Higgs search in H$\to$ZZ/WW decay channels with the CMS detector

Abstract

A prospective analysis for the search of the Standard Model Higgs boson decaying in vector boson pairs is presented with the CMS experiment in the context of the initial luminosity at the CERN LHC pp collider. Monte Carlo data corresponding to an integrated luminosity of up to 1 fb$^{-1}$ are analysed and the expected significance for a Standard Model-like Higgs boson in these channels is established. Comment: 4 pages, 4 figures, proceedings of the Lake Louise Winter Institute 2009, 16th-21st February 2009, Alberta, Canada

Full text

Page 1

Higgs search in H→ZZ/WW decay channels with the
CMS detector
C. Charlot, for the CMS collaboration
Laboratoire Leprince-Ringuet, Ecole Polytechnique and IN2P3/CNRS, Palaiseau, France
Abstract.
in vector boson pairs is presented with the CMS experiment in the context of the initial
luminosity at the CERN LHC pp collider. Monte Carlo data corresponding to an integrated
luminosity of up to 1 fb−1are analysed and the expected significance for a Standard Model-like
Higgs boson in these channels is established.
A prospective analysis for the search of the Standard Model Higgs boson decaying
1. Introduction
The Standard Model (SM) of electroweak and strong interactions predicts the existence of a
single physical Higgs boson, the quantum of the scalar field responsible for electroweak symmetry
breaking. Direct searches for the SM Higgs particle at the LEP e+e−collider have lead to a
lower mass bound of mH> 114.4GeV/c2(95% CL) [1]. Ongoing direct searches at the TeVatron
p¯ p collider by the D0 and CDF experiments set constraints on the production cross-section for
a SM-like Higgs boson in a mass range extending up to about 200GeV/c2[2, 3]. A consistency
fit including all the measured electroweak observables favours the mass range mH< 182GeV/c2
(95% CL) [4]. The inclusive production of SM Higgs bosons followed by the decay into di-bosons
and subsequently into leptons, H → ZZ(∗)→ lll?l?and H → WW(∗)→ lνll?νl? with l,l?= e or µ,
are expected to be early discovery channels at the CERN LHC pp collider over a wide range of
possible mHvalues. We present here the analysis strategies for the the Higgs search in di-boson
decay channels and leptonic modes in the context of an initial luminosity of 1 fb−1. Signal and
background datasets obtained with a detailed Monte Carlo simulation of the detector response,
including the limited inter-calibration and alignment precision expected at startup, are treated
using a complete reconstruction chain. Emphasis is put on the reduction of distinguishable
background rates and on methods allowing for a data-driven derivation of experimental and
background systematic uncertainties.
2. The CMS detector
A general description of the CMS detector can be found elsewhere [5]. This analysis relies mostly
on the tracker and the electromagnetic calorimeter (ECAL), both immersed in a 4T magnetic
field parallel to the z axis, and on the muon spectrometer hosted in the iron magnet return yoke.
The CMS tracker is a cylindrical detector equiped with silicon pixel detectors for the innermost
part and silicon strip detectors for the outer layers. The tracker acceptance for a minimum of 5
collected hits extends up to pseudorapidities η of about |η| < 2.4. The CMS ECAL is made of
quasi-projective PbWO4crystals with a granularity of approximately ∆η×∆φ = 0.0175×0.0175
in the barrel part (|η| ≤ 1.48) and of approximately ∆x×∆y = 1.3RM×1.3RMin the endcaps
arXiv:0906.1666v1 [hep-ex] 9 Jun 2009

Page 2

parts. The endcaps are equipped with a preshower device that cover the region 1.6 < |η| < 2.6.
The electron reconstruction efficiency varies from 85% to 95% depending on pTand η. The muon
spectrometer consists of Drift Tubes (DTs), Cathode Strip Chambers (CSCs), and Resistive
Plate Chambers (RPCs), which cover the angular region |η| < 2.4. The muon reconstruction
efficiency varies between 95% and 99%.
3. The WW(∗)→ lνll?νl? analysis
The analysis in this channel is performed using either a simple cut based approach and a more
involved multi-variate analysis based on a neural network (NN). Events passing leptonic trigger
paths with two identified and isolated high-pTleptons (e or µ) are selected. Standard lepton
reconstruction techniques are used, and the identification of electrons is relatively tight to reduce
the contamination from W+jets processes. The charged isolated leptons identified in this way
are combined into all possible pairs requiring leptons of opposite charge and within |η| ≤ 2.5
and pT ≥ 10 GeV or at least one lepton with pT ≥ 20 GeV. If none or more than one such
pair is found the event is rejected so to suppress WZ and ZZ backgrounds. A minimum missing
transverse energy of 30 GeV is required in accordance with the presence of two neutrinos in the
final state. An event containing any jet with pT> 15 GeV and |η| < 2.5 is rejected. This cut
removes the bulk of the t¯t background. Finally, m?? > 12 GeV is required to select events with
leptons coming from fully leptonic W pair decays. Additional variables are used in the final
kinematical selection both in the cut based and in the NN analysis: the angle ∆Φ??between the
two leptons in the transverse plane; the invariant mass of the lepton pair (e+e−and µ+µ−final
states); the missing transverse energy and the transverse momenta of the harder and the softer
lepton. In the NN analysis, following additional variables are used: the separation angle ∆η??
between the leptons in η, the transverse mass of both lepton-missing transverse energy pairs, the
η angle of both leptons, the angle in the transverse plane between the missing transverse energy
and the closest lepton and the flavour of the di-lepton final state. Systematic uncertainties
play an important role in this analysis where no strong mass peak is expected due to the
presence of two neutrinos in the final state. Experimental systematic uncertainties coming from
the luminosity measurement, lepton identification and efficiencies, missing transverse energy
resolution, jet efficiency and energy scale have been taken into account. The normalization
of the two main backgrounds, t¯t and W+W−has been addressed using various data driven
methods. The overall relative error depends on the final state and on the Higgs mass. It is
estimated as about 11% for the signal and 21% for the background. Fig. 1 (left) shows the neural
network outputs for the signal and the backgrounds for mH= 170 GeV/c2. The distributions
are representative of other mass regions. There is a clear shape difference between signal and
background events, although there is no region completely free of background. The vertical line
indicate the cut value used.
4. The ZZ(∗)→ lll?l?analysis
This channel is caracterized by the presence of two pairs of isolated primary electrons or muons,
with one pair generally resulting from the decay of a Z boson on its mass shell. After the
High Level Trigger, the event rates in the lepton paths are still dominated by ”fake” leptons
coming predominantly from QCD processes. To reduce the contribution of QCD multijets and
Z/W+jet(s) at a level comparable or below the contribution of the three main backgrounds,
t¯t, Zb¯b and ZZ(∗), a set of pre-selection cuts is applied. Events are kept only if at least one
combination of two matching pairs is found with an invariant mass greater than 100GeV/c2. To
further suppress the Z+jet(s) contamination, a loose track-based isolation is applied. The major
reducible backgrounds then remaining are Z+jet(s), t¯t → W+bW−¯b and Zb¯b with fake leptons
from jets or semi-leptonic decays of bottom mesons. The final selection incorporates tighter
lepton isolation complemented by three-dimensional impact parameter measurements. In order

Page 3

to best preserve the signal detection efficiency while acting on low p?
suppress the Zb¯b background, the isolation criteria for the leptons from the pair of lowest m?+?−
is made p?
preserve the simplicity of an mHindependent selection for the initial luminosity: it is required
a reconstructed ”Z” with 50 < mZ < 100GeV/c2and a ”Z*” with 20 < mZ < 100GeV/c2.
Contrary to the WW(∗)channel, the ZZ(∗)decay mode allows in presence of signal for a narrow
peak in the four lepton invariant mass distribution as shown in Fig. 1 (right). The Z+jet(s)
and t¯t backgrounds are completely eliminated. The Zb¯b background is considerably reduced and
only survives towards low masses.
Tlepton candidates to
Tdependent. The m2?masses observables are finally exploited with very loose cuts to
Neural Network Output
-1-0.50 0.5
events / bin
1
10
2
10
3
10
4
10
CMS Preliminary
=170 GeV
H
Signal, m
W+jets tW
di-boson
t tDrell-Yan

]]
22
[GeV/c [GeV/c
µµ
2e22e2
mm
00 50 50 100100150 150 200200250 250300 300350 350 400400 450 450500 500
]
2
/dm [fb/10 GeV/c
!
d
!
d
00
0.5 0.5
11
1.5 1.5
22
2.5 2.5
33
3.5 3.5
]
2
/dm [fb/10 GeV/c
bbb Zb Zb Zb
ZZ* ZZ*ZZ*
H130H130 H130
H150 H150H150
H200 H200 H200
H250 H250H250
µ
2e2
"
ZZ*
"
H
After Selection
Figure 1. left: NN outputs for signal (blue squares) and background (red circles) in the WW(∗)
channel for mH= 170 GeV; right: four lepton invariant mass after baseline selection in the
ZZ(∗)→ 2e2µ channel. Results are for an integrated luminosity of 1 fb−1.
5. Results
The same likelihood ratio technique as used at LEP and the Tevatron [6] is used here to evaluate
the significance of the experiment to the presence of a Higgs boson signal.
significance of an event excess under the assumption of the presence of a Higgs boson is shown
in Fig. 2 (left) for the WW(∗)channel and for the NN analysis. Systematic uncertainties for
the signal and background are taken into account. A SM Higgs boson can be found at 5σ in
this channel around mH= 160 GeV. In the case of the ZZ(∗)channel, the uncertainty on the
observations is dominated by statistical fluctuations for the considered integrated luminosity. It
is found unlikely that an integrated luminosity of 1 fb−1will yield an observation of a mass peak
with a significance well above 2σ. In absence of a significant deviation from the SM expectations,
an upper limit on the cross-section for the production of a SM-like Higgs boson can be derived.
The results are presented in Fig. 2 (right) for various mHhypothesis and expressed in terms of
the ratio of excluded over Standard Model cross sections R95%C.L.= σ95%C.L./σSM. Systematic
errors on the signal and background are taken into account, with an assumption of their 100%
correlation. One can see that there is a fair chance of excluding the SM-like Higgs at 95%C.L.
for the mass range 185−250GeV/c2. From the event counts for signal and background, together
with associated systematic errors, it is possible to combine results of expected sensitivity in all di-
boson decay channels. The combination of these channels is done using two different approaches,
a bayesian calculation and a method based on confidence levels, with different assumptions on
the correlation between systematic errors. As a result, combining the di-boson decay modes in
the leptonic final states, a SM-like Higgs boson can be excluded at the LHC with 1 fb−1for
all masses above ∼ 140GeV/c2. Finally, a preliminary evaluation of the combined sensitivity in
The expected

Page 4

2
Higgs mass, GeV/c
120 140160 180200
−1
Significance with 1fb
0
1
2
3
4
5
6
7
8
Multivariate Analysis
CMS Preliminary
Figure 2. left: expected significance as obtained from the NN analysis for the WW(∗)channel;
right: expected excluded cross-sections for a SM-like Higgs boson normalized to the SM Higgs
boson cross sections. The green (yellow) bands shows the 68% (95%) coverage range. Results
are for an integrated luminosity of 1 fb−1.
the case of a 10 TeV initial energy of the LHC has been performed. The dominant effect is the
change in cross-sections due to partonic luminosities. Using a simple rescaling of the signal and
background yields leads to a loss of about a factor 1.5 in sensitivity going from 14 TeV to 10
TeV. Detailed analyses at 10 TeV involving a complete reevaluation of systematics are ongoing.
6. Conclusions
The CMS experiment is actively preparing for the Higgs search with complete analyses based
on detailed Monte carlo simulations and full event reconstruction. The di-boson decay modes
with their leptonic final states represent the main discovery channels for integrated luminosities
of around 1 fb−1. A SM Higgs can be found at 5σ in the WW(∗)channel around mH= 160 GeV
for an integrated luminosity of 1 fb−1. Due to its lower σ×BR, the ZZ(∗)channel doesn’t allow
for a discovery with such integrated luminosities and 95% confidence level exclusions limits can
be set in absence of deviation from SM expectation. Combining both channels, it is found that
with an integrated luminosity of 1 fb−1a SM-like Higgs boson can be excluded at the LHC for
all masses above ∼140 GeV/c2. In the case of a 10TeV initial energy of the LHC machine, the
sensitivity would decrease by a factor of about 1.5.
[1] R. Barate et al. [LEP Working Group for Higgs boson searches], Phys. Lett. B 565 (2003) 61 [arXiv:hep-
ex/0306033].
[2] G. Hesketh [CDF and D0 Collaborations], Nucl. Phys. Proc. Suppl. 177-178, 219 (2008); The TEVNPH
Working Group [CDF and D0 Collaborations], Fermilab-Pub-08-270-E, e-print arXiv:0808.0534v1 (August
2008) 14pp.
[3] TheTEVNPH WorkingGroup [CDFand D0
arXiv:0903.4001v1 (March 2009) 25pp.
[4] S. Leone [CDF and D0 Collaboration], arXiv:0710.4983 [hep-ex].
[5] CMS Collaboration, “The CMS experiment at the CERN LHC”, Journal of Instrumentation 3 S08004, (August
2008) 361 pp.
[6] A.L. Read, in CERN Report 2000-005 p. 81 (2000).
Collaborations], Fermilab-Pub-09-060-E,e-print