Search for Higgs bosons decaying to WW in e+e- collisions at LEP

A search for Higgs bosons produced in association with a fermion pair, and decaying to WW, is performed with the data collected by the ALEPH detector at centre-of-mass energies ranging from 191 to 209 GeV. The data correspond to an integrated luminosity of 453.2 pb^-1. Thirteen exclusive selections are developed according to the different final state topologies. No statistically significant evidence for a Higgs boson decaying into a WW pair has been found. An upper limit is derived, as a function of the Higgs boson mass, on the product of the e+e- ->Hff cross section and the H ->WW branching ratio. The data on the search for H ->WW are combined with previously published ALEPH results on the search for H ->gamma gamma, to significantly extend the limits on the mass of a fermiophobic Higgs boson.


Introduction
A Higgs model [1] incorporating two doublets of complex scalar fields [2] generates five scalar Higgs bosons, three of which are neutral.In some types of models, for certain choices of parameters, one of these neutral scalars provides mass only to the fermions and another couples exclusively to the bosons, i.e. is a "fermiophobic" Higgs boson.Anomalous couplings in the Higgs sector can also enhance the bosonic branching fraction [3].
The search for a fermiophobic Higgs boson has been primarily carried out by the four LEP experiments in the H → γγ channel, in which the Higgs boson couples to photons via a W loop [4,5,6,7].A benchmark fermiophobic Higgs boson is defined by considering Standard-Model-like couplings to bosons, and null couplings to fermions.Current analyses exclude the benchmark fermiophobic Higgs boson up to a mass of 109.7 GeV/c 2 [8].For fermiophobic Higgs bosons heavier than 90 GeV/c 2 , the predicted H → γγ branching ratio becomes small relative to the predicted H → WW branching ratio (Fig. 1) motivating a search in this new channel.Such an analysis has already been carried out by the L3 collaboration [10] and is performed here with data collected by the ALEPH detector.
The main production processes at e + e − colliders for a fermiophobic Higgs boson are e + e − → Z * → ZH (Higgsstrahlung), WW and ZZ fusion.The cross sections of the boson fusion production processes are considerably smaller than that of the Higgsstrahlung process at LEP centre-of-mass (CM) energies.In the mass range kinematically accessible for Higgsstrahlung at LEP, one of the virtual W bosons is expected to be near on-shell, and the other (denoted W * ) to have a much smaller mass and energy.In this paper, all the signatures originating from the Z → qq, ν ν, ℓ + ℓ − decays and the W, W * → qq ′ , ℓ ± ν decays are searched for.For simplicity and conciseness, the term "lepton" (and the corresponding symbol ℓ) refers to electrons and muons only.Leptonic tau decays are not specifically addressed, but the corresponding selected events are included in the final results.The hadronic tau decays in e + e − → ZWW * → ℓ + ℓ − τ νqq ′ are also looked for.The analysis is performed on the data taken in the years 1999 and 2000 at CM energies ranging from 191 to 209 GeV.The luminosities and CM energies are shown in Table 1.
This paper is organized as follows.A brief description of the ALEPH detector is given is Section 2. The signal and background simulations are summarized in Section 3. The overall search strategy is presented in Section 4 and the specific selection algorithms in Section 5.The results are reported in Section 6.

The ALEPH detector
A detailed description of the ALEPH detector can be found in Ref. [11], and of its performance in Ref. [12].Here, only a brief description of the detector elements and the algorithms relevant to this analysis is given.The trajectories of charged particles are measured with a silicon vertex detector (VDET), a cylindrical drift chamber (ITC) and a large time projection chamber (TPC), all immersed in a 1.5 T axial magnetic field provided by a superconducting solenoidal coil.The energy of electrons, photons, and hadrons is measured with the electromagnetic (ECAL), the hadron (HCAL) and the luminosity (LCAL and SiCAL) calorimeters.The ECAL, placed between the TPC and the coil, is a highly segmented calorimeter, which is used to identify electrons and photons, and to measure their energy and position.The LCAL and SiCAL extend the calorimetric coverage down to 34 mrad from the beam axis.The HCAL consists of an instrumented iron return yoke.It provides the measurement of hadronic energy and, together with external chambers, muon identification.
In the following, good tracks (or simply tracks) are defined as charged particle tracks reconstructed with at least four hits in the time projection chamber, originating from within a cylinder of 20 cm length and 2 cm radius, coaxial with the beam and centred on the nominal collision point, and with a polar angle with respect to the beam such that | cos θ| < 0.95.
In this analysis, all searches make use of the same lepton identification criteria, where needed and applicable.Electrons are identified by comparing the momentum measured in the tracking detectors with the energy measured in the ECAL, by the depth and shape of the ECAL shower, and by the specific ionization information from the TPC, when available.Muons are identified by their characteristic hit pattern in the hadron calorimeter, and must have at least one associated hit in the muon chambers.Lepton identification is described in detail in Ref. [13].
Global event quantities, such as the total visible energy (E tot ), the total visible mass (M tot ), or the missing energy (E miss ) are measured with an energy-flow algorithm [12], which combines individual tracker and calorimeter measurements into energy-flow "particles".These objects are classified as photons, electrons, muons, neutral or charged hadrons.The jets used in the present analysis are obtained by clustering the energyflow particles with the Durham jet-finding algorithm [14], and allow various event topological variables to be determined.These include, for example, the acollinearity angle θ aco (respectively the acoplanarity angle Φ aco ) between two jets (respectively between their projections onto the plane transverse to the beam axis) in an event forced to form two jets, the transition values y ij of the resolution parameter y cut at which the number of jets in an event switches from j to i jets, or the good track multiplicity N ch,i in jet i.Other specific variables, related either to the event topology, to the jet properties, or to the lepton characteristics, are described in turn in Section 5.

Simulated samples
Signal samples were generated using HZHA [9] for Higgs boson masses between 90 and 117 GeV/c 2 and at seven different CM energies: 191.6, 195.5, 199.5, 201.6, 204.9, 206.5 and 208 GeV, including Higgsstrahlung and fusion processes.All decays of the Z and W bosons were considered.In the HZHA code, there is no spin correlation between the W bosons coming from the Higgs boson decay.The signal events were therefore re-weighted to take into account this spin correlation.The event weight was computed as the ratio between the full four-fermion matrix element and the HZHA matrix element [15].
Event samples of all Standard Model (SM) background processes relevant for the Higgs boson search were also generated: the Bhabha process was simulated with BHWIDE 1.01 [16], qq, dimuon and ditau events with KK 4.14 [17], γγ processes with PHOT02 [18].In the following, these processes will be grouped under the label "twofermions".WW production was simulated with KORALW 1.51 [19] and the remaining fourfermion processes with PYTHIA 6.1 [20].The background event samples were generated at the same CM energy values as the signal.The simulated sample sizes are at least a factor 20 greater than the data.A detailed simulation of the detector response was applied to both background and signal events.

Search strategy 4.1 Event classes, topological searches and targeted channels
The event selection is subdivided in a number of topological searches, each of which targets a specific final state (or channel) arising from the ZWW * production.The list of channels addressed in this paper is given in the second column of Table 2, with the Z, the W and the W * decays given in this order, together with the corresponding branching fractions.Altogether, almost 80% of the possible final states are targeted by the selection algorithms developed for this study.
Events are first separated in four exclusive classes according to the number and the energy of identified leptons in the final state.The four classes, further subdivided in different topologies (or subclasses), are defined as follows, and are displayed in the first column of Table 2.
Fully Hadronic (Class 1): in this class, only events with neither energetic nor isolated identified leptons are selected.It addresses final states exclusively with hadronic jets, with (class 1a) or without (class 1b) missing energy, depending on whether the Z decays in a pair of quarks or a pair of neutrinos.
Two Hard Leptons (Class 2): in this class, events with at least two energetic identified leptons are selected.It addresses final states in which the Z decays into a lepton pair.Five different topological searches were developed according to the W and W * decay modes.
One Hard Lepton (class 3): in this class, events with exactly one energetic identified lepton are selected.It addresses final states in which the W decays to ℓν.Four different topologies were defined according to the hadronic activity and the missing energy, to target the remaining Z and W * decays.
One Soft Lepton (class 4): in this class, events with exactly one isolated identified lepton that does not meet the momentum requirement of class 3 are selected, to address the W * → ℓν decays.Two different topological searches were developed according to the hadronic activity and the missing energy, to address the hadronic and invisible Z decays.These definitions, and the corresponding topological searches, were developed to minimize the cross-channel contamination between the different subclasses.

Topological search optimization
In each of the topological searches, the selection criteria were tailored to optimize the combined sensitivity to a Higgs boson mass hypothesis of 110 GeV/c 2 , which is near the expected experimental sensitivity in the fermiophobic scenario.To do so, the expected combined confidence level on the signal hypothesis, CL s , that would be obtained on average if no signal were present, is minimized [21] with respect to the position of the cuts on most of the selection variables.An estimate of the value of CL s is determined with a toy Monte Carlo method using the approximate formula of Ref. [22], with the algorithm of Ref. [23].
For each topology, the determination of CL s requires the expected number of background events (N b ), the number of signal events expected from the targeted final state N s , and the expected distribution of a variable D aimed at discriminating between the signal and the background.The observed values of the confidence levels on the signal and on the background hypotheses, CL s and CL b , are obtained in the same way from the number of events observed (N d ) and the value of D for each of these events.
In the optimization process the values of N b and N s correspond to the CM energies and the integrated luminosity collected in the year 2000.Lower energy data sets would indeed not contribute significantly to the combined sensitivity to a 110 GeV/c 2 Higgs boson signal, and are therefore absent from all distributions presented in Section 5. Since, however, they increase the sensitivity to smaller masses, the data taken in 1999 were included in the final result (Section 6).

Preselection and class assignment
Common preselection cuts are applied in all four classes in order to strongly reduce the γγ and ℓ + ℓ − backgrounds.The energy within 12 of the beam axis, E 12 , must be less than 40% of the CM energy.The acollinearity must be less than 170 for events with less than four tracks.Finally, the total invariant mass M tot and total transverse momentum P t of the event must satisfy M tot + 6P t > 0.2 √ s.
The event-to-(sub)class assignment is based on the energies E ℓ i and the isolations ) of the three most energetic leptons in the event.Here θ ℓ i T is the angle between the ith lepton direction and the closest track in the event.If less than three leptons are found, the corresponding energies and isolation variables are set to 0 and 10 −20 respectively.Events in which the most energetic lepton is "hard" (E ℓ 1 > 25 GeV) are assigned to class 2 or 3 depending on the energy of the second most energetic lepton.To separate the remaining events between classes 1 and 4, a linear discriminant D 14 is built with E ℓ 1 , the total missing three-momentum (P miss ), and the isolation of the most energetic lepton (I ℓ 1 ): Details of the partition process are presented in Table 3.The criteria used to define all the subclasses presented in Section 4.1 are also shown in the table.Subclasses 1a and 1b are separated by a cut on the missing mass M miss .Subclasses 2a and 2t are separated from subclasses 2b, 2c and 2d with a cut on E ℓ 3 .Subclass 2b is finally identified with a cut on the total hadronic energy, E had .Subclasses 2c and 2d are separated from each other by a cut on the number of tracks (N ch ) while subclasses 2a and 2t are distinguished using the hadronic activity.The separation between subclasses 3a, 3b and 3c, 3d is achieved by cutting on the total and missing invariant mass of the event.A cut on the energy of the second most energetic lepton is used to distinguish subclasses 3a, 3d from 3b, 3c.Finally, subclasses 4a and 4b are separated from each other by cutting on the total invariant mass of the event.
Corresponding preselection cuts are applied, in addition to those applied in the four classes, and are presented in Table 4, together with the numbers of selected data, background and signal events.After the preselection, the background is dominated by WW events.The final selection cuts are given in the lower section of  are assigned to the W * , and the remaining two jets to the W. The invariant mass of the Z, W and W * are denoted M 1 , M 2 and M 3 , respectively.

Subclass Selection Criteria
The discriminant variable M 2 enters the CL s computation and is shown in Figs.2a  and 2b, after the preselection and the final selection, respectively.In this subclass, the targeted-signal efficiency is 61% and the number of expected background events is 163.7.requiring the invariant mass of the two leptons (m rec Z ) to be in a window around the nominal Z mass (m Z ).The remaining ZZ background is reduced by cutting on the angle between the two leptons and on their total energy.The details of the preselection and selection are shown in Table 5.

Class
The discriminant variable, inspired from the Higgs boson mass is then computed as: where P ℓ 1 and P ℓ 2 are the momenta of the leptons associated with the selected pair and P tot is the total measured momentum.The discriminant variable is shown in Figs.3a and 3b after preselection and final selection, respectively.In this subclass, the targeted-signal efficiency is 74%.The expected background is 0.67 events.

Class 2t
: Two leptons and missing energy targeted channel: ZH → ℓ + ℓ − τ ν qq This is the only case where the one-or three-prong hadronic tau decays can be distinguished efficiently from other hadronic decays.The rejection of qq and WW events is achieved by requiring the reconstructed mass of the two leptons to be in a window around the Z mass.A cut on the transverse momentum of the lepton pair (P t (ℓ + ℓ − )) reduces the ZZ background.The selection cuts are shown in Table 5.In this subclass, the targeted-signal efficiency is 60%.The expected background is 0.46 events.

Class 2b:
Two leptons, two jets and one soft lepton This subclass is characterized by a third lepton and significant hadronic activity.The selection proceeds in a similar way as in the subclass 2a by concentrating on the two leptons associated with the Z boson decay.The isolation for the lepton that is the most anti-parallel to the missing momentum (I ℓ A ) is used in the selection.The lepton tends to be more isolated in the signal than in the background.The selection cuts are shown in Table 5.After the final selection the dominant background is ZZ events decaying into ℓ + ℓ − bb.The pair of leptons which have the same flavour and opposite charge and give the best estimate of the Z mass are associated with the Z boson decay.The following discriminant is used: where P Z and E Z are the momenta and energy of the Z boson, determined from the two assigned leptons.The additional term with respect to D 1 introduces the correction needed to take into account the undetected neutrino, assumed to be produced back-to-back to the third lepton.The discriminant variable is shown in Figs.3c and 3d before and after the full selection, respectively.In this subclass, the targeted-signal efficiency is 71%.The expected background is 0.16 events.

Class 2c: Three leptons and two jets targeted channel: ZH
This subclass is characterized by a third lepton and low hadronic activity.Compared to subclass 2b, the missing transverse momentum is higher, which makes it more difficult to distinguish from semi-leptonic WW events.The mass window for the reconstructed Z mass is also broader due to the larger combinatorial background.A cut on the transverse momentum of the Z boson is applied to reduce the ZZ background.
The signal and background distributions of the discriminant variable, D 2 , are shown in Figs.3e and 3f.The dominant background is ZZ events.In this subclass, the targetedsignal efficiency is 91%.The expected background is 0.17 events.
This subclass is characterized by a small branching fraction (the targeted channel represents 5% of the events in class 2) but with a clear topology: four leptons in the final state, one of them being soft.To reduce the ZZ background, cuts on the thrust and the event acoplanarity are applied.The remaining WW events are rejected by requiring that the most anti-parallel lepton with respect to the missing momentum is well isolated.
The selection criteria are detailed in Table 5.In this subclass, the targeted-signal efficiency is 57%.The expected background is 0.58 events.

Class 3: Final state with one hard lepton
In class 3, in order to reduce the γγ background, additional preselection cuts are applied on the total transverse momentum and on the cosine of the polar angle of the missing momentum cos(θ miss ).Depending on the subclass, a cut on the number of tracks and/or on the most energetic reconstructed photon (E γ 1 ) is required.

Class 3a: One lepton and four jets targeted channel: ZH → qq ℓν qq
To eliminate the γγ, ℓ + ℓ − and qq backgrounds, cuts are applied to the track multiplicity, the angle between the hard lepton and the total momentum θ ℓ,Σ , and the mass of the hard W boson, reconstructed as the invariant mass of the hard lepton and the missing momentum (M ℓ,P miss ).To suppress the remaining backgrounds, mainly WW, E ℓ 1 , the hadronic acollinearity (θ aco (no lept) ) the thrust computed without the hard lepton (T no lept ), y 45 , and the total transverse momentum are used.The selection criteria together with the numbers of signal, background and data events are summarized in Table 6.Figures.4a and 4b show the discriminant variable, M tot (hadr) , after the preselection and after all cuts, respectively.In this subclass, the targeted-signal efficiency is 39%.The expected background is 2.9 events.

Class 3b:
One lepton, two jets plus one soft lepton targeted channel: ZH → qq ℓν ℓν The selection procedure is very similar to that in subclass 3a, and is shown together with the numbers of signal, background and data events after each selection step in Table 6.
The γγ, ℓ + ℓ − and qq backgrounds are rejected by cuts on the number of tracks, on the total acollinearity and on the transverse momentum.To further suppress qq events, cuts on the angle between the hard lepton and the total momentum as well as on the event sphericity are applied.Finally, most of the WW events are removed by a cut on the hard lepton energy which is sensibly smaller for signal events than for WW events, on the hadronic acollinearity, on y 12 (no lept) , computed without the leptons, and on y 34 .Remaining background events are mainly semi-leptonic WW decays with a soft lepton produced in a jet.A good estimate of the Higgs boson mass can be obtained from twice the sum of the lepton energies.The following variable is therefore used as discriminant variable: The discriminant variable, D 3 , after the preselection and after all cuts is shown in Figs.4c and 4d respectively.In this subclass, the targeted-signal efficiency is 60%.The expected background is 0.76 events.

Class 3c: One lepton and one track targeted channel: ZH → ν ν ℓν ℓν
This subclass is characterized by a significant acollinearity between the two leptons, and by a small invariant mass.Only events with exactly two tracks are kept.The γγ events are rejected by a cut on the acollinearity and on the transverse momentum.The total mass and the acoplanarity are used to reject ℓ + ℓ − events as well as part of the WW background.
A final cut on the total missing mass and on θ ℓ,Σ removes most of the remaining WW and ZZ background.
The discriminant variable, D 3 , after the preselection and after all cuts is given in Figs.5a and 5b respectively.The selection criteria together with the numbers of signal, background and data events are summarized in Table 6.In this subclass, the targetedsignal efficiency is 58%.The expected background is 3.5 events.

Class 3d: One lepton and two jets
targeted channel: ZH → ν ν ℓν qq Events in this subclass are characterized by a single hard lepton with some soft hadronic activity.The γγ and ℓ + ℓ − events are rejected by cutting on the aplanarity, A [24].The dominant WW background is then reduced in two stages.First, the angle between the hard lepton and the total momentum and the energy of both the first and second most energetic leptons are used.Then, a final rejection is achieved by cuts on the sphericity, on M tot (hadr) and on y 12 (no lept) .The hadronic activity is included in the discriminant variable, defined as The discriminant variable after the preselection and after all cuts is shown in Figs.5c and 5d respectively.The selection criteria together with the numbers of signal, background and data events are summarized in Table 6.In this subclass, the targeted-signal efficiency is 58%.The expected background is 0.65 events.
5.5 Class 4: Soft-lepton final state 5.5.1 Class 4a: One soft lepton and four jets targeted channel: ZH → qq qq ℓν The event selection relies on y 45 , y 34 (no lept) and the lepton isolation, computed for the most anti-parallel lepton with respect to the missing momentum.This last variable reduces the background by a factor 6 (for a 42% efficiency in signal events).A χ 2 is then built which takes into account the W and Z boson masses, reconstructed from the four jets in the event.For this, jets are paired and the mass of each pair is compared either to the nominal W boson mass or to the nominal Z boson mass.The jet pairing that minimizes the χ 2 is retained: were M 12 and M 34 are the masses of each jet pair and σ is the estimated mean resolution on the reconstructed mass.The background is finally reduced by constraining the value of the χ 2 , the total hadronic mass and the lepton energy.The discriminating variable used for the estimation of the quoted confidence level is the reconstructed off-shell W mass, estimated from the missing momentum and the soft lepton momentum as follows: Details of the event selection are given in Table 7.The discriminant variable, after the preselection and the final selection is presented in Figs.6a and 6b, respectively.In this subclass, the targeted-signal efficiency is 35%.The expected background is 4.0 events.additional cuts on the total hadronic mass, the lepton isolation and on the hadronic acollinearity.Since at least three neutrinos are expected in this channel, a cut on the missing mass is used in the final selection.The dominant background after the selection is WW.The discriminant variable, D 5 , after the preselection and the final selection is presented in Figs.6c and 6d, respectively.The expected number of events after the signal selection is shown in Table 7.In this subclass, the targeted-signal efficiency is 54%.The expected background is 8.0 events.

Results
The numbers of signal (m H = 110 GeV/c 2 ), background and data events for the years 1999 and 2000, for each class, are given in Table 8.The numbers of observed events agree well with the expectations for all subclasses but 3a and 3c, where there is a small excess (∼ 2σ) Table 8: Numbers of signal events for a 110 GeV/c 2 Higgs boson (N s ), background (N b ) and data (N d ) events, as well as the value of the expected and observed confidence levels for each subclass, for the years 1999 and 2000 together.

Class
N over the expected background (CL b =0.97).For subclass 3a, 6.6 background events remain after the cuts, for 1.27 signal events, while thirteen candidates are observed.Similarly, for subclass 3c, 0.15 signal events are expected for 6.6 background events, and fourteen candidates are observed in the data.The best sensitivity is achieved for subclasses with a low expected CL s .These are subclasses 3d, 3a, 4a, 2a and 1b, in that order.The four classes are combined and the compatibility between data and background with and without the signal is evaluated with the log-likelihood ratio estimator ln Q [23].
The combined expected values of the signal and background confidence levels are 0.08 and 0.50 respectively.The observed values for the signal and background confidence levels are 0.26 and 0.87, respectively.

Systematic uncertainties
The main contribution to the systematic error comes from the simulated statistics.The uncertainties from the production cross sections, from the simulation of the calorimeter energy scale and from hadronization processes are also taken into account.The beam background is conservatively taken into account in the analysis (via event selection variable E 12 ).Uncertainties from the simulated statistics are included when combining the searches in the several subclasses by varying the expected signal and background levels bin per bin in a number of toy Monte Carlo experiments, while other uncertainties are taken into account in a correlated way by varying the expected signal and background levels coherently in the same direction for all subclasses and all bins.Each source of uncertainty is nevertheless handled independently, which corresponds to adding up their effect in quadrature.
For the four different classes, after applying the selection cuts, the main remaining background comes from WW pair production.The related uncertainty, associated to W decays into three jets, is taken to be 2%.This is a conservative estimate from the uncertainty on the α s measurement [25].The systematic errors related to the simulation of the absolute energy scale of the calorimeters are determined using hadronic Z events and are found to be ±0.9% and ±2% for the electromagnetic and hadronic calorimeters, respectively.The effect of a possible miscalibration of the calorimeters is evaluated on simulated samples by scaling the electromagnetic and hadronic part of the measured energy independently by these amounts.The largest of the observed shifts for each calorimeter is combined in quadrature.This leads to a mean uncertainty of the order of 3% on the remaining background level.The uncertainties originating from the hadronization model are evaluated by comparing WW events hadronized using the string model (JETSET 7.4 [26] Monte Carlo) and with the colour dipole model (ARIADNE 4.10 [27] Monte Carlo).The associated mean systematic error is 6%.Both the hadronization and calorimetric uncertainties are evaluated separately for each subclass.The uncertainties taken into account are presented in Table 9.

Combined Likelihood Ratio
The combined log-likelihood ratio as a function of the Higgs boson mass is presented in Fig. 7.A 1.5σ excess is observed.This excess does not depend on the hypothetical mass, and no supporting indication of a signal is seen.An upper limit on the signal cross section is set as a function of the Higgs boson mass.

Cross section upper limits
For each class in the analysis, an upper limit on the production cross section at a given Higgs boson mass is derived.Figure 8 presents the resulting 95% C.L. upper limit on ξ 2 = BR(H → WW)σ(e + e − → Hf f)/σ SM (e + e − → Hf f), as a function of the Higgs boson mass.The third class, providing the best compromise between a clean signature and signal sensitivity has the highest reach.The other three classes are nevertheless as important in order to get a competitive limit.Combining all four analyses, the 95% C.L. upper limit on ξ 2 as a function of the Higgs boson mass is given in Fig. 9.The benchmark fermiophobic Higgs model is drawn as the full line in the figure.The afore-mentioned excess in class 3 results in an observed limit which is less than the expected limit.

Mass exclusion limits
Lower limits on the Higgs boson mass can be extracted in the context of a given model from upper limits on the cross section presented in the previous section.If the branching ratio of the Higgs boson to W bosons is taken to be 100%, the expected limit is 107.5 GeV/c 2 .In the fermiophobic Higgs boson scenario, it is expected that Higgs boson masses between 97.5 GeV/c 2 and 104 GeV/c 2 can be excluded.For these two scenarios, due to the excess observed, no limit at 95% C.L. can be set.However, the present search for H → WW can be combined with the previously published ALEPH search for H → γγ [4] to significantly improve the limits on the fermiophobic Higgs boson scenario (Fig. 1).The combined expected limit on the fermiophobic Higgs boson mass is 111.4GeV/c2 , an improvement of ∼ 7 GeV/c 2 on the sensitivity from the H → γγ search alone [4]1 .The observed limit is 105.8GeV/c 2 .

ALEPH
A model-independent limit can be derived by scanning the H → γγ and H → WW branching fractions.This is conveniently parametrized as: BR bosons = BR H→γγ + BR H→WW + BR H→ZZ , R γγ = BR H→γγ /BR bosons , where R γγ represents the fraction of bosonic decays into photon pairs and ranges from zero to one.The best limit is obtained combining the present results with those previously published by ALEPH on the search for H → γγ [4].The 95% C.L. limit on BR bosons is determined at each point of the m H versus R γγ plane, resulting in the exclusion curves of Fig. 10.

Conclusions
A search for a Higgs boson produced in association with a Z and decaying into a WW pair has been performed with a dataset of 453.

Figure 2 :
Figure 2: Discriminant variable M 2 for class 1a events after the preselection (a) and after the final selection cuts (b).Discriminant variable M tot / √ s for class 1b events after the preselection (c) and after the final selection cuts (d).All distributions are obtained from year 2000 data.

Figure 3 :
Figure 3: Discriminant variable, D 1 , for subclass 2a after the preselection (a) and the final selection (b).In plots (c) and (d), the discriminant variable, D 2 , is shown after the preselection and the final selection, respectively, for events in class 2b.The two lower plots (e) and (f) show the same variable, after the preselection and the final selection, respectively, for events in class 2c.All distributions are obtained from 2000 data.

Figure 4 :
Figure 4: Discriminant variable, M tot (hadr) , for subclass 3a events after the preselection (a) and after the final selection cuts (b).Discriminant variable, D 3 , for subclass 3b events after the preselection (c) and after the final selection cuts (d).All distributions are obtained from 2000 data.

5. 5 . 2 Figure 5 :
Figure 5: Discriminant variable, D 3 , for subclass 3c events after the preselection (a) and the final selection (b).Discriminant variable, D 4 , for subclass 3d events after the preselection (c) and the final selection (d).All distributions are obtained from 2000 data.

ALEPHFigure 7 :
Figure 7: Log-likelihood ratio, −2ln Q, as a function of the Higgs boson mass hypothesis, m H , with all data collected between 191 GeV and 209 GeV.The solid line is the result obtained from the data.The expected background-only and signal-plus-background likelihoods are indicated by the dashed lines; the light and dark shaded bands around the background expectation contain 68% and 95% of the simulated background-only experiments, respectively.The 1σ bands for the signal-plus-background hypothesis are also shown.

Figure 8 :
Figure 8: Limit on ξ 2 (defined in the text) as a function of the Higgs boson mass hypothesis, m H , in the four different classes.The dashed line corresponds to the expected limit while the 95% C.L. excluded region is shown by the gray area.

ALEPHFigure 9 :ALEPHFigure 10 :
Figure 9: Limit on ξ 2 (defined in the text) as a function of the Higgs boson mass hypothesis, m H .The dashed line corresponds to the expected limit while the 95% C.L. excluded region is shown by the gray area.The benchmark fermiophobic Higgs model branching ratio is depicted by the full line.

Table 4
j6represent the smallest and largest angle between any pair of jets.The jet pair with the invariant mass closest to 91.2 GeV/c 2 is assigned to the Z.The two least energetic jets

Table 3 :
Details of the partition of events into each of the thirteen subclasses.Energies are expressed in GeV and masses in GeV/c 2 .

Table 4 :
Selection criteria for each subclass in class 1.The numbers of signal (N s ), background (N b ) and data (N d ) events are given in the table for the year 2000.Energies, momenta and masses are expressed in GeV, GeV/c and GeV/c 2 , respectively.

Table 4 .
[24]class 1b preselection has two additional cuts on the value of the cosine of the polar angle of the missing momentum cos(θ miss ) and on E 12 .After the preselection, the proportion of main background events is 43% WW, 30% qq, 12% Weν and 12% ZZ.The final selection cuts are detailed in the lower section of Table4.The acoplanarity and P t / √ s are used as well as the thrust T[24]of the event.The event is forced to form four jets.The smallest number of tracks in any jet is denoted N min ch,j4 , and the variables θ 1b: Four-jets and missing transverse momentum final state targeted channel: ZH → ν ν qq qq This subclass is characterized by a final state with a large missing mass.The preselection cuts used for subclass 1a are also relevant with, in general, weaker cut values, as shown in 5.3.1 Class 2a: Two leptons and four jets targeted channel: ZH → ℓ + ℓ − qq qqThe dominant backgrounds after preselection are qq, semi-leptonic WW as well as ZZ events where one of the Z bosons decays into hadrons and the other into leptons.The rejection of qq and WW events is achieved by applying a cut on the variable y 45 and

Table 5 :
Selection criteria for each subclass in class 2. The numbers of signal (N s ), background (N b ) and data (N d ) events are given in the table for the year 2000.Energies, momenta and masses are expressed in GeV, GeV/c and GeV/c 2 , respectively.

Table 6 :
Selection criteria for each subclass in class 3. The numbers of signal (N s ), background (N b ) and data (N d ) events are given in the table for the year 2000.Energies, momenta and masses are expressed in GeV, GeV/c and GeV/c 2 , respectively.

Table 9 :
Levels of systematic uncertainties considered for each subclass.The corresponding total systematic uncertainty is also given in the last column, excluding the impact of the limited simulated statistics.