculation by M. Spira [2]. In order to get a good NLO estimate for the Higgs-boson production through gluon fusion, the PYTHIA p t spectrum was reweighted to the MC@NLO [3] prediction, defining p t -dependent k-factors as proposed in [4] 2) . The total cross section was then scaled to the NLO cross sections listed in Table 1. Figure 1a) shows the effect of the reweighting on the generated Higgs-boson p t spectrum without any cuts applied, for a Higgs-boson mass of 165 GeV. While the effect of introducing the NLO correction changes the cross sections by about factor 2 for the Higgs produced through gluon fusion, it is almost negligible for the Higgs produced through vector-boson fusion. For the backgrounds, continuum vector-boson production (WW, ZZ, WZ) was generated using PYTHIA. The p t (WW) spectrum was reweighted to the NLO cross section using the same technique as for the signal, as shown 1) No particular decay of τ is assumed.
2) NNLO cross sections for the Higgs-boson production through gluon fusion were used in reference [4]. The present analysis uses NLO cross sections to be consistent with the other processes which are known at NLO. For Higgs-boson production, the NNLO cross sections are slightly larger than the NLO ones. in Figure 1b). A NLO cross section of respectively 16 pb and 77 pb was assumed for ZZ and WZ production. WW production via gluon fusion was generated using a Monte Carlo provided by N. Kauer [5], with parton shower simulation provided by PYTHIA. Figure 2 shows an example of a Feynman graphs for WW production.
Top production (double resonant, tt, and single resonant, tWb) was generated using the TopReX Monte Carlo program [6]. Figure 3 shows two Feynman graphs for the tt and tWb processes. TopReX includes the full spincorrelation structure and an exact treatment of the top mass. The parton shower simulation is provided by PYTHIA. NLO cross sections of respectively 840 pb and 33.4 pb were used for tt and tWb [7]. Table 2 lists the cross sections times branching ratios for the different background processes and compares them with the LO cross section.  The cross sections times branching ratio at LO given by PYTHIA (except for ggWW and tWb) and at NLO for the different background processes [8]. The ggWW cross section is given in reference [5] and is generated using a matrix-element program linked to PYTHIA for the showering. This process is only known at LO. The tWb cross section is from reference [7]. Process , νν, νννν 1.06 1.52

Event Reconstruction
The generated events were passed through a Geant simulation of CMS. Pile-up corresponding to the LHC low luminosity phase was also generated. The events were then reconstructed using the standard CMS software. Table 3 lists the number of signal and background simulated events used in this study together with the Monte Carlo generator version.

Trigger selection
First, the events are required to pass the global Level 1 (L1) trigger. The remaining events have to pass at least one of the following High Level Trigger (HLT) paths: single-electron, double-electron, single-muon or doublemuon trigger. The dotted curve in Figure 4a) shows the combined L1+HLT trigger efficiencies as a function of the Higgs-boson mass. The trigger efficiency is reduced since the W boson decays into τ are simulated but no specific τ trigger is used. The dashed curve shows the L1+HLT trigger efficiency for the events where the W bosons decay into electron and muons.
To estimate the numbers of 'useful events' rejected by this trigger requirement it is interesting to look at the trigger efficiency on events having exactly two leptons which fulfill the lepton selection cuts defined in the following section. In Figure 4a) the solid curve shows the trigger efficiency for such events. In this case, the trigger efficiency is higher than 95% on the full mass range. Figure 4b) shows this trigger efficiency for events with muon-muon, electron-muon and electron-electron final states. The trigger efficiency for muons is close to 100% whereas for electrons it is around 96%. This 4% loss could be recovered.

Lepton selection and reconstruction
Events are required to have exactly two opposite-charge leptons, electrons or muons, with p t > 20 GeV and |η| < 2, since the leptons from signal events are mainly central.
The efficiency to reconstruct a muons candidate with p t > 20 GeV and |η| < 2 close to a generated muon with p t > 20 GeV and |η| < 2 (∆R < 0.15) is 97%, when looking at events accepted by the HLT. Additional isolation requirements are imposed. The energy in the calorimeters around the muon candidate, within a ∆R = 0.3 cone, must be lower than 5 GeV and the sum of the p t of the tracks within a ∆R = 0.25 cone around the muon candidate  must be lower than 2 GeV. This muon selection has an efficiency of 94% for muons in events which pass the HLT trigger. The separate efficiencies of the selection cuts are listed in Table 4a).
Electrons are reconstructed by combining Super Clusters and tracks. A track is associated to a Super Cluster if their distance satisfies ∆R track−SC <0.15. Super Clusters are retained only if E t (SC) > 20 GeV and |η(SC)| < 2. The efficiency to find such a Super Cluster within ∆R = 0.2 of a generated electron with p t > 20 GeV and |η| < 2 is 92% for events that pass the HLT. An electron candidate must then fulfill the following identification requirements: • It deposits little energy in the HCAL: E hcal /E ecal < 0.05 • There is a precise matching of the electron track and cluster: in direction: |η track − η SC corr | < 0.005 and |φ track prop − φ SC | < 0.02 3) in magnitude: E/p > 0.8 and |1/E − 1/p| < 0.02 Finally, the electron candidate must be isolated by requiring tracks p t (track)/E t (SC) < 0.05 where the sum runs on all the tracks which have: This electron selection, applied on Super Clusters with an associated track, a transverse energy higher than 20 GeV and a pseudorapidity smaller than 2, has an efficiency of 87%. The efficiencies of each cut are listed in Table 4b).
Finally a cut on the impact parameter significance σ IP , is applied in order to reduce the bb background. Each lepton is required to have σ IP < 3. The two leptons are also required to come from the same vertex by demanding |z lep1 − z lep2 | < 0.2 cm.
The contribution of reducible background processes, like W+jet where one jet is misidentified as a lepton, is expected to be small.

Jet reconstruction
The inclusive cross section for tt background is at least 40 times larger than the signal cross section. This background can be strongly reduced by applying a jet veto. The reconstruction of jets is thus fundamental to ensure an efficient background rejection. An important issue is to differentiate between 'real' and 'fake' jets. Fake jets are expected to come from the underlying event, pile-up and noise. Their presence tends to affect signal and background in a similar way.
The jets are reconstructed using a cone algorithm of ∆R = 0.5 on calorimeter towers with E tow t > 0.5 GeV and E tow > 0. 8 GeV 4) . With the reconstruction software used for this study, the calibration constants are not valid for a jet transverse energy below 25 GeV. However, the real energy of the jets is not needed here, since we only want to apply a jet veto. Thus the raw energy of the jets is used for this study. For the jet energies relevant for our study, E t (jet) ≈ (1.5 − 2) · E t (raw). In addition the jets are required to be separated from the leptons (∆R jet−lepton > 0.5) to avoid that electrons and Bremsstrahlung photons are misidentified as jets.
The kinematics of the tt process favors central jets, while in the case of the Higgs, relatively more events have high rapidity jets, as shown in Figure 5a) and b). Therefore we apply only a veto on events with central jets. This avoids also potential problems with fake jets in the forward region, as shown in the following.  To estimate the contamination due to fake jets, it is useful to look at the generated jets. Here 'generated jets' means jets from the generator tree without detector simulation but using the same algorithms as the reconstructed jets. A jet is then considered 'real' if it can be matched to a generated jet, having ∆R gen−rec jet < 0.3. Figure 6 shows the fraction of matched jets over the total number of jets as a function of the raw transverse energy of the jet. The dots show the ratio of matched jets over the total number of jets without restricting the jet pseudorapidity and the dots shows the ratio of matched jet over all jets for jets with |η jet | < 2.5. The fraction of fake jets decreases when applying a cut on η. For a raw energy above 20 GeV, the number of fake jets is negligible.
[GeV] GeV for example, this means that around 35% of the reconstructed jets are fake jets, before a cut on |η jet | is applied, and around 18% after the cut is applied.  In order to increase the background rejection, jets with a raw transverse energy smaller than 20 GeV should also be vetoed. However, as Figure 6 shows, the number of jets without an associated generated jet, so-called 'fake' jets, is quite high when the raw energies are below 20 GeV. In order to reduce these fake jets, it is useful to consider the track content of the jets, defining the so-called alpha parameter 5) . Alpha is defined to be the ratio of the sum of p t of all tracks inside the jet divided by the transverse jet energy in the calorimeter: 5) This method is based on an idea proposed first by N. Ilina, V. Gavrilov and A. Krokhotin For a perfect detector, the alpha parameter of a jet would be on average 0.66, because two thirds of a jet are charged particles. This ratio is smeared and reduced by the detector energy resolution and the fact that particles need a minimal energy in order to be detected. In a fake jet, the alpha parameter looks different, as underlying events contain a lot of low p t particles, where the neutral particles leave energy in the ECAL but the charged tracks are curled up in the magnetic field and are then not seen in the tracker. This magnetic field effect leads to an alpha parameter around zero.
To calculate the alpha of a jet, the following tracks are selected: First they have to be 'inside' the jet, i.e. ∆R track−jet < 0.5. Then they have to come from the event vertex 6) , fulfilling |z trk − z vtx | <0.4 cm. Finally, these tracks should have more than 5 hits and p t > 2 GeV. Figure 7 shows the alpha parameter for E raw jet t < 20 GeV a) and E raw jet t > 20 GeV b). For a jet transverse energy above 20 GeV, almost all jets are matched to the generated jets, as shown in Figure 6. Below 20 GeV however, the unmatched jets tend to have an alpha parameter around 0. Therefore, fake jets can be reduced by requiring alpha > 0.2. Figure 8 shows the fraction of matched jets over the total number of jets as a function of E raw t for jets with |η| < 2.5. The dots show this fraction for jets with alpha > 0.2, whereas the triangles show jets without a cut on alpha. The matching efficiency increases by about 10% for jets between 15 and 20 GeV when the cut on alpha is applied. Adding a cut on alpha for jets with a raw transverse energy between 15 and 20 GeV allows to efficiently veto jets, keeping the fake rate low and a reasonable efficiency for the signal. This jet veto corresponds roughly to requiring that no quark or gluon with energy 40 GeV or above is produced 7) .
[GeV]  Figure 8: Fraction of matched jets to a generated jet over all reconstructed jets as a function of E jet T raw for jets built with a constant tower energy cut and requiring |η jet | < 2.5, shown for jets reconstruction from set 2. The dots are the ones where a cut on alpha is applied, whereas on the crosses no cut on alpha is applied.
In summary, for this study, jets are reconstructed using a cone algorithm with a cone size 0.5, using calorimeter towers with raw energy of E tow t > 0.5 GeV and E tow > 0.8 GeV. Currently no calibration is applied on the jet energy. In order to reduce fake jets, we require that jets have |η| < 2.5. and for jets with E raw t between 15 and 20 GeV, it is further required that the sum of transverse momenta of all tracks associated to the jet represents a substantial fraction of the transverse jet energy (alpha > 0.2). For the signal selection, the events containing such jets will be vetoed.

6)
The event vertex is defined as the mean z position of the two leptons. 7) Alternative methods were studied to implement the jet veto. For example, since the jets from tt are b-jets, a possibility would be to veto jets that have tracks with an impact parameter significance higher than 2.

Missing energy reconstruction
The missing energy is reconstructed by summing up the raw energy of all ECAL and HCAL towers, and correcting for muons. Since a jet veto is applied in the signal selection, further corrections to the missing energy did not lead to a significant improvement.

Kinematic Selection and Results
Starting with two oppositely-charged leptons fulfilling the cuts described above, these additional selection cuts were applied: • E miss t >50 GeV (missing transverse energy) • φ < 45 • (angle between the leptons in the transverse plane) • 12 GeV < m < 40 GeV (invariant mass of the two leptons) • No jet with E raw t > 15 GeV and |η| <2.5 (jet veto) • 30 GeV < p max t < 55 GeV (lepton with the maximal p t ) • p min t > 25 GeV (lepton with the minimal p t ) These cuts were optimized to discover a Higgs boson with a mass between 160 and 170 GeV 8) . Each cut value was varied and the value which maximized the signal significance and signal to background ratio was retained. Tables 5, 6 and 7 summarize, for the Higgs-boson signal at different masses and the backgrounds, the number of expected events after each selection cut for 1 fb −1 integrated luminosity. The relative efficiencies are given in parentheses. The last line shows the total selection efficiency together with the error coming from the limited Monte Carlo statistics. The cut on E miss t reduces the contribution from Drell-Yan (pp → γ, Z) events, which has a sharp peak around 15 GeV. It also reduces the WW contribution since the missing energy for this process peaks around 30 GeV, compared to 65 GeV for a 165 GeV Higgs boson, as shown in Figure 9a).
The cuts on φ and m reduce particularly WW continuum production. The cut m < 40 GeV reduces Drell-Yan events but also WW and tt events which have a longer m tail than the signal (Figure 9b)). The cut m < 12 GeV removes potential background from b-resonances. It has a 97% efficiency on a 165 GeV Higgs-boson signal.
The jet veto is mostly efficient against the tt background. A jet veto applied on the jet raw transverse energy at 15 GeV corresponds roughly to a veto on the corrected transverse energy of 25 GeV. For the ggWW background a jet veto is applied. However an efficiency of 1 is conservatively assumed for this cut, since this process is only known at LO [14]. Another possibility for further investigation would be to use the same k-factors as for the signal, since it is also produced with two gluons in the initial state.    Table 8 show the signal to background ratio, the signal significance with 5 fb −1 and the luminosity needed for a 5σ discovery for different Higgs masses. The signal significance is defined as the probability that the observed background, N B , fluctuates above the sum of the signal and background, N S + N B , following a Poisson distribution with mean µ = N B . A signal with more than 5σ statistical significance could already be observed with a luminosity of 5 fb −1 for a Higgs-boson mass between 150 and 180 GeV. In the next section systematic uncertainties on the backgrounds will be added to these results. For Higgs-boson masses between 160 GeV and 170 GeV the signal-to-background ratio is higher than one. The most important background component is continuum WW production. The inclusion of WW produced through gluon fusion increases this background source of about 30%.
The tt production is the dominant contribution from top, while tWb represents less than 15% of the tt contribution. The tt background was generated using TopReX. Two differences were observed when comparing the selection efficiency of this sample with a sample generated with PYTHIA. The efficiency of the φ cut is about 0.14 for the PYTHIA sample whereas it is 0.18 for the TopReX sample, without applying the other kinematic cuts. This is due to the fact that TopReX includes spin correlations in the top production, leading to a slightly different shape in the φ distribution. Figure 13 shows the φ distribution for events with two good reconstructed leptons predicted by TopReX (solid line) and PYTHIA (dotted line). 8) For a specific study of a Higgs with a lower or higher mass in this channel a better efficiency could be obtained by defining mass dependent cuts. This will be done in an upcoming CMS study. The efficiency to keep background events after a jet veto is applied is also about 20% higher in the TopReX sample when no other kinematic cuts are applied. This might be due to the fact that TopReX allows no hard gluon emission from the top whereas PYTHIA does. When data will be available it will be possible to discriminate between these two predictions 9) .
The background contribution from WZ and ZZ is expected to be small, respectively 1.8% and 1.2% of the total background. The Drell-Yan background was not considered in this analysis. It was however checked using a 9) Issues related to top background simulation and systematics are discussed in detail in [11].        2.5 fb −1 sample of pp → γ, Z → µµX (where the tau decays into muons were also simulated) that this background should represent less than 2% of the total background. Table 9 compares the selection cut efficiencies for H → WW decays into two electrons, two muons and one electron and one muon. The W decays into τ were not taken into account here 10) . The trigger efficiency for the two-muon final state, 83%, is higher than for the two electron final state, 73%. Globally, the electron selection efficiency is lower due to the tracker material. The efficiency of the E miss t cut is also higher for the two muon final state, 74%, than for the two electron final state, 69%. This is probably due to the increased activity in the calorimeters when there are two electrons in the final state. The efficiencies of all other cuts are similar. These differences in the selection efficiencies between electron and muon final states are also observed for the different background samples.
10) For a 165 GeV Higgs the WW decays including τ represent only 4% of the total signal. Table 9: The expected number of events for an integrated luminosity of 1 fb −1 for a 165 GeV Higgs boson for the two-electron, two-muon and electron-muon final states. The relative efficiency with respect to the previous cut is given in parentheses. The

Background Normalization and Systematics
At the LHC, it is desirable for the background to be determined from data whenever possible. This is relevant for the present study, which is a 'counting experiment'. In the case where backgrounds can be estimated using a normalization region in the data, the expected number of events for a background in the signal region, N signal region bkg , can be calculated using the following formula: If similar cuts are used to define the signal and the normalization regions, most systematic errors will cancel in the efficiency ratio. Moreover the efficiency ratio is better controlled than the predictions for the single efficiencies.
Such a procedure can be used to determine the tt, WW and WZ background. The other background components, ggWW and tWb, represent a small fraction of the total background. Currently no method has been proposed which allow to isolate them and their systematic uncertainties is determined using the Monte Carlo calculations. The values assumed for the systematic errors from the reconstruction are listed in Table 10.

tt background normalization
Two procedures are proposed to normalize the tt background and are explained in detail in reference [11]. The first one is based on a double b-tagging and the other one requires two hard E t jets in the final state.
A first normalization region is defined by replacing the jet veto in the selection cuts by the requirement of two b-tagged jets in the final state. The other selection cuts are kept unchanged. For the b-tagging, one requires first two jets with E t > 20 GeV and |η| < 2.5. A jet is then b-tagged if it has at least two tracks with an impact parameter significance higher than 2. The contribution from other backgrounds after the double jet veto is applied is expected to be less than 1% of the tt contribution.
A first source of systematics on this background normalization is the theoretical prediction of the ratio between the numbers of events in the signal and in the normalization region. By varying the factorization and normalization scale, this error is expected to be around 5% [12]. The effect of introducing higher order corrections is expected to be small, since the signal region is dominated by LO processes after the jet veto is applied.
The jet energy scale is another source of uncertainty. It is expected to be around 7.5% with an integrated luminosity of 5 fb −1 . The effect of the jet energy scale error on the tt background normalization can be estimated by recording how much the ratio ε sig reg bkg /ε norm reg bkg varies when the jet energy scale is varied 11) . This variation is about 12% for a jet energy scale error of 7.5%. An error of about 4% is expected on the use of the alpha parameter in the jet reconstruction. Finally the b-tagging uncertainty is expected to be around 9% for an integrated luminosity of 5 fb −1 . The statistical uncertainties from the number of events in the normalization region are negligible here. Summing these uncertainties in quadrature, a global uncertainty of 16% is expected for an integrated luminosity of 5 fb −1 .
This b-tagging method is very promising but might be hard to apply on early data as the systematics of b-tagging might need understanding. A second possibility would be to replace the b-tagging by simply requiring two additional hard jets with respectively E jet 1 t > 50 GeV and E jet 2 t > 35 GeV. In this case, in order to avoid a contamination from Drell-Yan background, only eµ final states are considered. The sources of systematics are 5% for the theoretical estimate, 15% for the jet energy scale and 4% for the jet reconstruction, for an overall uncertainty of 16%.
In both procedures, systematic uncertainties are larger than statistical uncertainties. It must be noted that this list of systematics might not be exhaustive once additional experimental and theoretical information will be available. With increasing luminosity, some improvements can be expected.

WW background normalization
A normalization region for WW can be defined by requiring φ < 140 and m > 60 GeV, keeping all other signal selection cuts unchanged. This mass cut eliminates essentially all events with a small φ . In this case and for an integrated luminosity of 5 fb −1 , the error on the theoretical prediction of the ratio between the number of WW events in the signal and in the normalization region is expected to be small compared to the other sources of uncertainty [13]. Only the eµ final state is considered in order to reduce the contribution of Drell-Yan and WZ. Figure 14 shows the φ distributions for the different processes in this normalization region. For an integrated luminosity of 5 fb −1 , about 124 WW events are expected in this signal-depleted region, 88 from tt, and 30 events from the other backgrounds (12 from tWb, 12 from ggWW and 6 from WZ). An additional 28 signal events for a Higgs-boson mass of 165 GeV are expected, compared to 230 events in the signal region.
Systematic uncertainties of 16%, 20%, 22% and 30% can be expected on the tt, WZ, tWb and ggWW backgrounds respectively, as discussed in the previous and following sections. The systematic uncertainty on WW events is given by adding the statistical and systematic error in quadrature: 242 + (88 · 0.16) 2 + (12 · 0.2) 2 + (12 · 0.3) 2 + (6 · 0.2) 2 = 21.5 This represents a relative systematic error of 21.5/124 = 17%. With data, the WW background should be normalized assuming that no signal is present. Therefore, in presence of a signal, the WW background will be overestimated in a first step by about 10%. This estimation can be then refined in an iterative procedure. The statistical uncertainty is the dominant uncertainty on the WW background. With 60 fb −1 , this error should be reduced by a factor of 2.
In a further study, the uncertainty on WW could be reduced by combining more normalization regions. Moreover with a suitable Drell-Yan simulation, the inclusion of ee and µµ final states in the normalization region could increase the statistics, reducing the statistical uncertainty.

WZ background normalization
The WZ background can be normalized by keeping the same selection cuts as in the signal region but requiring an additional lepton in the final state. To gain statistics, the cuts on φ and m are removed. For an integrated luminosity of 5 fb −1 , about 36 WZ events and 14 tt events are expected. Assuming a 16% accuracy on the tt  Figure 14: Distribution of the angle between the leptons in the transverse plane for the signal and the different backgrounds, for an integrated luminosity of 10 fb −1 . The WW normalization region is considered, with all signal cuts applied but m > 60 GeV. Only electron-muon states are kept.
background and an additional 3% for the selection of the additional lepton, an accuracy of about 20% is expected on WZ with 5 fb −1 .

ggWW and tWb normalization
It is difficult to define a normalization region for the ggWW and tWb backgrounds. The Monte Carlo models can be used to predict them, since they represent only a small fraction of the background events. The systematic uncertainty is then given by the following formula: The error on the jet veto and the missing energy was determined by the variation in the global selection efficiency after varying respectively the jet energy by ±2% and the missing energy value by ±2%. The theoretical uncertainty on ggWW is about 30% whereas for tWb it is about 20% [15]. Table 11 summarizes the different sources of systematic errors on these two backgrounds. The theoretical uncertainty are dominant. The ZZ background was not considered in the systematics study since it is expected to make a contribution of only 1.2% to the background in the signal region.

Effect of systematics on the signal significance
The signal significance is degraded by the introduction of systematics uncertainties on the background. This background uncertainty, ∆N B , is taken into account by applying a Gaussian smearing of σ = ∆N B on the number of expected background events [16]. For a large number of events, the significance, taking into account the background uncertainties, can be expressed with the following formula: In this analysis, two sources of background uncertainty are considered: background normalization and section and limited Monte Carlo statistics. Taking into account the sum of the different backgrounds, an overall uncertainty of 10% is found on the total background if only the background systematics are considered. Adding the contribution from limited Monte Carlo statistics, this uncertainty increases to 13%. These results are calculated for an integrated luminosity of 5 fb −1 . For integrated luminosities of 1,2 and 10 fb −1 , the total systematic uncertainties scales to 19%, 16% and 11%, respectively. Figure 12 and Table 8 show the signal significance after the systematic uncertainties are taken into account. For a Higgs-boson mass of 165 GeV, the inclusion of background systematics increases the luminosity needed for a 5σ discovery by a factor 1.8, going from 0.5 fb to 0.9 fb. Due to the high signal-to-background ratio of 1.7, the signal is less sensitive to background fluctuations, leading to this restricted change in the required luminosity for a discovery.
For Higgs-boson masses higher than 180 GeV and lower than 150 GeV, the inclusion of background systematics prevents a 5σ discovery. The reason is that the signal to background ratio is lower than 0.5 in these regions. In this study we concentrated on the mass region around m H = 165 GeV. The selection cuts were optimized for this region. In principle the sensitivity of this channel could be increased by specifically tuning the cuts on the p t of the leptons [1]. Moreover one might also expect that, with more luminosity, the background systematics can be further reduced.

Conclusion
Using a full CMS detector simulation and including background uncertainties, we showed that the Standard Model Higgs boson could be discovered in the H → WW → ν ν channel with less than 1 fb −1 if its mass is around 165 GeV. If it has a mass between 150 and 180 GeV, a 5σ signal could be seen with a luminosity of about 10 fb −1 .