UvA-DARE (Digital Academic Repository) Study of (W/Z)H production and Higgs boson couplings using H -> WW* decays with the ATLAS detector

: A search for Higgs boson production in association with a W or Z boson, in the H → W W ∗ decay channel, is performed with a data sample collected with the ATLAS detector at the LHC in proton-proton collisions at centre-of-mass energies √ s = 7 TeV and 8 TeV, corresponding to integrated luminosities of 4.5 fb − 1 and 20.3 fb − 1 , respectively. The W H production mode is studied in two-lepton and three-lepton ﬁnal states, while two-lepton and four-lepton ﬁnal states are used to search for the ZH production mode. The observed signiﬁcance, for the combined W H and ZH production, is 2.5 standard deviations while a signiﬁcance of 0.9 standard deviations is expected in the Standard Model Higgs boson hypothesis. The ratio of the combined W H and ZH signal yield to the Standard Model expectation, µ V H , is found to be µ V H = 3 . 0 +1 . 3 − 1 . 1 (stat . ) +1 . 0 − 0 . 7 (sys . ) for the Higgs boson mass of 125.36 GeV. The W H and ZH production modes are also combined with the gluon fusion and vector boson fusion production modes studied in the H → W W ∗ → `ν`ν decay channel, resulting in an overall observed signiﬁcance of 6.5 standard deviations and µ ggF+VBF+VH = 1 . 16 +0 . 16 − 0 . 15 (stat . ) +0 . 18 − 0 . 15 (sys . ). The results are interpreted in terms of scaling factors of the Higgs boson couplings to vector bosons ( κ V ) and fermions ( κ F ); the combined results are: | κ V | = 1 . 06 +0 . 10 − 0 . 10 , | κ F | = 0 . 85 +0 . 26 − 0 . 20 .


JHEP08(2015)137
In the Standard Model (SM) of fundamental interactions, the Brout-Englert-Higgs [1][2][3] mechanism induces the electroweak symmetry breaking that provides mass to elementary particles. The mechanism postulates the existence of an elementary scalar particle, the Higgs boson. The ATLAS and CMS collaborations at the CERN Large Hadron Collider (LHC) have observed the Higgs boson with a mass (m H ) of about 125 GeV [4,5]. The measurements of the Higgs boson couplings to SM particles, and its spin and CP quantum numbers, are essential tests of the SM [6-12]. Higgs boson production in association with a W or Z (weak) boson, which are respectively denoted by W H and ZH, and collectively referred to as V H associated production in the following, provides direct access to the Higgs boson couplings to weak bosons. In particular, in the W H mode with subsequent H→ W W * decay, the Higgs boson couples only to W bosons, at both the production and decay vertices. Searches for V H production have been performed at both the Tevatron and LHC colliders, in events with leptons, b-jets and either missing transverse momentum or two central jets. Evidence for V H production has been recently reported in the Tevatron combination [13] while no V H production has been observed so far at the LHC [14 -20].
In this paper, a search for Higgs boson production in association with a weak boson, followed by H→ W W * decay, is presented. The data were collected in 2011 and 2012 by the ATLAS experiment at centre-of-mass energies of √ s = 7 TeV and 8 TeV, respectively. In the SM, for m H = 125 GeV, the cross sections of the W H and ZH associated production modes, followed by the H→ W W * decay, are 0.12 pb and 0.07 pb at √ s = 7 TeV and 0.15 pb and 0.09 pb at √ s = 8 TeV [21], respectively. Four topologies are considered, with two, three or four charged leptons in the final state (only electrons or muons are considered). The analyses are optimised to search for both the W H and ZH production modes; a combined result for V H is also presented. The V H results are then further combined with the H→ W W * → ν ν analysis of gluon fusion (ggF) and vector boson fusion (VBF) production, for which the ATLAS Collaboration has reported the observation of the Higgs boson in the H→ W W * decay channel with a significance of 6.1 standard deviations [22].
The combination of the ggF, VBF and V H analyses, presented in this paper, is used to determine the couplings of the Higgs boson to vector bosons and, indirectly, to fermions, providing further constraints on the Higgs boson couplings.

Analysis overview
Higgs boson production in association with a W or Z boson, followed by H→ W W * decay, is sought using events with two, three or four charged leptons in the final state. Leptonic decays of τ leptons from H→ W W * → τ ντ ν are considered as signal, while no specific selection is performed for events with hadronically decaying τ leptons in the final state. In the present analysis events from V H(H → τ τ ) are considered as background. The analysis is designed to select events which are kinematically consistent with the V H(H→ W W * ) process, in order to enhance the signal-to-background ratio. Figure 1 illustrates the relevant tree-level Feynman diagrams of the studied processes, in which a Higgs boson is produced in association with a weak boson.
Four channels are analysed, defined as follows: (a) 4 channel (figure 1(a)): the leading contribution consists of a process in which a virtual Z boson radiates a Higgs boson, which in turn decays to a W boson pair. The decays of the weak bosons produce four charged leptons and two neutrinos in the final state. The lepton pair with an invariant mass closest to the Z boson mass is labelled as ( 2 , 3 ), while the remaining leptons are labelled as 0 and 1 and are assumed to originate in the H→ W W * decay. The main backgrounds to this channel are non-resonant ZZ * and ZW W * production.
(b) 3 channel ( figure 1(b)): the leading contribution consists of a process in which a virtual W boson radiates a Higgs boson, and the Higgs boson decays to a W boson pair. All the weak bosons decay leptonically producing three charged leptons and three neutrinos in the final state. The lepton with unique charge is labelled as 0 , the lepton closest to 0 in angle is labelled as 1 , and the remaining lepton is labelled as 2 . Leptons 0 and 1 are assumed to originate from the H→ W W * decay. The -3 -JHEP08(2015)137 data sample, due to the low sensitivity of this channel. The 8 TeV data were taken at a higher instantaneous luminosity (L 7 × 10 33 cm −2 s −1 ) than that for the 7 TeV data (L 3×10 33 cm −2 s −1 ) and with a higher number ( 21 versus 9) of overlapping protonproton collisions, producing higher out-of-time and in-time pile-up [24]. The increased pile-up rate, rather than the increased centre-of-mass energy, is the main reason for the differences between 8 TeV and 7 TeV analysis selections. Table 1 lists the Monte Carlo (MC) generators used to model the signal and background processes. For the Higgs production processes the production cross section multiplied by the branching fraction of the H→ W W * decay is shown, while for the background processes the production cross section, including effects of cuts applied at the event generation, is presented. The samples were simulated and normalised for a Higgs boson of mass m H = 125 GeV. The V H samples were simulated with Pythia and normalised to the nextto-next-to-leading-order (NNLO) QCD calculations [21,[44][45][46][47] with additional next-toleading-order (NLO) electroweak (EW) corrections computed with Hawk [48] and applied as a function of the transverse momentum of the associated vector boson. The gg → ZH samples were simulated with Powheg-Box1.0 interfaced with pythia8 and normalised to the NNLO QCD calculations [45]. Associated Higgs boson production with a tt pair (ttH) is simulated with Pythia8 and normalised to the NLO QCD estimation [21,44,45].
The matrix-element-level calculations are interfaced to generators that model the parton shower, the hadronisation and the underlying event, using either Pythia6, Pythia8, Herwig with the underlying event modelled by Jimmy [49], or Sherpa. The CT10 parton distribution function (PDF) set [50] is used for the Powheg-Box and Sherpa samples while the CTEQ6L1 PDF set [51] is used for Alpgen and AcerMC samples. The Z/γ * sample is reweighted to the MRSTMCal [52] PDF set. The simulated samples are described in detail in ref. [22] with a few exceptions that are reported in the following.
The Z/γ * processes associated with light-and heavy-flavour (HF) jets are modelled by Alpgen+Herwig with merged leading-order (LO) calculations. The simulation includes processes with up to five additional partons in the matrix element, or three additional partons in processes with b-or c-quarks. An overlap-removal procedure is applied to avoid double counting of HF in the light-jet samples. The sum of the two samples is normalised to the NNLO calculation of Dynnlo [53,54]. The ttW/Z and tZ backgrounds are simulated using Madgraph at LO interfaced with Pythia6. The production of four leptons from a pair of virtual Z or γ bosons, indicated by ZZ * in the following, contributes to the background in the 3 channel when one low-p T lepton is not detected. Since this background is more prominent when one lepton pair has a very low mass, a dedicated sample which requires at least one SFOS pair with m < 4 GeV, generated with Sherpa and normalised to the NLO QCD cross section from the parton-level MC program MCFM [55], is included. Production of triboson processes is a major source of background, in particular W W W * in the 3 channel and ZW W * in the 4 channel. They are modelled by Madgraph interfaced with Pythia6 and normalised to the NLO cross section from ref. [56]. section times branching fraction of the H→ W W * decay, σ × Br, are shown for the Higgs production processes, while for background processes the production cross section, including the effect of the leptonic branching fraction, and the m and p γ T cuts, as specified in the "Process" column, is presented. 'HF' refers to heavy-flavour jet production, and 'VBS' refers to vector boson scattering. When a lower cut on m is specified, it is applied to all SFOS lepton pairs, while when an upper cut is indicated it is applied to the SFOS pair of lowest mass in the event. For the Sherpa1.1 Z ( * ) Z ( * ) sample a lower cut of 4 GeV is applied, in addition, to the SFOS lepton pair of higher mass. Cross sections are computed to different levels of accuracy (LO, NLO, NNLO or next-to-next-to-leading-logarithm, NNLL), as specified by the last column.

JHEP08(2015)137
of the electromagnetic and hadronic calorimeters, and with Geant4 for other detector components. The events are reweighted to ensure that the distribution of pile-up observed in the data is correctly reproduced.

Event reconstruction
The primary vertex of each event is selected as the vertex with the largest value of (p T ) 2 , where the sum is over all the tracks associated with that particular vertex. Furthermore, it is required to have at least three tracks with p T > 400 MeV.
Muons are reconstructed in the region |η| < 2.5 by combining tracks reconstructed in the MS and ID [60]. This analysis uses muon candidates referred to as "Chain 1, CB muons" in ref. [60]. Electrons are identified within the region |η| < 2.47, except in the transition region between barrel and endcap calorimeters (1.37 < |η| < 1.52), through the association of an ID track to a calorimeter cluster whose shower profile is consistent with an electromagnetic shower [61]. Electron identification uses information from both the calorimetric and tracking system. In the 7 TeV analysis a cut-based approach is adopted while in the 8 TeV analysis a likelihood-based selection is also exploited as described in [62]. Following that reference, in the 4 and 3 channels, electrons with p T < 20 GeV are required to satisfy the "very tight" likelihood requirement, while electrons with p T > 20 GeV are required to satisfy the "loose" likelihood requirement. In the 2 channels, electrons with p T < 25 GeV are required to satisfy the "very tight" likelihood requirement, while electrons with p T > 25 GeV are required to satisfy the "medium" likelihood requirement.
Both a track-based and a calorimeter-based isolation selection are applied to leptons. The isolation criteria are chosen to maximise the sensitivity to the V H(H→ W W * ) process at m H = 125 GeV. The track-based isolation is built on the computation of the scalar sum of the p T of tracks associated with the primary vertex and inside a cone, constructed around the candidate lepton, of size ∆R = 0.2 2 and excluding the track of the candidate lepton. The calorimeter-based isolation uses the scalar sum of the transverse energies measured within a cone of ∆R = 0.2, excluding the energy of the calorimeter cluster associated with the lepton itself. For the 8 TeV data the electron calorimeter-based isolation algorithm uses topological clusters [62], while for the 7 TeV data it uses calorimeter cells. Cell-based isolation is used for muons in the calorimeter in both the 8 TeV and 7 TeV analyses. The calorimeter and track isolation criteria differ between the 8 TeV and 7 TeV data samples and are not the same for all the channels. The upper bound of the calorimeter-based isolation energy varies from 7% to 30% of the lepton p T , while the sum of the p T of the tracks in the cone is required to be smaller than 4% to 12% of the lepton p T , where tighter cuts are applied at low p T . Less stringent isolation criteria on energy and p T are required for the 7 TeV data sample, due to the lower level of pile-up compared to the 8 TeV data sample.
-8 -JHEP08(2015)137 to have p T larger than 25 GeV except for the forward region, |η| > 2.4, in which the threshold is raised to 30 GeV. In order to suppress the contamination of jets from pile-up, the following selection is applied: the sum of the p T of all tracks within ∆R = 0.4 of the jet axis and that of the subset of these associated with the primary vertex is computed. The ratio of the latter to the former is required to be larger than 0.5 (0.75) for the 8 (7) TeV data samples, for all jets with p T < 50 GeV and |η| < 2.4.
The MV1 b-jet identification algorithm is used to tag jets containing a b-hadron [65]. For b-jets with |η| < 2.5 and p T > 20 (25) GeV in the 8 (7) TeV data analysis, the selection has an efficiency of 85%, estimated using simulated tt events. It corresponds to a rejection of a factor of 10 against jets originating from light quarks or gluons [66,67].
When two leptons are reconstructed within a cone of ∆R = 0.1, or a lepton and a jet are reconstructed within ∆R = 0.3, they are considered to be the same physical object and one of the two is removed. In the rare occurrence of an overlap between two leptons of the same flavour, the higher-p T lepton is kept while the lower-p T lepton is discarded. The muon is retained in the presence of an overlap with an electron, the electron is retained in the presence of an overlap with a jet, and the jet is retained in the presence of an overlap with a muon.
Two variables describing the missing transverse momentum are employed in this study: one is calorimeter-based and the other is track-based. The former, which benefits from the large rapidity coverage of the calorimeter and its sensitivity to neutral particles, is referenced as E miss T [68]. The E miss T magnitude, E miss T , is used in the analysis selection. The quantity E miss T is calculated as the negative vector sum of the momenta of muons, electrons, τ leptons, photons, jets and clusters of calorimeter cells that are not associated with these objects (the "soft term"). In the 8 TeV analysis, to suppress the pile-up effect, the ratio of the scalar p T sum of all soft term tracks associated with the primary vertex to the scalar p T sum of all soft term tracks from all vertices is employed. This ratio is used to scale all soft-event contributions to E miss T [69]. The track-based missing transverse momentum measurement is used to reduce the effects of pile-up on the resolution of the calorimeterbased variant [70]. It is calculated as the vector sum of the transverse momenta of tracks with p T > 500 MeV that originate from the primary vertex. This quantity is called p miss T , and the analysis selections are applied to its magnitude, p miss T . In order to include neutral components in the calculation of p miss T in final states with jets, the sum of track momenta in jets is replaced by their energy measured in the calorimeter.

Event selection
Events are required to contain a primary vertex. The four channels are further split into eight signal regions, designed to optimise the sensitivity to the V H(H→ W W * ) process, with a specific set of selections applied to define each signal region. The selection criteria rely on the number of leptons and their properties such as charge, flavour, p T , and on the number of jets and b-tagged jets and on the magnitude of the missing transverse momentum. Leptons with p T > 15 GeV are selected and their number is used to divide the analysis in the various channels. Similarly the analysis channels are subdivided in categories according the number of selected jets. Of particular importance are the invariant masses and opening angles among the selected objects, most notably those of opposite-sign lepton -9 -JHEP08(2015)137 pairs. The spin-0 property of the Higgs boson, in conjunction with the V -A structure of the weak interaction, results in a preference for a small opening angle of lepton pairs from the H → W W * → ν ν decays. On the other hand, as described in section 2, major backgrounds often contain Z boson production or tt production which give rise to oppositesign lepton pairs with a large opening angle. In the 2 -SS channel, the lepton originating from the Higgs boson decay is selected by minimising the invariant mass of the lepton and jet(s); cuts are then applied to the opening angle between this lepton and the closest jet in the transverse plane. The definitions of the signal regions used for each channel are summarised in table 2 and further detailed in sections 5.2.1-5.2.4.
In all the 4 and 3 signal regions, events are recorded using inclusive single-lepton triggers, which are fully efficient for high lepton multiplicity signatures. For the 2 channels in 8 TeV data taking, dilepton triggers are also used. In all channels at least one lepton must match a candidate reconstructed at trigger level. This requires the leading lepton in an event to have p T greater than 24 GeV in the 8 TeV data sample, and greater than 18 GeV and 20 GeV for muons and electrons respectively in the 7 TeV data sample. Single lepton trigger efficiencies are measured with respect to offline reconstructed leptons using leptonic Z decays. The measured values are approximately 95% for electrons, 90% for muons in the endcap and 70% for muons in the barrel.

Four-lepton channel
Events in the 4 channel are required to have exactly four leptons. The p T of leading and sub-leading leptons must be above 25 GeV and 20 GeV, respectively, and the p T of each of the remaining two leptons must exceed 15 GeV. The total charge of the four leptons is required to be zero. Only events with at least one SFOS lepton pair are accepted, and events are assigned to the 4 -2SFOS and 4 -1SFOS SRs according to the number of such pairs.
In order to select final states with neutrinos, E miss T and p miss T are required to be above 20 GeV and above 15 GeV, respectively. In order to reduce the ttZ background, events are vetoed if they contain more than one jet. Top-quark production is further suppressed by vetoing events with any b-tagged jet with p T above 20 GeV. The invariant mass of 2 and 3 , m 2 3 , is required to satisfy |m 2 3 − m Z | < 10 GeV (where m Z is the mass of the Z boson), and the invariant mass of 0 and 1 , m 0 1 , is required to be between 10 GeV and 65 GeV. This requirement on m 0 1 greatly reduces the contamination from ZZ ( * ) production in events with two pairs of SFOS leptons.
The sensitivity is improved by exploiting two additional variables. The variable ∆φ boost is required to be below 2.5 rad. The magnitude of the vector sum of the lepton transverse momenta, p T4 , can discriminate against the main background, ZZ ( * ) , which has no neutrinos. A cut requiring p T4 > 30 GeV is introduced for the 4 -2SFOS SR. In this signal region the invariant mass of the four leptons is required to be above 140 GeV to remove events from the H → ZZ * → 4 decay, which are the target of another analysis [17]. In the signal extraction through the fit explained in section 8.4, the 4 -2SFOS and 4 -1SFOS SRs enter as two separate signal regions. -11 -JHEP08(2015)137

Three-lepton channel
For the 3 channel, exactly three leptons with p T > 15 GeV are required with a total charge of ±1. After this requirement, contributions from background processes that include more than one misidentified lepton, such as W +jet production and inclusive bb pair production, are negligible. Events are then separated into the 3 -3SF, 3 -1SFOS and 3 -0SFOS SRs, requiring three SF leptons, one SFOS lepton pair and zero SFOS lepton pairs, respectively.
In order to reduce the background from tt production, events are vetoed if they contain more than one jet. The background from top-quark production is further suppressed by vetoing events if they contain any b-tagged jet with p T > 20 GeV. In order to select final states with neutrinos, E miss T is required to be above 30 GeV and p miss T above 20 GeV in the 3 -3SF and 3 -1SFOS SRs. In the 3 -0SFOS SR, E miss T or p miss T selections are not imposed because the main backgrounds also contain neutrinos. The invariant mass of all SFOS pairs in the 3 -3SF and 3 -1SFOS SRs is required to satisfy |m − m Z | > 25 GeV. This requirement suppresses W Z and ZZ * events, and increases the Z+jets rejection.
A lower bound is set on the smallest invariant mass of pairs of oppositely charged leptons at 12 GeV in the 3 -3SF and 3 -1SFOS SRs, and at 6 GeV in the 3 -0SFOS SR. In addition, an upper bound on the invariant mass of oppositely charged leptons is set at 200 GeV in the three signal regions. These selections reject backgrounds from HF and reduce the number of combinatorial lepton pairs from the W Z/W γ * process. The latter could indeed give larger mass values with respect to the W H process since it can proceed through the t-and u-channels, in addition to the s-channel, which is also present in W H production.
The angular separation ∆R 0 1 is required to be smaller than 2 in the 3 -3SF and 3 -1SFOS SRs. This cut favours the Higgs boson decay topology relative to that of W Z/W γ * events.
In the 3 -3SF and 3 -1SFOS SRs, the shape of a multivariate discriminant based on a Boosted Decision Trees (BDT) [71], which produces a multivariate classifier ("BDT Score"), is used to achieve a further separation between signal and background. The main purpose of the multivariate classifier is to distinguish between the signal and the dominant W Z/W γ * and ZZ * backgrounds, and the BDT is trained against these two background processes. The BDT parameters are chosen in order to ensure that there is no overtraining, i.e. that the BDT is robust against statistical fluctuations in the training samples. The BDT input discriminating variables which provide the best separation between signal and background are the p T of each lepton, the magnitude of their vector sum, the invariant masses of the two opposite-sign lepton pairs (m 0 1 , m 0 2 ), ∆R 0 1 , E miss T , and p miss T . In the fit, the shape of the distribution of the "BDT Score", divided into six bins, is used to extract the number of observed events in the 3 -3SF and 3 -1SFOS SRs, while the shape of the distribution of ∆R 0 1 , divided into four bins, is used to extract the number of observed events in the 3 -0SFOS SR. In the other channels only the event yield in each signal and control region is used without shape information.

Opposite-sign two-lepton channel
In the 2 -DFOS channel, exactly two leptons with p T larger than 22 GeV and 15 GeV are required. Only opposite-sign eµ final states are considered in order to reduce the background from Z+jets, W Z and ZZ events. A cut on the invariant mass of the lepton pair, m 0 1 > 10 GeV, is applied to reject combinatorial dilepton backgrounds. In order to select final states with neutrinos, E miss T is required to be above 20 GeV. These selections reduce the background processes that contain jets faking leptons. The presence of at least two jets with p T > 25 GeV is required. The background from top-quark production is reduced by vetoing events if they contain any b-tagged jets with p T > 20 GeV. To reject the Z+jets production that leads to eµ final states through Z → τ τ decay, a requirement of m τ τ < (m Z −25 GeV) is applied, where m τ τ is the dilepton invariant mass reconstructed using the collinear approximation [22], namely under the assumptions that the lepton pair originates from τ lepton decays, the neutrinos are the only source of E miss T and they are collinear with the charged leptons.
Upper bounds on the invariant mass of the lepton pair, m 0 1 < 50 GeV, and the azimuthal angular separation of the lepton pair, ∆φ 0 1 < 1.8 rad, are applied to enhance the Higgs boson signal relative to the W W , tt and W +jets backgrounds. Requirements on the rapidity separation between the two leading jets, ∆y jj < 1.2, and the invariant mass of the two leading jets, |m jj − 85 GeV| < 15 GeV, are introduced to select jets from the associated W/Z bosons. The central value of the m jj selection interval is larger than the W boson mass in order to retain the acceptance for ZH production with Z → qq decay.
The selection m T < 125 GeV is applied, where m T is the transverse mass of the dilepton system and E miss T , defined as The selections on ∆y jj and m jj make this channel orthogonal to the ggF-enriched n j ≥ 2 category in ref. [22], while orthogonality with respect to the VBF category is ensured by explicitly vetoing the BDT signal region of the VBF analysis [22]. In the fit the 2 -DFOS channel enters as a single signal region.

Same-sign two-lepton channel
In the 2 -SS channel, exactly two leptons with the same charge are required. Lower bounds on lepton p T are set to 22 GeV and 15 GeV and both the same-flavour and different-flavour combinations are considered. A lower bound on m 1 2 is applied at 12 GeV for sameflavour lepton pairs and at 10 GeV for different-flavour lepton pairs. Despite the samecharge requirement, a wrong-charge assignment may allow background contributions from Z boson decays. Therefore a veto on same-flavour lepton pairs with |m 1 2 −m Z | < 15 GeV is introduced.
The 2 -SS2jet and 2 -SS1jet SRs require the number of jets to be exactly two or exactly one, respectively. Events with b-tagged jets having p T > 20 GeV are discarded. The E miss T is required to be larger than 50 GeV in the 2 -SS2jet SR and larger than 45 GeV in the 2 -SS1jet SR. Additional cuts are applied to events in the 2 -SS2jet and 2 -SS1jet SRs, on the following variables (see table 2  minimises the above variable and a jet, ∆φ min i j ; the transverse mass of the leading lepton and the E miss T , m lead )), where p T,lead and φ lead are respectively the transverse momentum and φ angle of the leading lepton. Lower values of m min i j (m min i jj ) and of ∆φ min i j favour Higgs boson decays relative to the major backgrounds. High values of m lead T help in reducing W +jets background. The p T threshold for the sub-leading muon in the µµ channel is increased to 20 GeV in both the SRs to suppress misidentified muons from W +jets and multijet production.
In the fit, the 2 -SS2jet and 2 -SS1jet SRs are further split into four signal regions according to the combination of lepton flavours in each event: ee, eµ, µe and µµ, where eµ refers to the case in which the electron has leading p T while µe refers to the case in which the muon has leading p T . This splitting is motivated by the expected differences in the background contributions, for example W γ, which is expected to be zero in the µµ channel but not in the other channels.

Signal acceptance
The number of expected V H(H→ W W * ) events surviving the event selections is presented for each channel in table 3. The total acceptance for W H(H → W W * → ν ν), W H(H → W W * → νqq) and ZH(H → W W * → ν ν) is 3.7%, 0.3% and 1.9%, respectively. The analysis acceptance for the ZH(H → W W * → νqq) process is negligible. The acceptance is defined as the ratio of the number of events in the SRs to the number of events expected according to the branching fractions for the various processes. Associated Higgs boson production followed by the decay H → τ τ cannot be completely isolated from the selected final states. Therefore the results presented in this paper, with the exception of section 8.5 for consistency of the analysed model, include this process as part of the background, with the production cross section (σ V H ) and Br(H → τ τ ) fixed to the SM value.

Background modelling
The background contamination in the signal regions results from various physics processes, each modelled by one of the following methods: • Pure MC prediction: rates and differential distributions (shapes) are extracted from simulation and normalised to the cross sections in table 1; • MC prediction normalised to data: rates are extracted from data in control regions but shapes are extracted from simulation; • Pure data-driven prediction: rates and shapes are extracted from data.
Misidentified-lepton backgrounds (W +jets, multijets) in the 2 channels are estimated by using a purely data-driven method, which utilises the rate at which a jet is misidentified as a lepton [22]. Table 4 summarises the method adopted for each process in each signal region. The labels "MC" and "Data" represent the pure MC prediction and the pure datadriven estimation. For backgrounds modelled by simulation with a normalisation factor -14 -JHEP08 (2015)  (NF) computed using data, the names of relevant control regions are shown as defined in tables 5 and 6. The ratio of tt yields to tW yields is found to be compatible between all the CRs and associated SRs, thus only one NF is computed per CR for the "Top" category. The ggF and VBF productions of Higgs bosons are treated as background as discussed in section 8.4. Definitions of control regions in the 4 and 3 analyses are presented in table 5, and those defined in the 2 analyses are shown in table 6. The CRs are made orthogonal to the corresponding SRs by inverting some selections with respect to the SR definitions. Such selections are in boldface font in the tables and are further explained in the following sections.

Background in the four-lepton channel
The main backgrounds that contribute to the 4 -2SFOS and 4 -1SFOS SRs are diboson processes, dominated by ZZ * with E miss T from Z → τ τ decay, and triboson processes, in particular ZW W * , which has the same signature as the signal. These processes respectively account for about 85% and 15% of the total background contamination. To normalise ZZ * a dedicated CR, the 4 -ZZ CR, is defined by inverting the requirement on the invariant mass of dileptons from the Higgs boson candidate. All the other minor background processes, listed in table 4, are modelled by simulation.

Background in the three-lepton channel
Three classes of backgrounds contribute to the 3 channel. The first class comprises diboson processes: W Z/W γ * , ZZ * with an undetected lepton mainly due to its low-p T , and Zγ, in which the photon converts to electron-positron pairs. The ZZ * contribution in this channel is mainly due to single-resonant ZZ * production where the three-lepton invariant Table 4. Summary of background modelling."V V V " represents the triboson processes W W W * , ZW W * , ZZZ * and W W γ * . "Top" processes include tt and single-top production dominated by tW with W → ν decay, as well as ttW/Z. Some backgrounds are normalised by rescaling the MC yields by the data-to-MC ratio measured in CRs. For these backgrounds the names of the most important CRs are listed. The symbol "-" denotes a negligible contribution to the total background in the signal region.    mass is just below the Z boson mass. The second class includes triboson processes, mainly W W W * . The last class of backgrounds are processes with a misidentified lepton, mainly Z+jets and top-quark pair production.
In the 3 -3SF and 3 -1SFOS SRs, W Z/W γ * and ZZ * represent the leading background contributions accounting for about 80% of the total background yields, with 65% from W Z/W γ * and 15% from ZZ * . Production of Zγ, V V V , Z+jets and top-quarks share the remaining background fraction equally. The 3 -0SFOS SR contains contributions of similar size from W Z/W γ * , V V V and top-quark production. In this SR the total background event yield is about eight times lower than in the 3 -3SF and 3 -1SFOS SRs.
A 3 -W Z CR is defined by reversing the Z-veto requirement, in order to select events with a Z boson decay. The 3 -ZZ CR and 3 -Zγ CR are defined by requiring low E miss T values to reflect the absence of final-state neutrinos in the background process under study. For these control regions, the invariant mass of the three leptons must be consistent with the Z boson mass. These regions are further distinguished according to the flavour combination of the three leptons, namely eee or µµe for the 3 -Zγ CR, and µµµ or eeµ for the 3 -ZZ CR.
The 3 -Zjets CR is defined by reversing the E miss T and the Z-veto selections. The properties of misidentified electrons and muons are different; therefore the 3 -Zjets CR is further split into the misidentified-electron component (eee+µµe events) and misidentifiedmuon component (µµµ + eeµ events) and an NF is assigned to each component. In the 7 TeV data sample, the predicted Z+jets event yield with misidentified muons in the 3 SRs is negligible. Furthermore, the number of events in the 3 -Zjets CR with such a lepton -17 -

JHEP08(2015)137
flavour combination is too small to reliably extract the NF. Therefore the estimation of the misidentified-muon component is taken directly from simulation.
The 3 -Top CR is defined by requiring at least one b-tagged jet. The Z+jets contribution is difficult to isolate from other processes that include Z bosons. Thus the NF for this process is constrained not only by the 3 -Zjets CR, but also in part by the 3 -W Z CR and the 3 -ZZ CR, as indicated in table 4.

Background in the opposite-sign two-lepton channel
The dominant background in this channel is top-quark production, which accounts for about 50% of the total contamination. The 2 -OSTop CR is defined by requiring a high invariant mass of the lepton pair in the final state. As the b-jet rejection criteria are the same in the CR and SR, the systematic uncertainties related to b-tagging largely cancel between the two regions. The second dominant background is Z → τ τ , which accounts for 20% of the total background in the SR. A dedicated control region, 2 -Zτ τ CR, is defined by requiring a large opening angle between the two leptons. The W W process constitutes the third largest background, accounting for 10% of the total. Due to a difficulty in separating this process from tt events over a wide kinematic region, no dedicated CR is defined, and this process is modelled purely by MC simulation.
The contribution of backgrounds with misidentified leptons, W +jets and multijet, accounts for 10% of the total background. The misidentified-lepton background rate has an uncertainty of 40%. Due to this large uncertainty, the misidentified-lepton background contributes significantly to this channel. The W Z/W γ * production and the ggF production followed by H→ W W * decay, each representing 5% of the total background, are modelled with MC simulation.

Background in the same-sign two-lepton channel
The W Z/W γ * and W +jets processes each account for one third of the total background. Some of the W Z/W γ * events with three leptons enter the selection when one of the leptons escapes detection. To normalise this process, the 2 -W Z CR is defined by selecting events with three leptons. The contamination from W +jets events with one misidentified lepton is estimated by using the same data-driven method used in the 2 -DFOS channel.
The remaining background processes contribute at the 10% level or less. The normalisation of W γ is based on the 2 -W γ CR, defined by requiring at least one electron consistent with a conversion, including a requirement that the electron does not have a hit in the innermost pixel layer. The background contribution to the 2 -SS channel due to lepton charge misidentification, in otherwise charge-symmetric processes, is found to be relevant only for electrons and affects top-quark production, opposite-sign W W and Z+jets. It represents 10% of the total background in the 2 -SS1jet SR and 3% of the total background in the 2 -SS2jet SR. The 2 -SSTop CR, 2 -W W CR and 2 -Zjets CR are defined selecting opposite-sign leptons to normalise these contributions. Moreover, in 2 -SSTop CR at least one b-tagged jet is selected. Due to the small production rate, no control region is defined to normalise the W W events from vector boson scattering with same charge, whose rate is taken directly from simulation.  Bkg. Uncert.
Others Top Bkg. Uncert. The hatched area on the histogram represents total uncertainty, both statistical and systematic (see section 7), on the total background estimate. The last bin includes overflows.

Normalisation factors and composition of control regions
NFs are computed through the signal extraction fit explained in section 8.4. Table 4 lists the main background processes, along with the CRs that contribute to the determination of their NFs. The NFs, which are specific to each signal region, are fitted taking into account only the total number of expected and observed events in each CR, separately for the 8 TeV and 7 TeV data samples, and are summarised in  Bkg. Uncert. In these tables and figures, each background process normalised using CRs is presented separately, while backgrounds that are not normalised using CRs are grouped together as "Others". The ggF and VBF production of Higgs bosons, and the V H(H → τ τ ) process are included in the "Others" category, assuming m H = 125 GeV and the SM value for the cross sections and for the branching fraction.

Systematic uncertainties
The sources of theoretical and experimental systematic uncertainty on the signal and background are described in this section, and summarised in table 9. The table shows the uncertainties on the estimated event yield after the fit in 8 TeV data samples. Similar values are obtained for the 7 TeV data analysis.
Systematic uncertainties have an impact on the estimates of the signal and background event yields in the SRs and CRs. The experimental uncertainties are applied to both the SRs and CRs. The extrapolation parameter from a SR to a CR is defined as the ratio of MC estimates in the SR and CR. The theoretical uncertainties are computed on the extrapolation parameters, and applied to SRs. Correlations between SRs and CRs are taken into account in the fit, and cancellation effects are expected for uncertainties correlated between SRs and CRs. The different uncertainty sources are treated according to their correlations across the data taking periods, among different analyses, between signal and background sources. Whenever an effect is expected to affect coherently the event yields in two event samples (SRs and/or CRs), for instance the muon reconstruction efficiency for both signal and background, a 100% correlation is assumed and the yields are varied coherently in the fit through the introduction of a single parameter (nuisance parameter). Alternatively when an uncertainty source is not expected to affect the event yields coherently, for instance trigger and luminosity uncertainties across different data taking years, different nuisance parameters are introduced to represent uncorrelated effects.
The "MC statistics" and "CR statistics" uncertainties in table 9 (b) arise from limited simulated events in the SRs and CRs and from the number of data events populating the CRs, respectively.

Theoretical uncertainties
The theoretical uncertainties on the total Higgs boson production cross section and branching fraction are evaluated by following the recommendation of the LHC Higgs cross-section working group [21,44,45]. Uncertainties concerning QCD renormalisation and factorisation scales, which are hereafter collectively referred to as QCD scales, PDF, the value of α S and branching fraction are estimated.   Table 9. Theoretical and experimental uncertainties, in %, on the predictions of the (a) signal and (b) total background for each category. Fake factor refers to the data-driven estimates of the W +jets and multijet backgrounds in the 2 channels. The dash symbol (-) indicates that the corresponding uncertainties either do not apply or are negligible. The values are obtained through the fit and given for the 8 TeV data sample. Similar values are obtained for the 7 TeV data sample.

JHEP08(2015)137
The main uncertainty on the V H (H→ W W * ) process, shown in the "V H Acceptance" row in table 9(a), accounts for the uncertainties in the acceptance of the signal processes and it is evaluated with MINLO-V Hj Powheg-Box1.0 [72] simulation. The leading contributions are the missing next to leading order QCD contributions in qq → V H simulated with Pythia8 and the parton shower uncertainty. The first is evaluated by comparing the Pythia8 prediction with the MINLO-V Hj interfaced with Pythia8 and amounts to 7%. The second is computed by comparing MINLO-V Hj interfaced with Pythia8 with MINLO-V Hj interfaced with Herwig and accounts for a further 7%.
The cuts on the number of jets are found to be the main source of such uncertainties. The uncertainty on the acceptance of the gg → ZH process is estimated to be 5% in the 4 channel, the only channel where this process is relevant.
Uncertainty on the Higgs boson branching fraction to W W * , which is particularly important for the V H and VBF production modes, amounts to 4%. The QCD scale uncertainty is 1% for W H production and 3% for ZH production. This uncertainty is larger for ZH production due to the gg → ZH contribution. The "VH NLO EW corrections" refers to additional uncertainties on the corrections [48] to the NLO differential cross section, applied as a function of the p T of the associated weak bosons in the LO W H and ZH production modes generated by Pythia8. The size of this uncertainty is 2%.
The uncertainties on ggF production are due to the inclusive cross section dependence from QCD scales, the PDF choice and α S , they range from 7% to 8%. In the 2 -DFOS channel, an additional uncertainty due to the jet multiplicity cut is computed with MCFM [55] and amounts to about 12%.
The uncertainties from the QCD scales of backgrounds are estimated by varying the scales up and down independently by a factor of two. MCFM is used to estimate this uncertainty in the 4 , 3 and 2 -SS channels. In the 4 channel the uncertainty on ZZ * is 4% in the SRs and CR. However, due to the cancellation between the SRs and CR, the resultant effect becomes negligible. In the 3 -3SF and 3 -1SFOS SRs the QCD scale uncertainty on W Z/W γ * is determined in each bin of the "BDT Score" and ranges between 3% and 6%. In the 2 -SS channel, due to cancellations, the QCD scale uncertainties on W Z/W γ * and W γ are found to be negligible with the exception of the 2 -SS2jet SR, in which 100% uncertainty is assigned to W γ. The QCD scale uncertainty on the same-sign W W +2jet, estimated using VBF@NLO [73], is of the order of 40%. In the 2 -DFOS channel, QCD scale uncertainties on top-quark and W W +2jet production with at least two QCD couplings, referred to as QCD W W in the following, are estimated as 9% and 17% by using MC@NLO and MadGraph, respectively.
The PDF uncertainties on backgrounds are calculated by following the PDF4LHC recipe [74] using the envelope of predictions from MSTW2008, CT10 and NNPDF2.3 PDF sets with the exception of top-quark production in the 2 -DFOS channel, which is evaluated with the same technique used in the ggF-enriched n j ≥ 2 category in ref. [22]. The uncertainties range from 1% to 6%, depending on the background process and the categories. An uncertainty of 33% on the NLO K-factor for the triboson process is evaluated by using VBF@NLO; in the 3 -0SFOS SR this uncertainty is estimated in bins of ∆R 0 , 1 and ranges from 1% to 6%.

JHEP08(2015)137
The "MC modelling" row in table 9(b) takes the yield variation observed between predictions of different MC generators. In the 3 and 2 -SS channels, the uncertainties on the W Z/W γ * event yield due to the modelling of the underlying event, parton shower and matching of the matrix element to the parton shower are evaluated by comparing the predictions of Powheg-Box+Pythia and MC@NLO+Herwig. In the 2 -DFOS channel, an uncertainty due to underlying event and parton shower modelling is assigned to top-quark production and Z → τ τ by comparing the expectations from Powheg-Box+Pythia and Powheg-Box+Herwig, and Alpgen+Pythia and Alpgen+Herwig, respectively. The QCD W W event yield in the 2 -DFOS channel is estimated using Sherpa. The uncertainty from the underlying event, parton shower modelling and matrix element implementation is evaluated through a comparison with MadGraph+Pythia. The uncertainty on the main background in the 4 channel, ZZ * , is dominated by the statistical component; a systematic uncertainty from the different models, underlying event and parton shower is assigned through a comparison between Powheg-Box+Pythia and Sherpa.

Experimental uncertainties
One of the dominant experimental uncertainties, labelled "Jet" in table 9, derives from the propagation of the jet energy scale calibration and resolution uncertainties. They were derived from a combination of simulation, test-beam data, and in situ measurements [75]. Additional uncertainties due to differences between quark and gluon jets, and between lightand heavy-flavour jets, as well as the effect of pile-up interactions are included. For jets used in this analysis, the jet energy scale uncertainty ranges from 1% to 7%, depending on p T and η. The relative uncertainty on the jet energy resolution ranges from 2% to 40%, with the largest value of the resolution and relative uncertainty occurring at the p T threshold of the jet selection.
The "Muon" and "Electron" uncertainties include those from lepton reconstruction, identification and isolation, as well as lepton energy and momentum measurements. The "Trigger efficiency" uncertainty in table 9 refers to the uncertainty on the lepton trigger efficiencies. The uncertainties on the lepton and trigger efficiencies are of the order of 1% or smaller.
The changes in jet, electron and muon energy scale uncertainties due to varying them by their systematic uncertainties are propagated to the E miss T evaluation and included in the "Jet", "Electron" and "Muon" rows in table 9. An additional "E miss T soft term" uncertainty is associated with the contribution of calorimeter energy deposits not assigned to any reconstructed objects in the E miss T reconstruction [68][69][70]. The "b-tagging efficiency" row refers to the uncertainties on the efficiency of tagging of b-jets and include contributions from b-jet identification and charm and light-flavour jet rejection factors [67,76]. The uncertainties related to b-jet identification range from <1% to 8%. The uncertainties on the misidentification rate for light-quark jets depend on p T and η, and have a range of 9-19%. The uncertainties on c-jets reconstructed as b-jets range between 6% and 14% depending on p T .

JHEP08(2015)137
The uncertainty labelled as "Fake factor" is associated with the data-driven estimates of the W +jets and multijet backgrounds in the 2 channels; it ranges between 35% and 45% depending on the sample and on categories. A "Charge mis-assignment" systematic uncertainty is estimated to account for the mismodelling of the charge flip effect by comparing the number of lepton pairs with same charge and opposite charge under the Z boson mass peak in data and MC simulation, resulting in a 16% relative uncertainty. The uncertainty is assigned to W Z/W γ * in the 3 -0SFOS SR, and top-quark production, Z+jets and W W in the 2 -SS channel.
The uncertainty labelled as "Photon conversion rate" is assigned to W γ in the 2 -SS channel, and is evaluated by comparing the yield in data and in MC simulation for events with two muons and one electron with no hit on the innermost pixel detector layer. It is relevant only for the 2 -SS channel and has a size of 6.5% for W γ.
The "Pile-up" field in table 9 includes the uncertainty on the weights applied to all simulated events to match the distribution of the number of pile-up interactions to that of data. The uncertainty on the integrated luminosity for the 2012 data is ± 2.8% and it is derived following the same methodology as that detailed in ref. [77]. For the 2011 data the uncertainty on the integrated luminosity is ± 1.8% [77]. The backgrounds normalised with CR data are not affected by the luminosity uncertainty.
The dominant systematic uncertainties on the V H(H→ W W * ) process in the 4 and 3 channels are due to uncertainties on lepton reconstruction and on the jet energy scale and resolution. In the 2 channels, the jet energy scale and resolution uncertainties are the most important.

Results
The data collected at √ s = 8 TeV and 7 TeV are analysed separately and then combined in all channels in order to search for the Higgs boson in W H and ZH production using H→ W W * decays. The analyses are optimised for a Higgs boson mass of m H = 125 GeV. For this mass, the selected Higgs bosons decay mainly as H → W W * → ν + X, but a small contamination from the V H production of Higgs bosons, from the decay chain H → τ τ → ν ν, is present. This contribution is treated as background and normalised to the SM expectation for the V H production cross section σ V H , and the H → τ τ branching fraction.
This section is subdivided in five sub-sections. Section 8.1 shows the event yields in each category of the 8 TeV and 7 TeV data samples. Furthermore, distributions of some of the relevant variables are shown for the 8 TeV data sample. Section 8.2 summarises the statistical treatment used for the signal extraction. Section 8.3 quantifies the agreement with the background only hypothesis in each analysis category and evaluates the observed and expected significance in each of them. Section 8.4 reports the significance for W H, ZH and combined V H production, moreover it shows the measurement of their cross sections divided by their SM expectations (µ). The V H categories are then combined with the categories of the ggF and VBF analysis using H → W W * → ν ν decays described in -28 -

JHEP08(2015)137
ref. [22]. In addition the ggF and VBF yields are extracted in a consistent manner together with the V H yields. Finally, section 8.5 shows the measurement of the couplings to vector bosons and to fermions obtained from the combination of the analyses sensitive to the three production modes using H → W W * decays.

Event yields and distributions
The number of events in each category is summarised in table 10 and in table 11 for the 8 TeV and 7 TeV data sample, respectively. In the 2 -DFOS and 2 -SS1jet categories the numbers of observed events are slightly larger than the expectation. The lepton flavour composition of the events in the 2 -SS channel is shown in table 12, in which the high event yield of the 2 -SS1jet SR mainly originates from the µµ and µe channels. The behaviour of the observed data yield as a function of the data period was studied. The events were uniformly distributed according to the acquired luminosity as expected, excluding temporary failures of the detector subsystems as an explanation of the excess. Moreover it was checked that known detector defects were not increasing the rate of particular background sources. Finally, the kinematic distributions of the events were analysed in order to look for striking features pointing to some particular missing background contribution. Because no particular problem was found, the data excesses were attributed to statistical fluctuations.
Distributions of some of the relevant variables after the event selection are presented for 8 TeV data in figure 6 for the 4 analyses, in figure 7 for the 3 analyses and in figure 8 for the 2 analyses. Figures 6(a)

Statistical method
The signal extraction is performed using the profile likelihood ratio method [78], which consists of maximising a binned likelihood function L(µ, θ | n). The likelihood is the product of Poisson distributions for each SR and CR. The mean values of the distributions are the sum of the expected yields of signal and background. The symbol n represents the observed events in each SR and CR. The signal and background expectations are functions of the signal-strength parameter, µ, and a set of nuisance parameters, θ. The signal strength µ multiplies the SM predicted signal event yield in all categories, while background normalisation factors, included as nuisance parameters, represent corrections for background sources normalised to data. Signal and background predictions are affected by systematic -35 -JHEP08(2015)137 uncertainties that are described by nuisance parameters. The normalisation factors are left free in the fit, while the constraints on the systematic uncertainties are chosen to be log-normal distributions.
The test statistic q µ is defined as The symbolθ µ indicates the nuisance parameter values at the maximum of the likelihood function for a given µ. The denominator is the maximum value of L obtained with both µ and θ floating. When the denominator is maximised, µ takes the value ofμ. The p 0 value is computed for the test statistic q 0 evaluated at µ = 0 in eq. (8.1), and is defined to be the probability to obtain a value of q 0 larger than the observed value under the backgroundonly hypothesis. There are no bounds onμ, although q 0 is defined to be negative ifμ ≤ 0. The equivalent formulation, expressed in terms of the number of standard deviations σ, is referred to as the local significance Z 0 . The signal acceptance for all production modes and decays are computed assuming m H = 125.36 GeV, which is the mass measured in H → γγ and H → 4 decays by ATLAS [79]. The acceptance for this mass is obtained interpolating between the values computed at m H = 125 and 130 GeV.

Signal significance extraction and determination of signal strengths
The V H-targeted categories are then combined with the categories of the ggF and VBF analysis using H → W W * → ν ν decays described in ref. [22]. The combination is again performed by building a likelihood function that includes the SRs and CRs of the ggF, VBF and V H analyses. The experimental and theoretical uncertainties affecting the same sources are correlated among different production modes. The µ values for each production mode (µ ggF , µ VBF , µ V H ) are correlated in all categories and fitted together while the background NFs are uncorrelated among the different analyses as they cover different phase-space regions. Therefore, when extracting µ V H , µ W H and µ ZH , the ggF and VBF productions are treated as background and their yields determined by the global fit.  The fit results for the µ values for the W H, ZH and V H production are: ). The uncertainties on the µ V H value are shown in table 14 (see section 7 for their description). The derivative of µ V H with m H has been evaluated to be -5.8 %/GeV at m H = 125.36 GeV. Figure 9 shows the value of the test statistic as a function of µ W H and µ ZH ; as shown, the correlation between the two parameters is weak. Table 15 summarises the signal strengths for each production mode and their combination at a value of m H = 125.36 GeV, together with the observed and the expected Z 0 . The combined signal strength is µ = 1.16 +0.16 −0.15 (stat.) +0.18 −0.15 (sys.), and the significance of the excess respect to the background only hypothesis is 6.5 σ, while the expected significance in the presence of a Higgs boson decaying to W W * is 5.9 σ. Figure 10 shows the value of the test statistic as a function of the signal strength of each production mode (µ ggF , µ VBF , µ V H ) and as a function of the combined signal strength µ HW W .
The obtained values are all compatible with the SM expectation within 1.4 standard deviations. Figures 11(a) and 11(b) show the two-dimensional dependence of the likelihood on µ V H and µ ggF , and on µ V H and µ VBF . The µ values not displayed are kept as free unconstrained parameters in the fit. The correlation between the parameters shown in figure 11 is small. The central values obtained for µ ggF and µ VBF are slightly different from those reported in ref. [22]. The shift represents a few percent of the quoted errors and is pulled up by the presence of a small contamination by V H(H→ W W * ) events in the 1-jet and 2-jets categories of the ggF and VBF analysis.   Table 14. Percentage theoretical and experimental uncertainties on the observed V H signal strength µ V H . The contributions from signal-related and background-related theoretical uncertainties are specified. The "VH acceptance" is evaluated using both the qq → (W/Z)H and the gg → ZH production. The statistical uncertainty due to the ggF and VBF subtraction measured in the categories of the ggF and VBF analysis are indicated with "ggF SR statistics" and "VBF SR statistics", for the contribution from the signal regions, and "ggF+VBF CR statistics" for the contribution from the control regions. The row "MC statistics" shows the uncertainty due to the statistics of the simulated samples. The values are obtained from the combination of the 8 TeV and 7 TeV data samples.

Measurement of the couplings to vector bosons and fermions
The values of µ ggF , µ VBF and µ V H can be used to test the compatibility of the bosonic and fermionic couplings of the Higgs boson with the SM prediction using the formalism developed in ref. [21]. Assuming the validity of the SU(2) custodial symmetry and a universal scaling of the fermion couplings relative to the SM prediction, two parameters are defined: the scale factor for the SM coupling to the vector bosons (κ V ) and the scale factor for the coupling to the fermions (κ F ). Loop-induced processes are assumed to scale as in the SM. The H → τ τ contribution is treated as signal and its yield is parameterised as a function of κ V and κ F . The total width of the Higgs boson can be expressed as the sum of the different partial widths, each one rescaled by the square of the appropriate scaling factor. Neglecting the small contribution from Γ(H → γγ) and rarer decay modes, the H→ W W * decay branching fraction is expressed as: where Γ SM (H → f f ), Γ SM (H → gg) and Γ SM (H → V V ) are the SM partial decay widths to fermions, gluons and weak bosons, respectively. The ggF (gg → H) process depends directly on the fermion scale factor κ 2 F through the top and bottom quark loops, while the VBF (qq → Hqq) and V H(qq → V H) production cross sections are proportional to κ 2 V , as expressed by the following relations: where the σ without subscript indicates the (κ V , κ F )-dependent cross sections and σ SM represents the SM cross sections. The gg → ZH production cross sections are more complex functions of both κ V and κ F [80]: The signal event yield is expressed as σ · Br(H→ W W * ) using the narrow-width approximation. Only the relative sign between κ V and κ F is observable and hence in the following only κ V > 0 is considered, without loss of generality. Sensitivity to the sign results from negative interference, in the gg → ZH process, between the box diagram in which both the Z and H bosons are produced directly from the heavy-quark loop and the triangle diagram in which only the Z * is produced and subsequently radiates a Higgs boson [81]. Because the relative weights of such processes depend on the √ s of the interaction, different coefficients appear in the expression for 8 and 7 TeV. The likelihood dependence on κ V and κ F is shown in figure 12. The product σ(gg → H) · Br(H→ W W * ), which is measured with good accuracy, does not depend on |κ F | in the limit |κ F | κ V . This explains the low sensitivity to high values of κ F . On the other hand µ VBF and µ V H , as measured for the H→ W W * decay, should vanish in the limit |κ F | κ V due to the increased value of the Higgs boson total width and the consequent reduction of the H→ W W * branching fraction. The observation of significant excesses in the VBF and V H production modes therefore leads to an exclusion of the |κ F | κ V region. The fit to the data results in two local minima and, although the negative κ F solution is preferred to the positive solution at 0.5σ, the observed results are compatible with the SM expectation, and the best fit values are:

Conclusions
A search for the Standard Model Higgs boson produced in association with a W or Z boson and decaying into W W * is presented. Associated W H production is studied in the final states in which the three W bosons decay to leptons or where one W boson decays to hadrons while the others decay leptonically. The two-lepton and four-lepton final states are used to search for ZH production. The dataset corresponds to integrated luminosities of 4.5 fb −1 and 20.3 fb −1 recorded by the ATLAS experiment with LHC proton-proton collisions at √ s = 7 TeV and 8 TeV, respectively. For the Higgs boson mass of 125.36 GeV, the observed (expected) deviation from the background-only hypothesis, which includes the Standard Model expectation for H → τ τ , corresponds to a significance of 2.5 (0.9) standard deviations. The ratio of the measured signal yield to its Standard Model expectation for the V H production is found to be µ VH = 3.0 +1.3 −1.1 (stat.) +1.0 −0.7 (sys.). A combination with the gluon fusion and vector boson fusion analyses using the H → W W * → ν ν decay is also presented. Including V H production the observed significance for a Higgs boson decaying to W W * is 6.5 σ with an expectation of 5.9 σ for a Standard Model Higgs boson of mass  [8] ATLAS collaboration, Measurements of fiducial and differential cross sections for Higgs boson production in the diphoton decay channel at    [76] ATLAS collaboration, Calibration of b-tagging using dileptonic top pair events in a combinatorial likelihood approach with the ATLAS experiment, ATLAS-CONF-2014-004, CERN, Geneva Switzerland (2014).
[77] ATLAS collaboration, Improved luminosity determination in pp collisions at √ s = 7 TeV using the ATLAS detector at the LHC, Eur. Phys. J. C 73 (2013)  [80] ATLAS collaboration, Measurements of the Higgs boson production and decay rates and coupling strengths using pp collision data at √ s = 7 and 8 TeV in the ATLAS experiment, arXiv:1507.04548 [INSPIRE]. -48 -