Doctoral Thesis by PHD - Javier Santaolalla

Centro de Investigaciones Energéticas, Medioambientales y
Tecnológicas
Departamento de Fı́sica Atómica, Molecular y Nuclear

Facultad de Ciencias Fı́sicas
Universidad Complutense de Madrid
Measurement of Electroweak processes

in muon
√ decay channels, in pp collisions
at s = 7 TeV, in the CMS experiment
at LHC
Author:
Javier Santaolalla Camino
Supervisors:
Dr. Juan Alcaraz Maestre
Dra. Begoña De la Cruz Martı́nez
Dra. Isabel Josa Mutuberrı́a
April 11, 2012

A mi querida familia,
especialmente Darı́o y Aurora
ya , miembros, socios
y simpatizantes
Contents
1 Introduction 1
2 The Large Hadron Collider and the CMS experiment 5

2.1 LHC physics program . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 The LHC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 CMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3 Muon detection and reconstruction in CMS 29

3.1 Importance of muons in particle physics analysis . . . . . . . . . . . . . 29
3.2 CMS muon system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2.1 Cathode Strip Chambers. . . . . . . . . . . . . . . . . . . . . . . 33
3.2.2 Resistive Plate Chambers. . . . . . . . . . . . . . . . . . . . . . . 34
3.2.3 Drift Tube Chambers. . . . . . . . . . . . . . . . . . . . . . . . . 34
3.3 Determination of the drift velocity . . . . . . . . . . . . . . . . . . . . . 40
3.3.1 Calibration of the drift tube . . . . . . . . . . . . . . . . . . . . . 41
3.3.2 Effect of the muon incidence angle in the drift velocity computation 44
3.3.3 Effect of the signal propagation delay along the DT wire . . . . . 46
3.3.4 Effect of the magnetic field in a φ-Superlayer . . . . . . . . . . . 48
3.3.4.1 Effect of the magnetic field in the drift velocity compu-
tation for all CMS wheels . . . . . . . . . . . . . . . . . 49
3.3.5 Effect of the magnetic field on the drift velocity computation
within a chamber . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.4 Determination of DT reconstruction efficiency. . . . . . . . . . . . . . . 54
3.4.1 Data and sample selection . . . . . . . . . . . . . . . . . . . . . . 54
iii
CONTENTS
3.4.2 Muon trajectory and crossing point determination in the test

chamber . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.4.2.1 Extrapolation for the 0T runs (linear method) . . . . . 57
3.4.2.2 Extrapolation for the 3.8T runs (STA based method) . 60
3.4.3 Quality criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.4.4 Inefficiencies due to geometrical effects . . . . . . . . . . . . . . . 67
3.4.4.1 Results for top sectors . . . . . . . . . . . . . . . . . . . 73
3.4.5 Results for bottom sectors . . . . . . . . . . . . . . . . . . . . . . 73
3.5 Validity tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4 Muon momentum measurement in CMS 87

4.1 Muon momentum measurement . . . . . . . . . . . . . . . . . . . . . . . 88
4.2 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.3 Validation of the method with MC samples . . . . . . . . . . . . . . . . 95
4.4 Results on MC using fake data sets . . . . . . . . . . . . . . . . . . . . . 97
4.4.1 Constant shift in 1/pT . . . . . . . . . . . . . . . . . . . . . . . . 98
4.4.2 Intrinsic resolution-like term . . . . . . . . . . . . . . . . . . . . 103
4.5 Results on collision data . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
4.5.1 Misalignment scenarios on MC samples . . . . . . . . . . . . . . 109
4.5.2 Scale factor and resolution parametrization in η and φ . . . . . . 113
4.5.2.1 Scale factor dependence on φ and η . . . . . . . . . . . 113
4.5.2.2 Resolution dependence in η . . . . . . . . . . . . . . . . 116
4.5.3 Scale factor and resolution results after the parametrization . . . 117
4.5.4 Comparison with MuScleFit method . . . . . . . . . . . . . . . . 121
4.6 Theoretical uncertainties and biases in the determination of muon reso-
lution parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
4.6.1 Parton density function (PDF) uncertainties . . . . . . . . . . . 124
4.6.2 Initial State Radiation (ISR) uncertainties and higher order QCD
effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
4.6.3 Weak and QED interference effects . . . . . . . . . . . . . . . . . 125
4.6.4 Final State Radiation (FSR) uncertainties . . . . . . . . . . . . . 126
4.6.5 Total systematics . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
iv
CONTENTS
5 Measurement of the inclusive W boson production cross section in

the muon channel 129
5.1 W and Z boson production in CMS . . . . . . . . . . . . . . . . . . . . . 134
5.2 W and Z boson decays . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
5.3 W signature in the muon channel . . . . . . . . . . . . . . . . . . . . . . 136
5.4 Inclusive W boson production cross section . . . . . . . . . . . . . . . . 140
5.4.1 Data and Monte Carlo samples . . . . . . . . . . . . . . . . . . . 142
5.4.2 Muon reconstruction and identification . . . . . . . . . . . . . . . 143
5.4.3 W → µν selection . . . . . . . . . . . . . . . . . . . . . . . . . . 144
5.4.4 Acceptance and efficiencies computation . . . . . . . . . . . . . . 150
5.4.5 W signal extraction . . . . . . . . . . . . . . . . . . . . . . . . . 152
5.4.6 Cross section measurement . . . . . . . . . . . . . . . . . . . . . 157
5.5 Systematical uncertainties . . . . . . . . . . . . . . . . . . . . . . . . . . 160
5.6 Kinematic distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
5.7 Discussion of the results . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
6 Measurement of associated production of W bosons with charmed

jets 173
6.1 Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
6.2 Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
6.2.1 Definition of W+c, W+b and W+ligh quark components at the
generator level . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
6.2.2 W+charm extraction . . . . . . . . . . . . . . . . . . . . . . . . . 186
6.3 Measurement of the σ(W+ + charm)/σ(W− + charm) ratio . . . . . . . 189
6.4 Measurement of the σ(W + charm)/σ(W + jets) ratio . . . . . . . . . . 194
6.5 Systematic uncertainties . . . . . . . . . . . . . . . . . . . . . . . . . . 194
6.6 Additional tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
6.6.1 Measurements with an alternative lifetime tagging variable . . . 200
6.6.2 Measurements using a MadGraph W+jets Monte Carlo as reference202
6.7 Final results and comparisons with theory . . . . . . . . . . . . . . . . . 203
6.8 Prospects for future improvements . . . . . . . . . . . . . . . . . . . . . 205
7 Conclusions 213
v
CONTENTS
References 217
vi
Agradecimientos
Los cuatro años de tesis dan para una larga lista de agradecimientos. Sin
embargo, los que me conocen bien saben que, entre lo despistado que soy y
lo poco que me gusta dilatarme, muchos agradecimientos no quedarán aqui
reflejados, pero también han sido importantes y no quedan en el olvido.
En primer lugar, agradezco enormemente el buen espı́ritu de trabajo y com-

panẽrismo que ha reinado en el grupo del CIEMAT durante los 4 años.
Agradecimiento especial a mis tres directores Isabel, Begoña y Juan por
lo mucho que he aprendido de ellos y la enorme paciencia que han tenido.
Pero también a Marcos, un gran jefe, Jesús, Mara, Juan Pablo, Ignacio... Y
por supuesto al resto de estudiantes: Aurelio, Nacho, Marı́a, etc. También
a mis compañeros de despacho en el CERN (Silvia, Carlo, Óscar, otra vez
Juan y Daniel) con los que ha sido divertido trabajar. Sı́, divertido. Todos
han hecho de los despacho del CIEMAT y del CERN mi segunda, o más
bien cuarta o quinta, casa.
Cientı́ficamente hablando también debo agradecer mucho a Mary Cruz por

la primera parte de la tesis, a Marı́a por la paciencia e infinito legado de
conocimiento en los W y a Mara por la maravillosa idea de estudiar el W+c.
Muy buena idea. Además de todos los coordinadores de los diferentes grupos
y todos los que han hecho posible que se generen suficientes W para que
yo tenga una tesis. También a la Fundación Mapfre Guanarteme, por ese
impulso en el momento más crı́tico. Mil gracias Julio.
Los que han hecho mis estancias en Madrid y Ginebra más llevaderas, mil
gracias. En Madrid, a todo el equipo de fútbol de fı́sicas y amigos de la
facultad (Jesús, David, etc.) y a Fete. En Ginebra a los copinos, personal
del CERN (Marı́a, Natacha, Marc, Nadege and co...), los Zapata invictos
hasta hoy, Jojo y el equipo de Challex, Raf, Cari y la salsa y los internos
de UN. Y de manera muy especial a la mejor persona del mundo, Susana.
Por supuesto todo no es trabajo, también al sol de Las Palmas, la Playa de
las Canteras y todos los que allı́ me esperaban para desconectar: Néstor y
demás noenfermos, Cansino y claretianos y en general a todos los canarios.
También a Pedro Hernández y el piano, salvadores más de una vez.
Un agradecimiento muy especial a los que han compartido casa conmigo,

pasando momentos inolvidables. Desde los orı́genes de LeClub con Albert,
y posteriormente los boludos Xabi, Martı́n, Gastón, Monti, Fran y el profe,
y finalmente con Daniel, Garo, Jose y mi primer amigo en el CERN, Tito.
A todos los socios, en especial a Carlos, Laurinha, Jabu, Bruno, Byron,
Josu y todos los simpatizantes que han hecho grande LeClub.
Burgos, Canarias, Madrid y Ginebra, a todos. Y por supuesto a mi familia,

la mejor del mundo. Y en especial gracias a la pequeña Aurora, que no sabe
lo que es una tesis pero que tan importante ha sido.
Finalmente, un párrafo especial para una mujer demasiado especial e im-

portante como para compartir párrafo con nadie. Y a toda su familia.
En fin, a todos. A Griffin, Perdue, J.L. Guerra, Cherito, Dieguito, todos

los MOWs, Justine, Marylin, Jimena, Dan, Blanka, Ken, Pepe y el Real
Madrid, Yayo, Gunther, TitoKing, a todos los cascos azules, Rosana,...
GRACIAS.
Viva LeClub.
1
Introduction
November 2009 marked the beginning of the operation era in the Large Hadron Collider
(LHC). More than 20 years of conception, design, prototyping, production, test, valida-
tion and installation were left behind, and numerous goals successfully accomplished.
Four experiments aiming for a wide physical coverage were installed at the LHC: AL-
ICE, ATLAS, CMS and LHCb. The Compact Muon Solenoid (CMS) collaboration
chose for the experiment a compact design: a 21 m long, 15 m high cylindrical detec-
tor, 12500 tons weight. The design and construction of the different CMS subdetectors
were carried out in the approximately 150 institutes of the CMS collaboration.
Proton-proton collisions at the LHC increased steadily in energy and intensity soon
after the beginning of the operation. In March 2010 collisions at 3.5 TeV per beam
were first ever produced. Operation of the machine during 2010 and 2011 runs has been
successful. A dataset with an integrated luminosity of 43 pb−1 of data was recorded in
the 2010 run, and almost 5 fb−1 in the 2011 run, exceeding all the expectations about
the LHC operation.
Many physical analyses have been carried out with data from the first two years of
operation. Among them, the electroweak measurements stand out. Electroweak bosons
are copiously produced in the collisions and their decay in leptons produce a very clear
signal for the analysis. Electroweak precision measurements are one of the first to be
accomplished.
The importance of the muon particle in this experiment is patent in its own name
(“M” of CMS standing for muons). Many of the physical goals of the CMS program are
based in searches with muons in the final state. Muon chambers are, indeed, of capital
1
1. INTRODUCTION
importance in the CMS design. A fourth of the chambers of one of the technologies
adopted as muon detectors in CMS were not only designed, constructed and built by
the Spanish CIEMAT institute, but also installed, calibrated and maintained by their
technicians and physicists. In November 2009, the performance of the barrel muon
chambers was already characterized, being ready for the first LHC collisions and the
analysis of thousand of million muons that would arise from the high energy collisions.
This thesis covers different aspects of the CMS operation and data analysis, always
with muons as the common factor. The calibration and characterization of the drift
tube chambers (DTs), one of the CMS muon detectors, carried out in the pre-collision
era, is presented in this thesis. To complete the calibration of the muon detection and
measurement, the calibration of muon momentum using electroweak data is shown. In
particular, the clean sample that Z bosons coming from collisions represents is used
to study the momentum scale and resolution of the CMS muon spectrometer. Finally,
electroweak precision measurements are shown. The inclusive electroweak boson pro-
duction cross section is the first electroweak measurement published by CMS. In this
thesis, the inclusive W boson production cross section in the muonic channel with the
whole 2010 dataset is measured. As a further step in the study of the W production,
this time accompanied by jets, the analysis of the exclusive final state W+charm jets
is both challenging and interesting from the physical point of view. The separation of
W+charm from other W+quark jets is successfully achieved using c-tagging techniques
that are new in CMS and useful for any other c jet studies. This analysis can constrain
the quark strange content in the proton. Prospects to improve the measurement with
2011 data, as well as important distributions with already 2 fb−1 are also reviewed in
the last part of the thesis.
The description of the LHC and its physical goals are shown in Chapter 2. The
CMS detector is also presented in detail in this Chapter.
Chapter 3 is devoted to muon chambers. The importance of muons in the CMS
physics program, the design of CMS DT chambers and the characterization and cali-
bration of this subdetector with cosmic muons in the pre-collision era are shown. The
fundaments of the detection using drift tube chambers are discussed. The calibration
procedure is revised and completed with an analysis based on the drift velocity mea-
surement and the factors that can modify the result of this measurement. The DT
chambers are characterized in terms of reconstruction efficiency in the same Chapter.
2
The proposed method is applied to scenarios with the solenoidal magnetic field on and
off to compare the results.
The muon transverse momentum calibration is the topic of Chapter 4. The method
presented aims to compute the extra momentum scale and resolution terms to complete
the already well performing simulation of physical processes through CMS and adequate
it to the experimental data. Many analysis can benefit from this method, since it
directly provides the distortion factors that can be used to make simulated samples
resemble data. The application of this method to evaluate systematic uncertainties
based on muon momentum scale and resolution is straight forward. This “bridge”
Chapter uses elements of the first part of the thesis (muon detection) and introduces
elements of the second part of the thesis (electroweak boson selection with collision
data).
Chapter 5 aims to underline the importance of electroweak studies in any collision
experiment, to set the basis for W boson selection and to compute the W boson produc-
tion inclusive cross section. This analysis benefits from high statistics recorded in the
2010 run (already dominated by the systematic uncertainties). This allows to present
a precision measurement already at the 2% level that is shown to be in agreement with
the Standard Model. The use of templates based on experimental data to model shapes
of distributions for signal and background processes, the efficiency computations and
other techniques that exploit the huge amount of electroweak boson recorded are a
key part to reduce these measurement uncertainties to the minimum achieved so far in
CMS.
Chapter 6 is devoted to the associated charm production in W boson final states. Its
application in terms of parton densities for the strange quark and the implication that
an improvement on this side could have in other precision measurements are presented
to motivate this study. A detailed selection and extraction of the W+c signal is dis-
cussed in this Chapter, leading to a clear signal signature by using c-tagging techniques.
The ratios, both of positive over negative W+charm production and the W+charm over
the W+jets, are presented and the uncertainties in this measurement shown in detail.
Prospects to exploit the possibilities to reduce the systematic uncertainties with larger
data samples and the sensitivity of this study to fulfill PDF groups requirements on
the strange quark composition of the proton are also discussed.
3
1. INTRODUCTION
4
2
The Large Hadron Collider and

the CMS experiment
The LHC (Large Hadron Collider) is a 27 km synchrotron that produces proton-proton

collisions at a new energy scale (1, 2, 3). It is the largest and most powerful particle
accelerator at present. The tunnel that hosts the accelerator is located near Geneva
(Switzerland), between 50 and 125 meters under the surface. During the first two years
of operation, the LHC has produced collisions at an energy of 7 TeV (center-of-mass
system). After these two years of operation, the amount of data recorded is enough to
perform high precision physical measurements.
2.1 LHC physics program
The current model for particle physics is called Standard Model. It is accurate and
almost complete (4). The SM embodies our current understanding of particles and
forces. It was developed in the middle 20th century. This theory has been tested with
an extraordinary precision in many different experiments. It has explained a great
deal of experimental results and predicted a host of varied phenomena. Today this
model has become established as a well tested physical theory and a reference model
to describe most of the phenomena.
The SM has successfully described hundreds of particles and their complex interac-
tions with few ingredients: 12 matter particles and their corresponding antiparticles (see
Table 2.1); and force carrier particles (see Table 2.2). The matter particles (fermions)
5
2. THE LARGE HADRON COLLIDER AND THE CMS EXPERIMENT
Charge First Generation Second Generation Third Generation

-1 Electron (e) Muon (µ) Tau (τ )
Leptons
0 Neutrino(νe ) Neutrino (νµ ) Neutrino (ντ )
-1/3 Down (d) Strange (s) Bottom (b)
Quarks
2/3 Up (u) Charm (c) Top (t)
Table 2.1: Matter content as described in the SM (5).
Electromagnetic Weak Strong Gravitation

Photon (γ) Weak Bosons (W + , W −, Z 0) Gluons (g) Graviton (g) ?
m=0 m ∼ 100 GeV m=0 m=0
Table 2.2: Force carrier particles as described by the SM.
are classified as quarks or leptons depending of whether or not they feel the strong
force respectively. The interactions in this model are described by the interchange
of a force carrier particle (bosons). The SM has successfully integrated three of the
forces known in nature: electromagnetic, weak and strong. Forces in this theory are
described as quantum fields in which the Lagrangian is invariant under a continuous
group of local transformations. Quantum electrodynamics, the theory describing the
electromagnetism within this model, is described using an abelian gauge theory with
the symmetry group U(1) with one gauge field, and the photon as the gauge boson. By
adding to this theory the symmetry group SU(2) the SM has been able to describe also
the weak force, reaching the unification of both forces, that are shown to be different
aspects of the same force. At the energy of the order of 100 GeV both forces appear to
be the same. The force carrier particles of this unified gauge group U(1)×SU(2), after
the symmetry breaking, are called photon, W+ , W− and Z.
The strong force is included in this model using a similar description: a quan-
tum field with a symmetry group. In this case SU(3) is the non-abelian group that
successfully describes the strong processes. This quantum theory is called quantum
chromodynamics (QCD) and is added to the SM by increasing the gauge group to form
the final U(1)×SU(2)×SU(3).
In the U(1)×SU(2) Lagrangian the force carrier particles appear to be massles.
However, experimentally the masses of the W and Z bosons result to be different from
0. The electroweak symmetry breaking mechanism was introduced in this model as a
6
solution to this problem. The Higgs mechanism, implemented by a Higgs field and its
interaction with the other fields, is the missing element within the theory in order to
produce the mass terms. The way these bosons acquired mass explained at the same
time the masses of all particles within the SM. It is the simplest mechanism capable of
giving mass to the gauge bosons while remaining compatible with gauge theories. The
Higgs field would consist of 4 component fields, two charged and two neutral. Three of
them are used to give mass to the electroweak bosons (W+ , W− and Z). Both charged
scalar and one neutral components are goldstone bosons which act as the longitudinal
third polarization components of the massive W+ , W− and Z bosons. The remaining
degree of freedom would become detectable through a boson, the Higgs boson: a 0-spin
massive particle. All the properties of this particle, except the mass, are predicted by
the theory.
Despite the accuracy of the SM describing many of the physical phenomena, our
understanding of the Universe is still incomplete. In order for the SM to be a complete
theory, the Higgs boson has to be discovered. Direct searches in the former CERN
main accelerator (LEP) excluded the existence of such a particle if its mass is smaller
than 114 GeV (6, 7). Measurements carried out in the proton-antiproton collider at
Fermilab (Tevatron) excluded the existence of this boson in the mass window [158 -
176] GeV (8). More recently, in the LHC, the ATLAS and CMS experiments combined
exclusion range for 2.3 f b−1 of proton-proton collision analyzed spans from 141 to 476
GeV (9). In Fig. 2.1 the exclusion region for CMS and ATLAS (combined), LEP and
Tevatron is shown. In Fig. 2.2 the preferred Higgs boson mass, following a fit to the
measured parameters of the SM is shown. From both figures we can conclude that
even if the Higgs boson is excluded for most of the window mass where it could exist,
the preferred region needs still some time to be analyzed. In the short future, due to
the high luminosity-high energy collisions at LHC, the experiments will cover the full
theoretically permitted mass spectrum for the Higgs boson. At LHC both ATLAS and
CMS experiments were specially designed to optimally search for this missing particle.
If the Higgs boson is proved not to exist, many of the different theoretical proposals
based on a higgless scenario should be tested experimentally. LHC may shed light
over all these possibilities. The SM, despite being a successful theory to describe many
physical phenomena, is today considered a phenomenological theory part of a more
fundamental one yet to be described. Among SM limitations, the most important are:
7
Figure 2.1: Exclusion region (LEP, TEVATRON and the combination of AT-
LAS and CMS) - Excluded region with 2.3 f b−1 of proton-proton collision available.
Exclusion bands coming from LEP and TEVATRON are also shown.
• it does not include gravity. Gravitation is the only known force which can not be
described using SM formulation: quantum fields with gauge symmetry. Gravity
is accurately formulated using Einstein’s Relativity, which is a relativistic, non-
quantum theory. Every attempt as of today to quantize gravitation has not been
successful. The description of all physical phenomena within a unique framework
is the ultimate goal of physics, called unification theory. Such kind of theories,
globally called GUT, have been actively looked for during the last decades. The
LHC by exploring a new energy range will provide useful information in the chase
of this theory.
• it has too many parameters. The inclusion of many parameters in the theory
makes the model less predictable and elegant, without an apparent reason behind
each particular value. The existence of so many parameters in the SM make the
physicists community consider SM not to be the fundamental theory of nature.
• even if the missing piece of the SM is discovered in the LHC, there is a problem
that will remain unsolved. Unless there is a fine-tuning cancellation in loop
diagrams, the large quantum contributions to the square of the Higgs boson mass
would make this to be bigger than its expected value. This is the so called
8
Figure 2.2: Exclusion region - The χ2 for a global fit to electroweak data is shown as a
function of the Higgs mass. The solid line corresponds to the result of this fit with the blue
band indicating the impact of theoretical uncertainty. The yellow-shaded area indicates
the region of Higgs masses excluded by experiment.
9
hierarchy problem. Cancellations provided by a new group of particles is the

solution proposed by Supersymmetry to solve this problem.
• there is no energy scale at which electromagnetic, weak and strong coupling con-
stants converge in a unique value. Again, Supersymmetry is a proposed theory
which would solve this problem.
Many different theories were proposed during the last half century to substitute
the SM. Among them supersymmetry is the most popular (10). Sypersymmetry is
an extension of the SM. By doubling the number of particles it is able to solve the
hierarchy problem, and, at the same time, unify the three described forces. It adds a
new symmetry to those of SM: fermion-boson symmetry. In this theory, the number
of bosonic number of degrees of freedom equals the fermionic ones. As a consequence,
integer spin particles with the same properties as leptons and quarks should appear:
the so called sparticles. If this theory is correct, the sparticles may appear in the
proton-proton collisions at the LHC. Other theories such as string theory (11), extra
dimensions or technicolor (12) are also expecting new LHC data to be confirmed or
rejected.
Following cosmological observations, the amount of matter in the universe described
by SM is only 4%. The other 96% remains unknown, and it is shared between two
different kinds of energy/matter whose origin is unknown, called dark matter (23%) (13)
and dark energy (73%). The existence of the dark matter was proposed after several
indirect cosmological measurements. So far, no hints of any dark matter candidate
have appeared in any of the different experiments carried out for such a purpose. At
LHC these particles may be created and studied in the two major experiments (CMS
and ATLAS).
Another open question in particle physics is the matter-antimatter asymmetry. If
matter interacts exactly in the same way antimatter does, only radiation would have
remained after matter annihilation at the Big Bang. Due to the fact that there is
matter in the Universe and matter and antimatter are not coexisting, there should be a
difference in their interaction, namely, that there exists a violation of the CP (charge-
parity operator) conservation (14). The solution to this problem is also one of LHC
goals, and a dedicated experiment, LHCb, has been setup for it.
10
2.2 The LHC
Besides new discoveries of particles in the collisions, the physics program at the
LHC is wider, including studies in many fields of particle physics. Electroweak studies
are important for several reasons. First of all, they provide a test to the validity of the
SM at a new energy regime. The improvement on the knowledge of some Electroweak
boson properties, e.g. W boson mass, is useful to better understand and constrain the
SM (15). Several Electroweak measurements and production channels can improve the
knowledge on the proton parton composition (PDFs). Electroweak processes are the
main background of many discovery channels, and their understanding directly affects
these searches. They also provide high mass resonances whose decay products are used
to calibrate the detector response.
Due to the available energy in the collisions and the dominant gluon-gluon fusion
production process, the LHC provides a huge amount of top-antitop events, making
LHC a top quark factory. Top quark was the last particle of the SM to be discov-
ered (16). By measuring its properties, we are sensitive to new physic phenomena. Top
quark production is also a main background for many interesting processes and a tool
for calibration and physics commissioning purposes.
The study of medium-mass-quark production in high-energy hadronic interactions
plays a critical role in testing next-to-leading order (NLO) Quantum Chromodynamics
(QCD) calculations. Additionally, the hadroproduction of quarkonia is not fully under-
stood. For example, none of the existing theoretical models satisfactorily describes the
prompt J/ψ differential cross section. LHC results are improving our understanding
of the quarkonium productions mechanisms. Low mass resonances like the J/ψ also
contribute to the calibration of the detector.
2.2 The LHC
The LHC is the most energetic particle accelerator built up to now (1, 2, 3). It collides
bunches of protons at almost the speed of light. Before acceleration in LHC tunnel,
from 450 GeV till nominal energy (3.5 TeV), protons are pre-accelerated using differ-
ent CERN machines (see Fig. 2.3). Protons are extracted from hydrogen atoms. The
first accelerator is linear, Linac2, in which protons reach an energy of 50 MeV before
being injected in the following step: the PS Booster. The PS Booster is a synchrotron
11
accelerator of 25m radius which accelerates protons to feed the next step in the accel-
erator sequence: the Proton Synchrotron (PS). Protons reach PS with an energy of 1.4
GeV. PS is a 628 m circumference ring which accelerate protons that are injected at
an energy of 25 GeV into the super Proton Synchrotron (SPS). The SPS accelerates
them to reach an energy of 450 GeV per bunch and transfers them to the LHC.
Figure 2.3: CERN accelerator complex - LHC and all the accelerators involved in
the experimental complex
The LHC is hosted in the LEP tunnel, a 27 km long circumference, with 25% of
its perimeter in Switzerland, the rest under French surface. The tunnel was built 100
m underground in average. At the same time, Earth provides a good shielding for
radiation. It was built with a 1.4% tilt making its depth vary between 175m and 50m.
LHC accelerator is made of arcs and insertions. The arcs contain each 154 bending
superconducting magnets. Insertions are straight sections plus two transition regions
placed in each end. Depending on the use of the insertion, its layout could vary. There
are insertions for physics, injection, beam dumping and beam cleaning. A sector is a
segment of LHC with an arc and an insertion. LHC consist of 8 sectors, which are
defined as the accelerator segment between two insertion points.
The high mass of the particles of interest makes necessary collisions with particles
at high energy. The small probability for these particles to be created with respect
12
2.2 The LHC
to other non-physically-interesting processes makes necessary a high collision rate and

a small interaction area. In other words we need collisions at high energy and high
luminosity. These are the two most important parameters for an accelerator.
To reach this high energy, it was necessary to build a long accelerator ring (27 km)
with high magnetic field to bend the proton beam (8.3 T). To accelerate the particles,
superconducting radiofrequency cavities are used. Currently the protons are accelerated
to an energy of 3.5 TeV. In the near future (2014 or 2015) it is foreseen to reach 7 TeV,
at full magnets power.
Luminosity depends on beam parameters (2). For a gaussian beam profile distribu-
tion:
Nb2 nb frev γr
L= F (2.1)
4πǫn β ∗
where Nb is the number of particles per bunch (1011 protons at LHC), nb the number
of bunches per beam, frev the revolution frequency (approx. they turn 11 thousand
times per second), γr the relativistic gamma factor, ǫn the normalized transverse beam
emittances, β ∗ the beta function at the collision point and F the geometric luminosity
reduction factor due to the crossing angle at the interaction point. Given the number
of particles per bunch and number of bunches in the beam, the luminosity can be
increased by reducing the emittances and the β factor. LHC uses corrector magnets
to avoid the dispersion of the beam and thus increase the luminosity. By the end
of September 2011, after two years of proton-proton collisions, LHC reached a peak
luminosity of 3 · 1033 cm−2 s−1 and it has kept working at this value until the end of
2011 proton-proton data taking, in November 2011.
These high values made the design of the LHC a technological challenge, implying
development of new technologies in different fields:
• Vacuum. Protons in the accelerator must travel in the beam pipe in the absence
of matter to avoid scattering or energy loses. LHC has three vacuum systems:
for cryomagnets, to avoid heat load by gas conduction, it requires a pressure of
only about 10−6 mbar; for the helium distribution line and for the beam vacuum
(reaching 10−9 mbar).
• Magnets. There are about 9600 magnets in the LHC, 1232 of them are 15 m long,
35 tons dipoles. Dipoles use NbTi cables working at 1.9 K. The magnet coils
13
for the LHC are wound from ’Rutherford’ cable. This cable consists of up to 36
twisted 15-mm strands, each strand being made up in turn of up to 8800 individual
filaments, each filament having a diameter as small as 7 micrometres. At this
temperature these cables conduct electricity without resistance (superconductor).
The current, 11850 A, is enough to create the 8.3 T magnetic field required
by design. The other magnets (quadrupoles, octupoles, decapoles,...) are used
to focus and squeeze the beam and optimize its trajectory. Most of them are
embedded in the main dipoles or quadrupoles.
• Cryogenics. The LHC is the biggest cryogenic system in the world. To maintain
its 27 kilometre ring at superfluid helium temperatures, the LHC’s cryogenic
system have to supply an unprecedented total refrigeration capacity of some 150
kW at 4.5 K and 20 kW at 1.9 K distributed around the ring.
• Cavities. They are aimed to deliver radiofrequency power to the beam during
energy ramping. They also keep the proton bunches in LHC as tight as possible
in order to reduce the bunch spread, increasing as a consequence the instant
luminosity. LHC uses superconducting cavities due to their small losses and large
stored energy. In total, 8 cavities per beam are used, placed in a long straight
section.
Collisions are designed to take place at LHC in the 4 interaction points, where the
tunnel turn into a cavern to albergate the detectors. The 4 main LHC experiments are
placed in these caverns: ALICE (A Large Ion Collider Experiment) (17), ATLAS (A
Toroidal LHC AparatuS) (18), CMS (Compact Muon Solenoid) (19) and LHCb (LHC
experiment for b-physics) (20). TOTEM and LHCf, two forward smaller experiments,
are placed nearby the interaction point of the two big experiments at LHC (CMS and
ATLAS) and aimed to study very forward physics processes.
ATLAS and CMS are general purpose detectors, they aim to look for any hint of
new physics in the new available energy spectrum. LHCb will study the physics of b
quarks (trying to solve the matter-antimatter mystery). ALICE will study heavy ion
collisions, trying to observe a new state of matter called quark-gluon plasma.
LHC operation started the 23th November 2009 with the first proton-proton colli-
sions. In these collisions, the proton beam consisted in only one bunch of protons and
14
2.3 CMS
collided at the injection energy (450 GeV per proton). Shortly after, the 30th March
2010, the beam energy was increased to an energy of 3.5 GeV per proton, starting the
so called 2010 run. It finished the 8th November 2010 with a peak luminosity of 2 · 1032
cm−2 s−1 . During this period, an integrated luminosity of 36 pb−1 of proton-proton
collisions at 3.5 GeV per proton were collected and certified as good data for analysis
in CMS and ATLAS.
The 2011 data taking period started on the 13th of March of 2011 with collisions
at 3.5 GeV per proton beam at a peak luminosity of 1.2 · 1030 cm−2 s−1 . The last run
considered in this thesis involved collisions at 3.5 GeV of beam energy and was taken
the 7th of September 2011 at an instantaneous luminosity of 3 · 1033 cm−2 s−1 . This
period (2011 runA) was followed by a technical stop aimed for improvement of the
accelerator. The total luminosity collected and certified by CMS during this period is
2.3 fb−1 .
2.3 CMS
CMS is a cylindrical detector which is placed in one of the LHC interaction points. It
is a 21.6 m length, 14 m diameter cylinder of 12500 tons, with a solenoidal field of 3.8
T (21) (see Fig. 2.4). This design guarantees a full coverage around the interaction
point and a good transverse momentum resolution, for an accurate measurement of the
physical processes generated at the collisions.
Being a general purpose experiment, it explores any hint of new physics at the LHC,
besides performing precision measurements on existing particles. The goals pursued by
this experiment determine its main characteristics (21, 22):
• Muons. Redundant and accurate muon system, with excellent muon identifica-
tion, over a wide range of momenta in the pseudorapidity region below 2.5. Good
dimuon mass resolution and negligible charge miss-identification for muons with
momentum lower than 1 TeV/c. Excellent trigger efficiency. The importance of
muons in CMS will be seen in Chapter 3.
• Tracker (pixel and silicon strips). Good performance in the tracker: excellent
momentum resolution and reconstruction efficiency. High trigger efficiency and
b-jet tagging efficiency. CMS tracker is the biggest silicon detector ever built.
15
Figure 2.4: CMS detector - Representation of CMS detector
• Electromagnetic calorimeter. Exceptional energy resolution, diphoton and dielec-

tron mass resolution. Wide geometric coverage, good direction measurement, π 0
rejection and efficient photon and lepton isolation. Its design was driven by the
needs to cover the low mass (≈ 100 GeV) region for the Higgs boson (H → γγ
channel).
• Hadronic calorimeter. Reasonable dijet mass resolution and missing transverse

energy resolution. Large coverage and fine lateral segmentation.
In CMS different detector technologies are applied in order to optimize particle

identification. These detectors are placed concentrically, and in two regions: barrel
(central part of CMS covering till 1.1 in pseudorapidity) and endcaps. In the central
region (the barrel, see Fig. 2.5), going from the innermost part to the outermost one
we find: a tracker detector, built with silicon strips and pixels, the electromagnetic
calorimeter (ECAL), the hadronic calorimeter (HCAL) and the muon chambers. In
the barrel region drift tube chambers and resistive plate chambers are used as muon
detectors, and are located in 5 wheels. In the 2 endcaps, cathode strip chambers are
16
2.3 CMS
installed instead due to occupancy and magnetic field conditions in this part of CMS.
A brief description of each of the subdetectors follow.
Figure 2.5: CMS detector transversal slice - Slice of CMS detector showing the
different technologies used in CMS barrel.
• Magnet. The aim of CMS magnet is bending the charged particles in the trans-
verse plane to compute, by means of the radius of curvature, the momentum of
the particle in this plane. CMS magnet is a 13 m long, 5.9 m diameter supercon-
ducting solenoid producing a 3.8 T magnetic field. The muon chambers in the
barrel, in the outer side of the solenoid, benefit from a high return field. They are
placed in the 1000 ton return yoke, in the barrel region. A 2 T residual field in
this region allows a good momentum resolution in the muon chambers. The sta-
bility of the magnetic field in this outer region is also an asset, to ease the muon
reconstruction and to allow a good momentum resolution (23). The calorimeters
and tracking detectors are placed within the solenoid. CMS magnet consist of 5
modules. Each module contains 4 layers winding of NbTi cables. With this design
(19.5 kAmperes current, 2.6 GJ stored energy) the quality factor (stored energy
over mass) is improved by a factor two with respect to the previous state-of-art
magnet designs.
• Inner Detectors. They are built to accurately measure the trajectory of the
charged particles, in a region where the magnetic field is intense. They aim
17
to compute the momentum of the particles in the transverse plane, to detect

the vertices and for triggering purposes. CMS inner detector consist of silicon
detectors (24, 25). It uses three different designs depending on the expected flux
in the three different regions (see Tab. 2.3). In the innermost part, close to the
interaction point, where the flux is very high (10 < r < 20 cm) pixel detectors
are used. Given the size of each pixel (100 × 150 µm2 ) and the luminosity in
nominal operation mode, the occupancy per channel for this subdetector is about
10−4 . In the intermediate region (20 < r < 50 cm) CMS makes use of silicon
microstrips. The expected rate of particles in this region allows the use of 10
cm × 80 µm detectors, keeping the occupancy low enough (2-3%). Larger-pitch
silicon microstrip are used in the outermost region of the inner detector, where
the flux of particles has substantially decreased (r > 55 cm). The maximum cell
size (25 cm × 180 µm) is still small enough to keep occupancy below 1%.
Detector Radius Size Occupancy

Pixels r > 10 cm 100 × 150 µm2 < 10−4
Silicon microstrip 20 < r < 50 cm 10 cm × 80 µm 2-3%
Silicon microstrip r > 55 cm < 10 cm × 80 µm 1%
Table 2.3: Description of the different technologies and sizes of the tracker detector
components in CMS.
Pixel detector. The total amount of pixels in CMS is 66 millions, covering an

area of 1m2 . They are distributed in two areas: barrel and endcap (see Fig. 2.6).
In the barrel, 3 pixel layers are placed at 4.4 cm, 7.3 cm and 10.2 cm of average
radius. 768 pixel modules, arranged into half-ladders of 4 modules each are used
in this region. The total length of the pixel detector in this part is 53 cm. In each
of the 2 endcaps CMS uses two pixel disks. The detector in this region extends
from 6 to 15 cm in radius, with 672 pixel modules, with 7 different modules in
each blade. The spatial resolution obtained is 10 µm in r-φ plane (transverse
plane), 15-20 µm in the z coordinate (longitudinal axis).
Strip detector. It consist of 9.6 million silicon strips located in 15400 modules,
covering an area of 200 m2 . Strips in CMS are placed in 2 regions in the barrel
(TIB, Tracker Inner Barrel, and TOB, Tracker Outer Barrel) and two regions in
18
2.3 CMS
Figure 2.6: Layout of the pixel detector - Layout of CMS pixel detector. It is
divided in two regions: barrel and endcap
each of the endcaps (TID, Tracker Inner Disks and TEC, Tracker End Cap) (see
Fig. 2.7 and 2.8). In the barrel region, the tracker detector of CMS consists of 10
layers: 4 layers in the TIB, and 6 in the TOB. The TIB covers up to z < 65 cm.
Strips in this region are 320 µm thick, and 80-120 µm pitch. Resolution in the
TIB is 24 µm approximately in both r-φ plane and z direction. In TOB thicker
(500 µm) and wider pitch (120-180 µm) detectors are used, since the flux in this
region is lower. Resolution in r-φ plane varies from 32-52 µm, and in z direction
52 µm. Each TEC consists of 9 disks. They are arranged in rings, having different
pitches. The thickness is 320 µm in the innermost regions and 500 µm for the
rest of the TEC. TID comprises 3 small disks. Strips in this region are 320 µm
thick. The performance of the tracker detector is shown in Fig. 2.9 and Fig. 2.10.
In Fig. 2.9 the resolution computed with first LHC data is shown. The resolution
is computed after a fit to the mass of dimuon events candidates to come from
J/ψ resonance, versus the muon (lepton) pseudorapidity. A ≈ 1% resolution is
observed in the central region whereas a ≈ 3% resolution is computed for the
endcaps. In Fig. 2.10 the primary vertex resolution is plotted versus the number
19
of tracks for different average pT and in the 3 spatial axis. When the number
of tracks used to determine the primary vertex is sufficiently high, the resolution
achieved in the transversal plane is of the order of 10 µm and ≈ 15 − 20µm in
the longitudinal one.
Figure 2.7: Layout of the tracker detector - Layout of CMS tracker detector.
• Electromagnetic Calorimeter (ECAL). It is essential in the identification of

electrons and pions and in the measurement of the energy in hadronic showers (21,
28). This detector is placed inside the solenoid, covering a pseudorapidity range
of |η| < 3.0. Lead tungstate (PbWO4 ) crystals are used for being fast (80% of the
light is emitted within 25 ns), compact, and their fine granularity and radiation
resistant, having an excellent energy resolution. However, these crystals present
the drawback of a relatively low light yield (30 γ/MeV). Photodetectors with
intrinsic gain, and that can operate in a magnetic field, are used to detect the
radiation coming from the electromagnetic cascade and its amplification. Silicon
Avalanche Photodiodes (APDs) are used in the barrel (EB), Vacuum Phototriodes
(VPTs) in the endcaps (EE). In total, 61200 lead tungstate crystals in the barrel
part and 7324 in the 2 endcaps comprise CMS ECAL. Details on the EB and EE
elements are given in Tab. 2.4.
20
2.3 CMS
Figure 2.8: Representation of CMS tracker detector - Pictorial representation

of CMS tracker detector.
Figure 2.9: Resolution on transverse momentum. - Resolution on transverse

momentum with data (blue line) compared to the MC resolution computed from MC
truth (red points) and from the fit using dimuon events from the decay of the J/ψ
resonance (26).
21
Figure 2.10: Vertex resolution - Primary vertex resolution in the X axis (top
left), Y axis (top right) and Z axis (bottom) versus the number of tracks for different
average pT (27).
22
2.3 CMS
In the barrel (EB) 22 × 22 mm2 front face cross-section crystals are used. Their
length (230 mm) corresponds to 25.8 radiation lengths. EB has an inner radius
of 129 cm and consist of 36 identical modules, each containing 1700 crystals.
It covers a pseudorapidity range |η| < 1.479. The granularity obtained is of
δη × δφ = 0.0175 × 0.0175.
The endcaps (EE) use 28.6 × 28.6 mm2 front face cross-section crystals, 220
mm length (24.7 radiation lengths). They are placed 314 cm away from the
interaction point, covering a η range of 1.479 < |η| < 3.0. Crystal are grouped in
“supercrystals”, units of 5×5 crystals. A preshower device, 2 planes of silicon strip
detectors with disks of lead absorber, is placed in front of the crystal calorimeter
to improve the position determination of the particles, and pion identification.
Cross section Length (Radiation Lenghts) Coverage (η)

EB 22 × 22 mm2 230 mm (25.8) |η| < 1.479
EE 28.6 × 28.6 mm2 220 mm (24.7) 1.479 < |η| < 3.0
Table 2.4: Description of the elements of the ECAL detector.
In Fig. 2.11, the energy resolution for electrons from simulation as a function
of the incident particle energy is shown. This curve is parametrized as a func-
tion of energy for energies below 500 GeV, by fitting a gaussian function to the
reconstructed energy distribution:
σ 2 2 2
E S N
= √ + + C2 (2.2)
E E E
where S is the stochastic term, N the noise and C the constant term.
In Fig. 2.12 the energy spectra of the ECAL channels in the barrel and the endcaps
in proton-proton collisions are shown. In Fig. 2.13 the observation of π 0 when
decaying to two photons coming from proton-proton collisions is shown.
• HCAL. The aim of the hadronic calorimeter is to measure the energy of hadron
showers (21, 30). Its requirements are mostly imposed by an accurate measure-
ment of the missing transverse energy (MET). As a result, hermeticity is an asset.
Minimizing the non-gaussian tails in the energy resolution also imposes stringent
requirements in the HCAL design. The absorption capacity is another constraint,
23
Figure 2.11: ECAL supermodule energy resolution. - ECAL energy reso-

lution as a function of electron energy. The upper and lower points correspond to
different data taking conditions. The parameters of the fit can be seen in the figure
legend.
Figure 2.12: Energy spectra of ECAL. - Energy spectra of the individual chan-
nels in the barrel (left) and in the endcaps (right) from 7 TeV collision events (29).
24
2.3 CMS
Figure 2.13: Invariant mass of π 0 candidates. - Mass spectrum of reconstructed

pairs of photons showing a clear peak corresponding to the π 0 mass (29).
since it is located inside the solenoid, surrounding the ECAL. Hence, the HCAL
design maximizes material inside the magnet coil in terms of radiation lengths.
The material used as absorber is brass due to its short interaction length, being
easy to machine and non-magnetic. The active medium uses plastic scintillator
tiles connected with wavelength-shifting fibers (WLS) and hybrid photodiodes to
detect the emitted light.
The HCAL consists of 4 modules: the inner barrel, the outer barrel, the endcaps
and forward calorimeter.
The barrel calorimeter (HB) covers the pseudorapidity region −1.4 < |η| < 1.4.
It consists of 2304 towers with a segmentation δη × δφ = 0.087 × 0.087. It is
constructed out of flat brass absorber plates, parallel to CMS z axis, interleaved
with scintillation plastics.
The outer part (HO) covers the region |η| < 1.26. It contains scintillators with a
thickness of 10 mm. The tiles are grouped in φ matching DT segmentation (angu-
lar sectors of 30 degrees in φ). The outer part increases the hadron calorimetry to
10 interaction lengths. MET resolution is also improved by HO. It is divided in 5
25
rings. The central one, in which the absorber length is minimal, uses 2 scintillator
layers (18 cm thick). The other rings have 1 layer each.
The hadron endcap (HE) covers the pseudorapidity region 1.3 < |η| < 3.0, a
region in which about 34% of the particles from the primary vertex are produced.
It consists of 2304 towers. They are distributed in 14 different η regions, covering
5 degrees in φ each the outermosts and 10 degrees the innermosts.
The hadron forward (HF) gives coverage in the 3 < |η| < 5 region. Its design is
determined by the high particle fluxes in this pseudorapidity region. It consists
of 900 towers and 1800 channels in the 2 modules. They are grouped in 13 η
regions, covering each tower 10 degrees in φ (except the highest η tower, covering
20 degrees). The front face is situated at 11.2 m from the interaction point. The
absorber is 1.65 m deep. The Cherenkov light emitted is transmitted by quartz
fibers to reach the photomultipliers.
The performance of the hadronic calorimeter was studied with simulation and
with test beams before its installation underground (see Fig. 2.14). In Fig. 2.15,
the pT resolution as a function of the pT and for three different jet reconstruction
algorithms is shown (31) in simulation. In data, jet pT resolution is shown in
Fig. 2.16, for γ+jet and dijets.
Figure 2.14: HCAL energy resolution - The jet transverse energy resolution as
a function of the simulated jet transverse energy in 3 different η regions.
26
2.3 CMS
Figure 2.15: HCAL pT resolution - pT resolution as a function of the pT is

shown for four representative η ranges.
27
Figure 2.16: - Jet pT resolutions from γ+jets (red triangles) and dijet asymmetry
(blue points) measurements (32).
• Muon system. The aim of the muon system is the measurement of the trajectory
of the muons and their identification, as well as to provide with seeds for the
different levels of the trigger system. Three different detector technologies are
used to detect muons: cathode strip chambers, resistive plate chambers and drift
tubes. The importance of muons in CMS, a description of these three detectors,
and calibration and parametrization results carried out are presented in the next
chapter.
28
3
Muon detection and

reconstruction in CMS
Muons are charged elementary particles belonging to the second lepton family. They
are unstable particles, that decay in electrons and neutrinos with a mean lifetime of 2.2
µs (typical muons in high energy collisions travel more than 600 m before decaying) (5).
The mass of the muon, 105.7 MeV, is 200 times bigger than that of the electrons. Muons
could be thought of as heavy electrons. As a consequence, they do not emit as much
bremsstranhlung radiation as electrons do. Thus, muons penetrate into matter further
than electrons, being able to go through long distances before being absorbed.
3.1 Importance of muons in particle physics analysis
The importance of muons in the CMS experiment is patent in its name. The “M” of
CMS refers to the “muon” particles, underlining the key role of this particle in the
experiment. There are several reasons that make this particle so interesting from the
physical analysis point of view. First of all, concerning detection, high energy muons are
not absorbed by the detector as they are low interacting particles. This makes possible
to study them in a low occupancy environment, since most other particles produced
in the primary interaction are absorbed in the CMS calorimeters. As a consequence,
physical processes involving muons in their final state are clean, and can be detected
and triggered on with high efficiency (33).
Muon detection is non-destructive and thus by placing multiple layers of muon
29
3. MUON DETECTION AND RECONSTRUCTION IN CMS
detectors, an accurate and robust momentum measurement is achieved. The excellent

transverse momentum resolution of muons in CMS and the almost 4π coverage of the
detector guarantees a good reconstruction of any process related with muons.
Muons are interesting from the physics point of view. The clean muon detection
provides experiments with a powerful tool, as many interesting processes contain muons
in their final state. High transverse momentum muons come often from the decay of
high mass particles. Since these particles are potentially interesting, muon detection is
a reliable way to explore new resonances. Many of the interesting physical processes to
be studied at the LHC can be studied in a muon channel:
• Higgs boson. As described in the previous chapter, the discovery (or rejection)
of the Higgs boson is one of the fundamental goals of the LHC experiment (34).
The Higgs boson, if it exists, will be detectable by its decay into less massive
particles. Its decay into two Z bosons (H → ZZ (∗) ), in which both bosons decay
into two muons is specially interesting. This final state with 4 high momentum,
isolated muons, two positive and two negative, with a dimuon invariant mass near
the Z peak, is one of the most promising channels due to the clean way the muons
are detected (35) despite its low cross section compared to other possible decay
channels.
Other interesting channels in the Higgs boson search are H → W W → lνlν,

H → W W → lνjj and H → ZZ → llνν. The first one is specially important
in the [150 - 180] GeV Higss boson mass window, where the branching ratio for
the golden channel (decay into Z bosons) is not significant enough. These four
channels underline the importance of muon detection in almost the full Higgs
boson mass spectrum.
• High dimuon mass resonances. Possible SM extensions predict the existence

of new gauge fields. High dimuon mass resonances (Z’, W’) could appear as
a consequence, confirming this prediction. The LHC will be able to study the
existence of such resonances up to energies of the order of 4 TeV. This could
be achieved by means of the measurement of high momentum muons, following
decays like Z ′ → µ+ µ− or W ′ → µν.
30
3.2 CMS muon system
• Supersymmetry. Supersymmetry (SUSY) is one of the most famous SM ex-

tensions. Muons are a key part in this search. Due to their properties, they are
important in the discovery of the “sparticles” (squarks, gluinos, sleptons...) and
the measurements of their properties. Examples are a possible decay of neutrali-
nos in dimuon pairs, or the decay of the Minimal SuperSymmetric Model (MSSM)
Higgs bosons, similar to that of the SM.
• Electroweak physics. Electroweak physics is important, especially in the early

phases of the LHC operation. Their study contribute to improve the knowledge
on the proton composition, are sensitive to new physical signatures, are back-
ground of many searches and are useful to test the SM in a new energy regime.
Electroweak physics are mostly studied in the electronic and muonic channels,
thus highlighting the importance of muon detection in this field.
• Top physics. Top quark are produced massively in LHC proton-proton colli-
sions. They decay almost exclusively in a W boson and a b quark. The study
of the muonic decay of the W is preferred over the hadronic one, due to the
clean muon detection and the relatively easy QCD multijet background reduction.
Muons are also important in the b-tagging methods, by analyzing the properties
of the muon inside the jet.
3.2 CMS muon system
As briefly described in the previous chapter, CMS muon system is required to be

accurate in the muon transverse momentum measurement, with excellent muon iden-
tification, over a wide region of momenta in the pseudorapidity region covered. Muons
transversing CMS are first measured in the tracker as charged particles (innermost part
of CMS) and then in the muon system (outermost part of CMS).
CMS uses three different technologies to detect muons (33).
The choice of each technology is mainly driven by two factors: the flux of particles
and the magnetic field in the region considered. In the barrel region (|η| < 1.1), due
to a low arrival flux of particles, a relatively slow detector can be used (≈ 100 ns).
The uniform magnetic field in this region allows the use of Drift Tubes (DTs). In the
endcaps, where the flux of particles and the magnetic field is higher Cathode Strip
31
Chambers (CSC) are deployed, covering the region up to |η| < 2.5. A third technology,
resistive plate chambers (RPC), is used in both barrel and endcaps for timing purposes.
They are fast detectors with good time resolution and coarser position resolution than
the other 2 kind of muon detectors. They are important devices to correctly identify
the bunch crossing from which the detected particles come and for triggering purposes.
The layout of the muon system in one quarter of the longitudinal plane of CMS can be
seen in Fig. 3.1.
Figure 3.1: Layout of the muon system - Distribution of muon chambers in a quarter
of the longitudinal plane of CMS. DT and RPC chambers are used in the barrel, whereas
CSCs and RPC ones in the endcaps
The muon system is also a key part of the trigger system. It operates within the
first level, providing two independent and complementary sources of information (CSC
or DT, and RPC). The high number of chambers in the muon system, and the use of
two different technologies to provide information make the muon trigger to be a robust,
precise and reliable source of information.
32
3.2 CMS muon system
All muon detectors in CMS are gaseous detectors. When a muon traverses the gas
it may produce the ionization of nearby atoms. The difference of electrical potential
applied in the chamber accelerates the free electrons creating an avalanche. The cloud
of electrons reaches the anode wire, being detected as an electrical signal.
3.2.1 Cathode Strip Chambers.
These are the detectors chosen to detect muons in the forward part of CMS. CMS uses
468 CSC to form the Muon Endcap (ME) system. The CSC chambers are trapezoidal
detectors. Each of them consists of 6 gas gaps, with a plane of radial strips and a plane
of anode wires running perpendicularly to the strips (see Fig. 3.2). While electrons are
accelerated towards the anode wire, ions go to the cathodes, measuring 2 coordinates
in each plane. To avoid gaps in the muon system, most of the chambers are overlapped
in the azimuthal angle. By adding information of the different planes of a chamber, a
3D reconstruction of the original trajectory is built. This information is also used for
the Level-1 Trigger. The spatial resolution is about 200 µm and the angular resolution,
10 mrad.
Figure 3.2: CSC representation - Schematic view of an endcap CSC: a six plane
chamber of a trapezoidal shape with strings running radially and wires running across.
33
3.2.2 Resistive Plate Chambers.
They are used for timing purposes due to their fast response. They consist of 2 parallel
planes of plastic material, with a 2 mm gap, with a high voltage applied between them
(see Fig. 3.3).
RPCs can either operate in streamer or avalanche mode. In CMS this second mode
is chosen due to the high flux of particles expected (10 kHz/cm2 ).
In the barrel, there are 1 or 2 RPCs placed coupled to the Drift Tube chamber,
depending on the station. In each of the two endcap rings 36 RPC chambers are used.
Figure 3.3: RPC representation - RPC detectors in CMS are made of 2 parallel planes
of plastic filled with gas.
3.2.3 Drift Tube Chambers.
Since these detectors are a capital part of the thesis, they will be explained with more
detail. The barrel muon system at CMS consists of 250 drift tube (DT) chambers
allocated in 5 iron wheels, which serve to guide the return of the magnetic field lines of
the solenoid, named YB-2 consecutively to YB+2. As a consequence, drift tubes are
under the influence of the residual magnetic field created by the solenoid. Each wheel
is structured in 12 sectors named counter clockwise from S1 to S12, each one covering
30 degrees in the r − φ plane (see Fig. 3.4). Four stations are placed in concentric rings
around the beam line in each sector to measure the muon trajectory. They are labeled
MB1 to MB4 from the innermost to the outermost ones.
34
3.2 CMS muon system
Figure 3.4: Schematic view of the CMS detector. - Transverse view of CMS. The
barrel muon detector elements are denoted as MBZ/N/S where Z is the wheel number, S
one of the 12 sectors and N the station number.
35
The basic detection unity is the drift cell, shown in Fig. 3.5. These drift tubes are
put together in rows (called layers). 4 rows are glued one on top of each other to form
a so called SuperLayer (hereafter SL), measuring either the r − φ coordinate (SL − φ)
or the r − θ coordinate (SL − θ) depending on the wire orientation. These 4 rows are
placed staggered to diminish the probability of losing a muon. Stations MB1, MB2 and
MB3 combine 2 SL − φ and 1 SL − θ, whereas MB4 measures only the r − φ plane (see
Fig. 3.6). In order to confer rigidity to the ensemble a non-sensitive structure, called
honey-comb, is inserted between the SuperLayers.
Figure 3.5: CMS barrel muon drift cell - Representation of CMS drift tube chamber
detection unity.
The drift cell is a 2.5 meters long tube of rectangular section (42 mm long, 13 mm
high). Right in the middle, an anode wire runs along the tube. At a distance of 21 mm,
the cathode strip runs in the same direction. The drift cell is filled with a gas mixture
(15% CO2, 85% Ar) optimized to detect a particle going through it. The 2.3 kV/cm
electric field created inside the tube makes the free electrons ionized after the passage
of the muon to drift towards the wire. The time it takes the electrons to reach the wire
is measured as an absolute time (TDC counts). This time is converted into the drift
time using the calibration procedure detailed in the following sections. This drift time
measured is related to the relevant quantity, the distance, to calculate the passing point
of the charged particle inside the cell. This is done using the electron drift velocity in
the cell. The electric field inside the cell is required to be highly uniform in order to
have a constant drift velocity, and avoid any non-linearities. In one of the two ends of
36
3.2 CMS muon system
the drift tube, the high voltage is provided. This end will be called along the text as
high voltage side (HV). At the other end, the read-out electronics are placed. It will be
called hereafter the front-end side (FE).
Each drift tube gives information of one point in the muon trajectory. When a
muon crosses the barrel region, the DT chambers provide up to a total of 44 points
(32 in the r − φ plane and 12 in the r − θ plane) which are used to build the muon
trajectory. This information is later combined with the data coming from the tracker
to conform the complete muon trajectory. The way the local information given by a
DT chamber is transformed in a full muon path in the barrel region is hierarchical,
building more complex objects as we associate information coming from neighboring
detectors. All the hits coming from a superlayer are fitted to a straight line producing a
so called 2Dsegment (two-dimensional information). All the 2Dsegments in a chamber
are then used to create a 4Dsegment, with 3-dimensional information. They are, in
turn, again associated to obtain the highest order object using DT chambers alone: the
Stand Alone Muon Track (STA) (36). To build the STA track it is necessary to bear
in mind the effect of the magnetic field on the muon path. Since the magnetic field
inside the drift cell is negligible, the path within a muon chamber can be considered
as linear, and reconstructed fitting a straight line. However, in the iron that separates
the drift chambers in a same sector, the magnetic field is of the order of 2 T, bending
the muon trajectory. A dedicated algorithm reconstructs the complete muon trajectory
considering the muon scattering in this dense material and the curvature of the muon
due to the magnetic field.
The muon chambers were tested and commissioned in the laboratory using cosmic
rays. Then, they were assembled at CERN and tested both in the surface and in the
cavern. To check the synchronization with the other detectors, the chambers were
tested using cosmic muons with the full CMS in operation mode. These tests were
performed with the magnetic field switched off and at 3.8 T. The studies presented in
this Chapter correspond to the calibration of the detector carried out using all CMS
subdetector (both at 0 and 3.8 T). To do so, cosmic muons were detected during several
data taking campaigns. CRUZET (Cosmic Run at Zero Tesla) gives name to a series
of campaigns in which the magnetic field of the solenoid was switched off. CRAFT08
(Cosmic Run at Four Tesla 2008) is another of these multiple tests that the CMS
subdetectors have undergone. It was the first global run with the magnetic field risen
37
Figure 3.6: CMS DT chamber - A DT chamber consist of 2 φ SL, a non-active structure

(honey-comb) to give rigidity to the ensemble and, for MB1, MB2 and MB3, a θ SL.
38
3.2 CMS muon system
at its design value. During this data taking period, spanning around one month of
continuous running, more than 300 million cosmic muon events were registered. The
details of some of these runs are presented in Tab. 3.1. The conditions of the detector
(magnetic field) during these data taking periods is shown in Fig. 3.7.
Figure 3.7: Magnetic field - Status of the solenoidal magnetic field in CMS during the
period in which the CRAFT08 was taken. Data with and without magnetic field were
taken for CRAFT08.
These tests served to characterize the response of the drift tubes. Several studies
were carried out using cosmic muons: the local reconstruction in the drift chambers
of the muon tracks, the angular and spatial resolution in the different steps of the
reconstruction, charge misidentification rate, DT trigger performance, reconstruction
efficiencies and calibration and drift velocity measurements are the most important (37).
The spatial and angular hit resolution depend on the drift velocity in the detector
and a good knowledge of it is needed. Consequently, a deep understanding of the
different effects that may have an impact on the drift velocity determination is essential,
in order to keep the resolution as high as possible. Drift velocity in the DTs is studied
in section 3.
CMS muon system is required to be a robust, accurate system. In the barrel
region, DTs should detect and reconstruct the incoming muon with high efficiency.
The reconstruction efficiency of this detector is studied in section 4.
39
# RUN B Field # events analyzed Data Taking

57553 0T 508 705 CRUZET
57539 0T 436 392 CRUZET
557498 0T 154 364 CRUZET
67818 3.8 T 20 303 938 CRAFT
67838 3.8 T 34 316 877 CRAFT
68021 3.8 T 35 822 927 CRAFT
68141 3.8 T 3 520 000 CRAFT
70147 0T 960 000 CRAFT
70170 0T 1 300 000 CRAFT
70675 0T 2 470 000 CRAFT
Table 3.1: CRAFT08 runs analyzed in this study.
3.3 Determination of the drift velocity
The method used in this study to measure the drift velocity is called meantime method,
and it is based on computing the maximum drift time within a cell.
For every 3 consecutive layers in a SL (for example layers i, j and k) the value
Tmean = (Ti + Tk )/2 + Tj (3.1)
corresponds to the time the ionization cloud takes to reach the wire when the muon
crosses the cell next to the cathode (see Fig. 3.8) and it is called Tmean . In this formula,
Ti is the drift time on a cell in the i layer. The meantime is the maximum drift time in
a cell, for muon incident angles below 30 degrees, given the geometry of the drift cell
and their displacement inside a SL, shown in Fig. 3.8. This value is of the order of 400
ns.
Provided the electrical field is uniform enough, the drift velocity can be taken as
constant inside a cell and can be easily computed as half the cell size over the Tmean ,
21000
vd = (µm/ns) (3.2)
Tmean
Time given by the read-out does not only account for the drift time but also for
other contributions as the propagation delay along the cables and the trigger latency
(the delay produced by the trigger system in the data acquisition). The conversion of
40
Figure 3.8: Meantime method computation in a SL - This method is based in the

determination of the maximum drift time on a cell.
this raw time (TDC counts) into a drift time is done through a dedicated calibration
process.
3.3.1 Calibration of the drift tube
The meantime method relies on an accurate calibration of the SL, which essentially gives
the time origin, signaling the passage of a muon through the detector. There are several
sources that contribute to this raw time. Small differences among the propagation time
in different cells in layers are taken into account in the so called relative t0 (few ns).
This time is computed sending LED pulses to every cell, in dedicated calibration runs.
In practice, in the DT chambers, this procedure is applied for a group of cells (2 layers
of cells or a SuperLayer). Once the relative t0 values are taken into account, a raw time
histogram or time distribution (tbox) can be plotted for a group of cells (Fig. 3.9 (left)).
The time offset, in this case around 1400 ns, is known as global t0 or time pedestal and
it is related to the trigger latency and the cabling from the detector to the readout
units. Both relative and global t0 factors integrate the final t0 (37). A typical time box
after calibration can be seen in Fig. 3.9 (right). The time pedestal or t0 corresponds,
41
thus, with the time read when a muon passes close to the wire. This value is computed
for a reference run and stored for each cell in a data base. After calibration procedure,
the intrinsic drift tube resolution is of 4-5 ns.
Figure 3.9: Time box - A typical time distribution (or tbox) (left) before calibration,
(right) after calibration.
In collision mode, muons are expected to arrive every 25 ns. Windows of 25 ns

width are then used to distinguish among different bunch crossings. However, when
collecting cosmics, the muon arrival time to the drift chambers is not known a priori
and thus, is not synchronized with LHC clock. This fact adds an uncertainty of 12.5 ns
in the muon arrival time that propagates to the drift time determination and therefore
affect the drift velocity computation and the chamber resolution.
In this section, 4 different runs, belonging to the CRAFT08 final days, are chosen
to analyze the drift velocity with and without magnetic field. The details of these runs
are presented in Tab. 3.1. Some runs from CRUZET4 (previous to CRAFT08) were
also used for specific studies. All the results presented correspond to one of the the
SL-φ as representative.
The value of Tmean has been computed for each SL − φ1 of MB1, MB2 and MB3
stations in sectors 4 and 10 of each wheel and in every run, by means of equation
3.1. Sectors 4 and 10 are chosen since they are the horizontal ones. Due to the fact
that cosmic rays arrive mostly in the vertical direction, the horizontal chambers will
reconstruct a bigger amount of cosmic muons. An example of a Tmean distribution is
plotted in Fig. 3.10. The graphic on the left shows the Tmean distribution obtained after
the t0 calibration. To compute the right plot, a correction on the arrival time is applied
to make the plot look like the one we have in proton-proton collisions. In this case, the
42
arrival time is introduced as a new parameter in the fit to the muon hits, improving
the resolution. In both cases, the distribution is fitted to a gaussian function, and the
mean value is used to compute the drift velocity by means of equation 3.2. In Fig. 3.11
the drift velocity for all SL−φ1 in every MB1 station of every sector in CMS is plotted,
without the arrival time correction (left), and with it (right). The plots including the
arrival time correction are shown as an illustration of the collision mode plots and this
method is not applied on the rest of the Chapter. In order to quantify the systematic
errors when applying this method the Tmean using layers 1, 2 and 3 of a SL and layers 2,
3, and 4 of the same SL is computed. The r.m.s. of the difference distribution between
these two quantities gives an idea of the error associated with this method. An error
of 0.12 µm/ns and 0.04 µm/ns is obtained in both cases (without and with correction
respectively). However, this procedure used to obtain systematic errors is not sensitive
123 and T 234 .
to a bad calibration, as it affects equally Tmean mean
4000
1000 Entries 35374
Entries 36658
Mean 394.2
800 Mean 392.9
RMS 21.22
RMS 13.2
600
2000
400
200
0 0
300 350 400 450 500 300 350 400 450 500
Time (ns) Time (ns)
Figure 3.10: Tmean distribution - Tmean distribution for the W-2 S4 MB3 station
(run 70147) (a) without and (b) with t0 event-by-event correction.
It is important to notice that the drift velocity absolute magnitude is very sensitive
to the t0 determination procedure. This is the reason why the attention of this study
is mainly focused in the drift velocity variation in different cases and not in its actual
value.
In the following subsections the different factors that may affect the measurement
of the drift velocity are revised.
• The muon incidence angle on the chamber is a factor to bear in mind. CMS is
designed to optimize detection of muons with a maximum angle of 30-35 deg.
43
Drift Velocity (µm/ns)

58 58
56 56
54 54
52 52
50 50
48 W-2 W-1 W0 W1 W2
48 W-2 W-1 W0 W1 W2
S11
S11
S11
S11
S11
S11
S11
S11
S11
S11
S1
S1
S1
S1
S1
S1
S1
S1
S1
S1
S2
S3
S4
S5
S6
S7
S8
S9
S2
S3
S4
S5
S6
S7
S8
S9
S2
S3
S4
S5
S6
S7
S8
S9
S2
S3
S4
S5
S6
S7
S8
S9
S2
S3
S4
S5
S6
S7
S8
S9
S2
S3
S4
S5
S6
S7
S8
S9
S2
S3
S4
S5
S6
S7
S8
S9
S2
S3
S4
S5
S6
S7
S8
S9
S2
S3
S4
S5
S6
S7
S8
S9
S2
S3
S4
S5
S6
S7
S8
S9
S10
S12
S10
S12
S10
S12
S10
S12
S10
S12
S10
S12
S10
S12
S10
S12
S10
S12
S10
S12
Figure 3.11: Drift velocity - Drift velocity for every SL − φ1 in MB1 stations in a run
without magnetic field (a) without and (b) with t0 event-by-event correction.
These are the conditions in the nominal p-p collision mode. Muons with higher
incidence angle will create a different pattern of hit cells by ionized electrons
affecting the measured distance.
• The length of the drift tube, 2.5 meters long, will introduce additional indetermi-
nations. The time taken by the signal created after the avalanche to propagate
along the wire will be added to the drift time, making measured distances longer.
The total drift time will differ in a few ns, depending on where it was produced
along the wire.
• The drift chambers are placed inside the CMS magnet return yoke, thus under the
influence of the residual magnetic field created by the solenoid. This field varies
considerably depending on the chamber position, as can be seen in Fig. 3.12, its
radial component reaching around 0.8 T in the innermost stations (MB1) near the
endcaps. As it will be further explained, this magnetic field affects the electron
drift path.
3.3.2 Effect of the muon incidence angle in the drift velocity compu-
tation
When a muon traverses a DT chamber with a certain incident angle, with respect to
the normal incidence, the electron cloud created by ionization will be lead by electrons
44
1
Br[T]
MB1
0.8 MB2
MB3
MB4
0.6
0.4
0.2
-0.2
-0.4
0 2 4 6
z[m]
Figure 3.12: Radial magnetic field component along the CMS global z coordi-
nate. - In the external parts CMS for MB1 stations a non negligible residual magnetic
field is observed.
45
closer to the wire. In this case, the drift time computed will be smaller than in the case
of vertical incidence, and thus the drift velocity will appear to have bigger values.
This effect, which is of the order of 3% for high incidence angle (greater than 20
degrees), is clearly seen for a 0T run. For that purpose, a group of CRUZET4 runs
(57553, 57539, 57498) has been studied, dividing the angle spectrum in 4 intervals:
muons with incidence angle between 0 and 5 degrees, 5 and 15, 15 and 25 and finally
higher than 25 degrees.
An apparent drift velocity increase when increasing the incidence angle is observed
in Fig. 3.13 for the SL − φ. The drift velocity is plotted using data from CRUZET4,
once CMS was installed inside the cavern. A good agreement between this plot and
previous studies with the detector in surface is observed [REF]. The drift velocity
variations reach around 3% for high incident angles.
Drift Velocity (µs/ns)
56
MB1
MB2
MB3
55
54
0 5 10 15 20 25
Angle (degrees)
Figure 3.13: Drift velocity versus the incidence angle -. Drift velocity versus the
muon incidence angle with data from CRUZET4 runs.
3.3.3 Effect of the signal propagation delay along the DT wire
The time stored by the DT electronics after a muon crossing contains not only the
drift time and trigger latency but also the time the signal takes to reach the front-end.
46
Whereas trigger latency effect is removed when carrying out the calibration process,
this latter contribution remains in the final time used in the reconstruction. In spite
of the high speed of the signal propagation along the wire, the 2.5 meters long wire is
enough to make this effect noticeable when comparing a muon crossing the DT near
the front-end and other from the other side. The delays for two muons separated 1
meter along the wire direction is of the order of 3 ns. It is sizable compared with the
intrinsic time resolution in the drift cell (4-5 ns) and consequently observable.
In order to study this effect, the drift tubes are divided in 5 regions in the wire
direction, as can be seen in Fig. 3.14. For each single region, the Tmean is computed
and the drift velocity is extracted. To know the region the muon traversed we make
use of the information coming from the SL − θ.
Figure 3.14: Chamber division - Graphical description of the chamber split carried
out to study the effect of the signal propagation in the wire and the effect of the magnetic
field in the drift velocity within a chamber.
In order to study this effect independently from the magnetic field effect, several
runs taken at 0 T from the final part of CRAFT are chosen. Details from the runs
analyzed in this section are shown in Tab. 3.1.
The drift velocity computed for the stations W-2 S4 MB1, MB2 and MB3 is pre-
sented in Fig. 3.15 for each of the 5 regions defined. The read-out electronics are placed
in the negative extreme of the local X coordinate, where the computed drift velocity is
maximal, since the length to be traveled by the signal is minimal. The further a muon
crosses the cell towards the opposite side, the bigger the distance to be traversed by
the signal, decreasing the computed drift velocity.
47
55

54
53
52
51
-100 -50 0 50 100

X local (cm)
Figure 3.15: Drift velocity versus chamber local X coordinate - Drift velocity
computed for stations W-2 S4 MB1 (black dots), MB2 (red squares) and MB3 (green
triangles) versus the chamber local X coordinate. Similar plots were obtained for the rest
of the CMS DT chambers with analogous results.
The velocity of the signal propagation along the wire can thus be computed as it is
directly related to the slope of the linear relation in Fig. 3.15, through the expression
2vd
vw = (3.3)
−Tmean · m(vd )
where Tmean is the mean value of the Tmean distribution in each region, vw is the signal
propagation velocity, vd the average drift velocity in the chamber, and m(vd ) the fitted
value for the slope in the drift velocity versus X coordinate plot. The average value for
the signal propagation velocity is compatible with the speed of light.
3.3.4 Effect of the magnetic field in a φ-Superlayer
The magnetic field is expected to affect the drift path of the electrons in their way
to the anode. The CMS superconducting solenoid provides a magnetic field of 3.8
T, essentially parallel to the Z-axis in the global CMS system of reference, inside the
solenoid, with small radial and axial components in the DT chambers. This residual
radial field, orthogonal to the electron drift velocity in a φ-superlayer, will give rise to a
non-zero vz component of the electron drift velocity, making the drift path longer. As
a result, the apparent drift velocity (e.g. the one measured with the meantime method)
will be reduced, in such a way that the higher the radial component of the magnetic
48
field, the lower the apparent drift velocity in that chamber. The axial magnetic field
component is not expected to change notably the drift path, as it will be roughly 0
inside the DT chambers, giving rise to a 0 net force (only in a SL − φ). The typical
effect of the magnetic field on a CMS DT cell over the drift lines is shown in Fig. 3.16.
Figure 3.16: Effect of the magnetic field on the CMS cell drift lines. - The radial
component in a r − φ SL makes the measured drift velocity to be smaller.
Fig. 3.12 shows the radial component of the magnetic field as a function of the
distance to the center of CMS along the beam axis. The radial component is not
uniform and ranges from 0 T to 0.8 T depending on the station and the wheel (z global
coordinate) considered. As a result, the drift velocity may be different when comparing
0 T and 3.8 T runs, the effect becoming more evident in the stations where the radial
component of the magnetic field is highest.
3.3.4.1 Effect of the magnetic field in the drift velocity computation for all
CMS wheels
A comparison between drift velocities obtained in CRAFT runs (see Tab. 3.1) with
and without magnetic field are presented in this section. In the B off case, data from
the three runs analyzed, which share the same conditions, are accumulated in order to
increase the statistics used. The drift velocities are computed for the superlayers φ1 in
stations MB1, MB2 and MB3 from sector 4, in all the 5 wheels, using the meantime
123 ) are used. The results
method previously explained. Hits from layers 1,2 and 3 (Ttmean
are shown in Fig. 3.17. The errors in these plots are statistical and systematical, the
123
latter computed, as previously explained, using the relative variation between Tmean
234 .
and Tmean
On the left plot of Fig. 3.17 the drift velocity for each station under study in both
(B on/off) cases in sector 4 is presented. Comparing B on/off scenarios, a significant
49
decrease of the drift velocity when the magnetic field is switched on, is observed for
both MB1 chambers from the outer wheels (wheels -2 and 2).
55
Drift Velocity Variation (%)

54
2 MB1
53 MB1 MB2
MB2 MB3
1
52 MB3
51 B off B on 0
50 W W W W W W W W W W
-2 -1 0 1 2 -2 -1 0 1 2 W-2 W-1 W0 W1 W2
Figure 3.17: Drift velocity in a 0 and 3.8 T run -(Left) Drift velocity for superlayers
φ1 in stations MB1 (circles), MB2 (squares) and MB3 (triangles) in sector 4 for each wheel.
This plot is divided in two parts by a dashed line. Values on the left correspond to the
0 T case whilst right values are from the 3.8 T case. (Right) The variation between both
magnetic field cases is shown: MB1 (circles), MB2 (squares) and MB3 (triangles).
In Fig. 3.17 (right) the percentile variation of drift velocity (B off case - B on case
/ B off case) is plotted, for the results presented in the upper plot.
The same procedure is applied for the rest of the CMS sectors, obtaining similar
results in the overall effect of the magnetic field on the apparent drift velocity.
3.3.5 Effect of the magnetic field on the drift velocity computation

within a chamber
As already shown in Fig, 3.12, the radial component of the magnetic field changes
considerably along the CMS global z coordinate, mainly in the first station (MB1) of
the outer wheels.
The values for the drift velocities presented in Fig. 3.17 for these chambers can be
understood as an average value within the chamber.
In order to study the magnetic field effect with more detail, the chambers have been
divided in 5 regions or slices along z (see Fig. 3.14) and the drift velocity computed in
each one of them. This study has been done with the hits collected in the superlayer
φ1 in stations MB1, MB2 and MB3 of sector 4, in Wheel -2. In the φ superlayers the
50
wires are disposed parallel to the z axis, with the front-end side placed in the outer
part of each wheel. The effect of signal propagation (bigger apparent drift velocity
for muons crossing near the front end side) acts in an opposite way with respect to
the magnetic field effect. Results are shown in Fig. 3.18. where the drift velocity for
each station of W-2 S4, split into 5 regions, is shown in 0 T (left) and 3.8 T (right)
cases. In the 0 T scenario (absence of magnetic field) there is a decrease of the drift
velocity when increasing the local X coordinate in all three stations (in other words,
when moving from the front-end side of the chamber to the high voltage side), due to
the time invested by the electronic signal to propagate along the wire.

54 54
52 52
50 MB1 50 MB1
MB2 MB2
48 MB3
48 MB3
46 46
-100 -50 0 50 100 -100 -50 0 50 100
Local X (cm) Local X (cm)
Figure 3.18: Drift velocity within a chamber - Drift velocity in SL − φ1 in stations

MB1 (circles), MB2 (squares) and MB3 (triangles) in sector 4 W-2, in (left) a 0 T scenario,
(right) a 3.8 T scenario.
In the presence of magnetic field (3.8 T scenario) the same description applies to
drift velocities in MB2 and MB3, where the radial component of the magnetic field
is almost negligible. In the case of the MB1 station, nevertheless, the effect of the
magnetic field acts in the opposite way with respect to the signal propagation along
the wire, and thus, the points are aligned in a 0 slope straight line, as a result of
the counteract effect of both the magnetic field (decreasing the drift velocity near the
front-end) and the signal propagation delay (increasing the drift velocity near the front
end).
In Fig. 3.19 the same plots are shown, now for W-1. In this case no significant
51
differences can be appreciated between 0 T and 3.8 T scenarios. The radial component
of the magnetic field in this wheel is not high enough to affect the drift velocity mea-
surement, making 0 T and 3.8 T cases quite similar. The same behavior is observed in
wheels 0 and 1, whereas results from W2 led to the very same conclusions extracted
when studying Fig. 3.18.

54 54
52 52
MB1 MB1
50 50
MB2 MB2
48 MB3 48 MB3
46 46
-100 -50 0 50 100 -100 -50 0 50 100
Local X (cm) Local X (cm)
Figure 3.19: Drift velocity within a chamber - Drift velocity within the W-1 S4
SL − φ1 MB1 (circles), MB2 (squares) and MB3 (triangles) in (left) a 0 T scenario, (right)
a 3.8 T run.
To summarize, in Fig. 3.20 the slope extracted from the previous plots (after a
straight line fitting) is depicted for each wheel and station with and without magnetic
field. As it has already been pointed out, only slopes from MB1 W-2 and W2 vary
in both cases (B on/off). For any other station the slope is roughly constant and
falls in a range between −0.008 and −0.006 µm/ns/cm as a consequence of the signal
propagation along the wire.
To isolate the magnetic field effect we have compared the drift velocity in every z
region with and without magnetic field.
Starting from the computation carried out in this section (see Fig. 3.18), we have
worked out the variation for each region between 0 T and 3.8 T scenarios in all CMS
wheels. The analysis is now extended to sectors S3, S4 and S5 and the results are
averaged. In Fig. 3.21 this variation, (B on case - B off case) / Bon case, is shown
for stations MB1, MB2 and MB3 for all the CMS wheels. The variation for chambers
MB1, MB2 and MB3 from W0, W1 and W-1, and for MB2 and MB3 from W-2 and
52
0.01
Slope (µm/ns/cm)
MB1
MB2
MB3
0
-0.01
B off B on
-0.02 W W W W W W W W W W
-2 -1 0 1 2 -2 -1 0 1 2
Figure 3.20: Signal propagation velocity along the wire - Slope from a linear fit
of the drift velocity versus local z position in 0 T and 3.8 T scenarios for sector 5 of each
wheel and stations MB1 (circles), MB2 (squares) and MB3 (triangles).
53
W2 are compatible with 0, as we expected since the radial component of the magnetic
field is very low in these chambers. In MB1 from W-2 and W2, the variation behaves
in the same way that the radial component of the magnetic field does, being maximal
in the wheel outer region and minimal in the inner one (see Fig. 3.12).
These results can be compared with those drawn out during the MTCC [REF]
observing a good agreement with those previous results.
3.4 Determination of DT reconstruction efficiency.
The reconstruction of the muon trajectory is done in several steps. Each of the steps is
characterized by the efficiency of the reconstruction in that step. The hit reconstruction
efficiency is bigger than 99.5% inside the cell and about 97.5% taking into account
border effects of the drift cell. The 2D and 4DSegment reconstruction efficiencies of
the DT chambers is reported in this Section, using CRAFT08 data (cosmic muons).
The computed efficiency is quantified as the fraction of muons crossing a DT cham-
ber in which a 4Dsegment is successfully built. To determine whether a muon actually
traverses a chamber, an extrapolation is performed making use of the hits recorded in
the other stations. In order to guarantee that the extrapolation has been robust, some
quality cuts are applied. The procedure is first carried out in top CMS sectors and
then in the bottom ones. An alternative procedure to compute the segment efficiency
is described in (37). Additional results obtained with CRAFT08 data, as the efficiency
dependence with muon transverse momentum, are also reported here.
3.4.1 Data and sample selection
This study is carried out using data in conditions with and without magnetic field. Data
from 5 different CRAFT08 runs (see Tab. 3.1) have been used: run 70147, belonging to
CRAFT08 final days where the magnetic field was switched off; and runs 67818, 67838,
68021 and 68141 belonging to the last CRAFT08 period with the magnetic field on.
The run at 0T is chosen to compare. The four 3.8T runs are shown to be good runs
(good reconstruction in the whole CMS), and they had all the DT chambers operative
(chambers from S7 and S1 are excluded, since they are vertical sectors, the amount
of muons crossing this chambers and fulfilling the quality conditions are insufficient).
54
2 2
Drift Velocity variation (%)

1 1
0 0
-1 -1
-2 -2
-3 MB3 -3 MB3
-4 MB2 -4 MB2
-5 Wheel -2 MB1 -5 Wheel -1 MB1
-6 -6
-100 -50 0 50 100 -100 -50 0 50 100
X local (cm) X local (cm)
2 2

1 1
0 0
-1 -1
-2 -2
-3 MB3 -3 MB3
-4 MB2 -4 MB2
-5 Wheel 0 MB1 -5 Wheel 1 MB1
-6 -6
-100 -50 0 50 100 -100 -50 0 50 100
X local (cm) X local (cm)
2
1
0
-1
-2
-3 MB3
-4 MB2
-5 Wheel 2 MB1
-6
-100 -50 0 50 100
X local (cm)
Figure 3.21: Drift velocity variation (CRAFT) - Drift velocity variation between
the 0 T and the 3.8 T cases within the SL − φ1 of MB1 (circles), MB2 (squares) and MB3
(triangles) averaged over sectors 3, 4 and 5.
55
A total of 5 million events from each run were analyzed, which are, after the sample
selection and quality cuts, enough to quantify the chamber efficiency.
The aim of this study is to measure the proportion of events where a particular DT
chamber reconstructs a 4Dsegment when a cosmic muon traverses it. First events with
a muon crossing each particular DT chamber are used. To determine the passing point
of the muon we rely on the information delivered by the neighboring DT chambers.
Hits from the other DT chambers in the same sector and wheel are used to build a
muon trajectory. This track is then extrapolated to the test chamber to determine
whether the muon crosses it.
To ensure a good extrapolation point we require the other 3 chambers (e.g. of
MB1, MB2 and MB4 when testing MB3) to have a valid reconstructed 4Dsegment.
Also, in order to simplify the 4Dsegment search and without generality loss, we will
avoid using muons that cross different sectors and wheels, considering only muons that
traverse the CMS bottom/top part within the same wheel and sector. Only events with
one 4Dsegment per chamber are kept. In order to uncouple the 4Dsegment efficiency
from any possible trigger effect, only events triggered by the opposite hemisphere to
that of the test chamber are considered (events triggered by the top part to evaluate
the bottom chamber efficiency and vice versa). We consider only events with a certain
trigger quality: in the transverse plane HH or HL trigger quality (correlated) is required
while in the θ plane any quality is accepted.
This selection process reduces the sample to around 10% of the initial one. Roughly,
from trigger considerations one third of the initial sample is selected when studying the
top sectors. And then, around another third of it fulfils the 4Dsegment requirement,
having 3 4Dsegments in the same wheel and sector. Given the high statistics available,
the selected sample has proved to be enough to test the efficiency in each single station,
and without introducing any bias in the sample.
3.4.2 Muon trajectory and crossing point determination in the test

chamber
The muon crossing point is computed by means of two different approaches. In the
first one, valid only for data taken with no magnetic field, the muon trajectory and
the subsequent extrapolation to the test chamber is done with a linear fit. The other
method relies on the official CMS muon trajectory fitter and propagator in the presence
56
of a magnetic field. This last method can be used both for runs taken at 0T and 3.8T.
A pictorial representation of both extrapolation is shown in Fig. 3.22.
Figure 3.22: Pictorial representation of the reconstruction and extrapolation followed

for both 0 T and 3.8 T cases. Chambers A, B and C represent the chambers used to
extrapolate to the test chamber.
3.4.2.1 Extrapolation for the 0T runs (linear method)
We have implemented a simple linear extrapolation. Hits in the neighboring chambers

were all globally fitted to a unique linear trajectory via a weighted least squares method.
A weight was assigned to every hit in order to account for the uncertainty due to
multiple scattering, depending on the amount of iron traversed by the muon from that
particular hit to the test chamber. This is carried out by assuming a deviation angle
from the straight path equal to the scattering angle distribution width. The scattering
angle distribution width was taken as (5)
r
13.6MeV x x
θ= z 1 + 0.038 ln (3.4)
βcp X0 X0
We take 20 GeV as the typical momentum for a muon crossing CMS (see Fig. 3.42).
In this formula, z is the charge of the incident particle, x the distance traversed and X0
the radiation length of the material traversed, which is 1.757 cm for iron. The weight is
computed as the inverse of the spatial deviation after assuming this scattering angle. In
57
Table 3.2: Width of the multiple scattering angle distribution for a 20 GeV muon, de-
pending on the amount of iron traversed.
Stations Distance (cm) Multiple Scattering Angle
Dist. Width, θ (deg)
MB1 - MB2 45.1 0.17
MB3 - MB4 69.68 0.34
MB2 - MB3 70.95 0.35
MB1 - MB3 116.05 0.74
MB2 - MB4 140.63 1.00
MB1 - MB4 185.73 1.53
Tab. 3.2 the scattered angle distribution width, in degrees, depending on the distance
traversed by the muon between chambers is shown. A perpendicular incident muon is
assumed. Once the fit is performed, the muon crossing point at the test chamber is
determined extrapolating the muon path to the chamber half-height. This procedure
is followed in both r-φ and r-θ planes independently.
In Fig. 3.23 the error in the extrapolated position is shown for each chamber in both
planes (only for top sectors). As expected, in Fig. 3.23 (left), the errors for internal
chambers (from the extrapolation point of view we call internal those being in the
middle of the iron, i.e. MB2 and MB3, and external ones those in the borders, i. e.
MB1 and MB4) are smaller, of the order of few millimeters. The errors in MB2 are
smaller than in MB3 as a consequence of the iron thickness between MB1 and MB2,
being smaller than between MB3 and MB4.
In the r-z plane the picture is different, as only 3 out of the 4 stations contain
information in this plane1 . In this view, Fig. 3.23 (right), the relative values of the
errors can also be explained in terms of the number of hits and their positions when
performing the fit. In this case, MB2 is the only internal station. As a result, the muon
crossing point in this chamber can be extrapolated with the minimal error. Between
MB1 and MB3, both external stations, errors are smaller in MB1, as a consequence of
the iron width between stations. Errors in the MB4 case are rather difficult to predict.
MB4 is the most external station, with a high uncertainty due to the multiple scattering.
1
4Dsegments in r-φ plane contain up to 8 hits while in r-z view contain up to 4 hits
58
×103 ×103
800 1000 MB1
MB2
MB1 MB3
600 MB2 MB4
MB3
MB4
400 500
200
0 0
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Extrapolation error (cm) Extrapolation error (cm)
Figure 3.23: Extrapolation error (linear case) - Error in the extrapolated point in
the r-φ plane (left) and in the r-z plane (right) for all stations in sectors 3, 4 and 5, of all
wheels.
But, on the other hand, the number of hits used to perform the extrapolation is higher
than in the other stations, resulting in a better extrapolation. As a result of these two
opposite effects, the errors for MB4 lie somewhere in between MB1 and MB3.
The extrapolated points can be compared with the actual position of the recon-
structed 4Dsegment at the chamber half-height in both views r-φ and r-z. If we com-
pute the distance between the reconstructed point (Xreco ) and the extrapolated one
(Xextrap ) (hereafter Xreco − Xextrap distribution) in the r-φ plane, the result is shown
in Fig. 3.24 (left) in this case for chamber W-1 S4 MB1. This distribution is fitted
to a gaussian function. The sigma of the gaussian fit (around 3 mm) is higher than
the extrapolation error shown in Fig. 3.23 (left). This may be a consequence of the
simplifications carried out when applying 3.4. In Fig. 3.24 (right) this distribution is
superimposed to those from MB2, MB3 and MB4. The sigma after a gaussian fit of
these distributions (0.27 cm, 0.18 cm, 0.21 cm and 0.37 cm for MB1, MB2, MB3 and
MB4, respectively) qualitatively agree with the errors from Fig. 3.23 (left).
In Fig. 3.25 the mean value (top) and sigma (bottom) for all MB1 CMS top sector
chambers of the gaussian fitted Xreco −Xextrap distributions in the r-φ plane are shown.
The mean values are around 0 and not bigger than half a cm. The sigma values are in
agreement with the errors previously computed in most of the cases, being around 2
mm in average.
The same has been done in the r-z view. In Fig. 3.26 the Yreco − Yextrap distribution
59
Entries 84711
Constant 6233 MB1 W-1 S4
Mean 0.09668 MB2 W-1 S4
10000 MB3 W-1 S4
Sigma 0.2612
5000 MB4 W-1 S4
5000
0-3 -2 -1 0 1 2 3 0
-3 -2 -1 0 1 2 3
Xreco - Xextrap (cm) Xreco - Xextrap (cm)
Figure 3.24: Residuals (linear case) - Residual between the extrapolated muon cross-
ing point and the reconstructed 4Dsegment using the linear extrapolation (left) for the
W-1 S4 MB1 chamber and (right) for all the stations in W-1 S4, in the r-φ plane.
is depicted for chamber W-1 S4 MB1. The sigma of the gaussian fit in the r-z plane is
higher than in the r-φ one. This is a consequence of the lower number of hits used to
build the extrapolation in this case.
In Fig. 3.27 the mean value (top) and sigma (bottom) of the gaussian fitted Yreco −
Yextrap distribution for all MB1 CMS top sector chambers are shown. As previously,
the average values are around 0 and not bigger than half a cm. The values of the
width of the distributions reflect again qualitatively the behavior described regarding
the extrapolation error in Fig. 3.23 (right).
3.4.2.2 Extrapolation for the 3.8T runs (STA based method)
For runs taken with magnetic field on, the muon does no longer follow a linear trajectory
in the r-φ plane. Therefore, we have followed a different approach. We rely on the official
methods developed in CMS, and available in CMSSW for muon trajectory fitting and
extrapolation (36). The position extrapolated is computed propagating from the track
innermost (outermost) position to test MB3 or MB4 in top sectors (MB1 or MB2 in
bottom sectors). In this way, the distance from the propagation origin to the test
chamber is minimal. One has to notice that in the linear case hits from the test
chamber do not take part in the fitting track. However, when computing the fit and
the extrapolation using the official code, by means of the so called StandAlone (STA)
track, hits from the test chamber are used.
60
0.4 0.4
MB1 MB2
0.2 0.2
Mean (cm)
Mean (cm)
0 0
-0.2 -0.2
-0.4 W-2 W-1 W0 W1 W2 -0.4 W-2 W-1 W0 W1 W2
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
0.4 0.4
MB3 MB4
0.2 0.2
Mean (cm)
Mean (cm)
0 0
-0.2 -0.2
-0.4 W-2 W-1 W0 W1 W2 -0.4 W-2 W-1 W0 W1 W2
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
0.5 0.5
MB1 MB2
0.4 0.4
Sigma (cm)
Sigma (cm)
0.3 0.3
0.2 0.2
0.1 0.1
W-2 W-1 W0 W1 W2 W-2 W-1 W0 W1 W2
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
0.5 0.5
0.4 MB3 0.4
MB4
Sigma (cm)
Sigma (cm)
0.3 0.3
0.2 0.2
0.1 0.1
W-2 W-1 W0 W1 W2 W-2 W-1 W0 W1 W2
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
Figure 3.25: Residuals (linear case) - Mean and sigma values of the gaussian fit to the
residual between the extrapolated muon crossing point and the reconstructed 4Dsegment
(top) mean and (bottom) sigma for all CMS top sector DT chambers, using the linear
extrapolation in the r-φ plane.
61
2000
MB1 W-1 S4
Entries 14898
1000 Constant 917.7 MB2 W-1 S4
Mean -0.0852 MB3 W-1 S4
Sigma 0.3098
1000
500
0 0
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
Y reco - Yextrap (cm) Y reco - Yextrap (cm)
Figure 3.26: Residuals (linear case) - Residual between the extrapolated muon cross-
ing point and the reconstructed 4Dsegment using the linear extrapolation (left) for the
W-1 S4 MB1 chamber and (right) for all the stations in W-1 S4, in the r-z plane.
0.8 0.8 0.8

0.6 0.6 0.6
MB2 MB3
0.4 MB1 0.4 0.4
Mean (cm)
Mean (cm)
Mean (cm)
0.2 0.2 0.2
0 0 0
-0.2 -0.2 -0.2
-0.4 -0.4 -0.4
-0.6 W-2 W-1 W0 W1 W2 -0.6 W-2 W-1 W0 W1 W2 -0.6 W-2 W-1 W0 W1 W2
-0.8 -0.8 -0.8
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
0.5 0.5 0.5
MB1 MB2 MB3
0.4 0.4 0.4
Sigma (cm)
Sigma (cm)
Sigma (cm)
0.3 0.3 0.3
0.2 0.2 0.2
0.1 0.1 0.1

W-2 W-1 W0 W1 W2 W-2 W-1 W0 W1 W2 W-2 W-1 W0 W1 W2
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
Figure 3.27: Residuals (linear case) - Mean and sigma of the gaussian fit to the
(top) mean and (bottom) sigma for all CMS top sector DT chambers using the linear
extrapolation in the r-z plane.
62
In Fig. 3.28 (left) the extrapolation error for all MB1, MB2, MB3 and MB4 stations
in top sectors in r-φ is shown. The propagation error is minimal for MB1 and MB4
(as the distance to propagate from the innermost or outermost position is minimal for
them), and maximal for MB3, where the propagating distance is maximal.
In Fig. 3.29 (left) the difference between the extrapolated local coordinate and the
reconstructed one in the r-φ plane is depicted for W-1 S4 MB1. In Fig. 3.29 (right)
the same quantity is plotted for W-1 S4 MB1, MB2, MB3 and MB4. The ordering of
the width of these distributions is in agreement with the ordering of the propagation
error shown before. Compared with the previous linear method, a higher accuracy in
the position determination is reached, since the residual distribution is shown to be
narrower. One has to remember that in the STA case an optimized fit procedure is
established (a successive Kalman filter procedure going back and forth in the track)
[REF], and applied several times, and hits in the test chamber are considered in the
STA method, consequently reaching better results.
MB1 MB1
104 104
MB2 MB2
MB3 MB3
103 103
MB4 MB4
102 102
10 10
1 1
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Extrapolation error (cm) Extrapolation error (cm)
Figure 3.28: Extrapolation error (3.8 T run case) - Extrapolation error for MB1
(black solid line), MB2 (red dashed line), MB3 (green dotted line) and MB4 (blue dashed-
dotted line) in (left) r-φ and (right) r-z planes.
In Fig. 3.30 the mean value (top) and sigma (bottom) of the gaussian fitted to the
Xreco − Xextrap distribution for all CMS top sector chambers are shown, now for the
STA track extrapolation. The widths in this case are of the order of half the widths in
the linear case.
In the r-θ plane the situation is very similar. In Fig. 3.28(right) the extrapolation
error in this plane is shown for all the top sectors MB1, MB2, MB3 and MB4. The
63
Entries 6654
1500 Constant 1389
1500 MB1 W-1 S4
Mean -0.04677 MB2 W-1 S4
Sigma 0.08642 MB3 W-1 S4
1000 MB4 W-1 S4

1000
500 500
0 0
-1.5 -1 -0.5 0 0.5 1 1.5 2 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2
Xreco - Xextrap (cm) Xreco - Xextrap (cm)
Figure 3.29: Residual (3.8 T run case) - Residual between the extrapolated muon
crossing point and the reconstructed 4Dsegment using the StandAlone fitter (left) for the
W-1 S4 MB1 and (right) for all the stations in W-1 S4 in the r-φ plane.
distribution Yreco − Yextrap is also shown (but now for the r-z plane), in Fig. 3.31, for
W-1 S4 MB1 (left) and all the stations in W-1 S4 (right). This distribution is wider
than in the r-φ case, as expected, since the number of hits per fit is smaller. In Fig. 3.32
the mean value (top) and sigma (bottom) of the previous distributions are shown for
all CMS top sector chambers, now for the STA track extrapolation.
3.4.3 Quality criteria
The necessary conditions to guarantee a robust efficiency determination are presented

in this section. This efficiency is defined as a ratio of cases where a particular DT cham-
ber reconstructs a 4Dsegment over the total possible number of cases where a muon
traverses this chamber. Based on the results presented in the previous section, a max-
imum residual value of 5 cm between the reconstructed position and the extrapolated
one is allowed, in both planes and both magnetic field configurations. All chambers
with a 4DSegment reconstructed beyond this 5 cm around the extrapolated coordinate
will be considered as non-efficient ones. In this way, there are three possible sources of
inefficiency: lack of a 2D segment (one of the r-φ or r-z views) inside the 4D segment,
lack of the 4D segment itself or a distance between extrapolated and reconstructed
position above the threshold. Since we are interested in measuring the DT chamber
efficiency, not its accuracy, this threshold of 5 cm comes out to be a rather loose re-
64
1 1
0.5 0.5
Mean (cm)
Mean (cm)
MB1 MB2
0 0
-0.5 -0.5
W-2 W-1 W0 W1 W2 W-2 W-1 W0 W1 W2
-1 -1
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
1 1
MB3 MB4
0.5 0.5
Mean (cm)
Mean (cm)
0 0
-0.5 -0.5
W-2 W-1 W0 W1 W2 W-2 W-1 W0 W1 W2
-1 -1
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
0.3 0.3
0.25 MB1 0.25 MB2
Sigma (cm)
Sigma (cm)
0.2 0.2
0.15 0.15
0.1 0.1
0.05 W-2 W-1 W0 W1 W2
0.05 W-2 W-1 W0 W1 W2
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
0.3 0.3
0.25 MB3 0.25 MB4
Sigma (cm)
Sigma (cm)
0.2 0.2
0.15 0.15
0.1 0.1
0.05 W-2 W-1 W0 W1 W2
0.05 W-2 W-1 W0 W1 W2
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
Figure 3.30: Residual (3.8 T run case) - Mean and sigma of the gaussian fit to the
(top) mean and (bottom) sigma for all CMS top sector DT chambers using the StandAlone
fitter and the SteppingHelix propagator in the r-φ plane.
65
1000 1000
Entries 6556
Constant 847.6 MB1 W-1 S4
Mean 0.07628
MB2 W-1 S4
Sigma 0.144
MB3 W-1 S4
500 500
0 0
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2
Y reco - Yextrap (cm) Y reco - Yextrap (cm)
Figure 3.31: Residual (3.8 T run case) - Residual between the extrapolated muon
crossing point and the reconstructed 4Dsegment using the StandAlone fitter and the Step-
pingHelix propagator (left) for the W-1 S4 MB1 and (right) for all the stations in W-1 S4
in the r-z plane.
0.4 0.4 0.4

MB1 MB2 MB3
0.2 0.2 0.2
Mean (cm)
Mean (cm)
Mean (cm)
0 0 0
-0.2 -0.2 -0.2
-0.4 W-2 W-1 W0 W1 W2 -0.4 W-2 W-1 W0 W1 W2 -0.4 W-2 W-1 W0 W1 W2

S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
0.5 0.5 0.5
0.4 MB1 0.4 MB2 0.4 MB3

Sigma (cm)
Sigma (cm)
Sigma (cm)
0.3 0.3 0.3
0.2 0.2 0.2
0.1 0.1 0.1

S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
Figure 3.32: Residual (3.8 T run case) - Mean and sigma of the gaussian fit to the
residual between the extrapolated muon crossing point and the reconstructed 4Dsegment:
(top plot) mean and (bottom plot) sigma for all CMS top sector DT chambers using the
StandAlone fitter and the SteppingHelix propagator in the r-z plane.
66
quirement (around 20 sigmas of the distance extrapolated to reconstructed position

distribution).
The efficiency values to be computed are essentially 100%, finding only very few
cases where the chamber is not efficient. If, by any reason, the extrapolation is wrongly
determined and gives rise to requiring a (non-existent) 4Dsegment to be present in
the chamber, the error induced in the efficiency calculation is sizeable. In order to
guarantee good quality extrapolations, a cut in the χ2 /ndof of the linear fit is required
to be less than 1 in both planes, for the linear method . In Fig. 3.33 (left)a histogram
of the χ2 /ndof of the linear fit is depicted for both r-z and r-φ planes for the linear
method. Setting a maximum value of 1 is shown to be suitable for this case and not
really restrictive, as most of the sample fulfils this requirement. This cut in the χ2 /ndof
is rather loose but it allows to remove pathological cases.
In addition to this cut, in order to ensure a good precision in the extrapolated
position, a lower cut in the minimum number of hits that have been used to perform
the extrapolation is applied. In Fig. 3.33 (right) the number of hits used to perform the
extrapolation is shown for the STA method in both the r-φ and r-z planes, where the
number of hits from the test chamber are not included in the hit-counting. In Fig. 3.34
the efficiency versus the number of hits in the r-φ plane of the STA track is shown for
the STA based method. We can appreciate that for more than 13 hits (the minimum
number of hits set) the efficiency reaches a plateau for all the stations, therefore a cut
at 14 (9) hits is set for the STA (linear) method respectively, not taking into account
the number of hits from the test chamber. In the r-z plane only tracks with at least
5 hits, apart from those in the test chamber, have been considered in both methods.
Requiring a minimum of 5 hits means, in practice, imposing the existence of at least
two chambers to perform the extrapolation.
3.4.4 Inefficiencies due to geometrical effects
The efficiency is computed in two steps. First we will show the efficiency computation
for the top sectors (all the sectors placed in the top hemisphere of CMS: S2, S3, S4, S5,
and S6). Then we will see the particularities for the bottom sectors (sectors placed in
the bottom hemisphere of CMS: S8, S9, S10, S11 and S12). S1 and S7 are not considered
in this study. Being vertical sectors, the amount of cosmic muons reconstructed is not
enough to perform a complete study on the reconstruction efficiency.
67
×103 10000
r-φ
2000
r-φ r-θ
r-θ
5000
1000
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 0 5 10 15 20 25 30
χ2 / ndof # hits
Figure 3.33: Quality criteria - χ2 /ndof of the linear fit for both r-θ (red dashed line)
and r-φ (black solid line) planes using the linear extrapolation (left) and number of hits in
the STA track (not considering the number of hits in the test chamber) in r-φ (black solid
line) and r-θ (red dashed line) planes (right).
Figure 3.34: Efficiency versus the number of hits - DT 4DSegment efficiency versus
the number of hits in the r-φ view of the STA track (no hits from the test chamber are
considered in the hit-counting) for MB1 (black), MB2 (red), MB3 (green) and MB4 (blue).
68
The previously defined 4DSegment efficiency is computed over the resulting good-
quality muon sample. The results are shown in Fig. 3.35, where the efficiency for all
CMS top sector stations is plotted. This figure shows that efficiencies are high (more
than 90% almost everywhere), but also some other unexpected effects are revealed.
Some differences between stations (e.g. MB3 being more efficient than MB1) can be
appreciated. Also, within the same station, a certain pattern can be observed, depend-
ing on the sector, which is repeated for every wheel. Since the chamber construction
and installation process is the same for all stations and sectors, no intrinsic differences
among efficiencies are expected.
100 100
MB1
Efficiency (%)
Efficiency (%)
95 95
MB2
90 90
85 85
W-2 W-1 W0 W1 W2 W-2 W-1 W0 W1 W2
80 80
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
100 100
MB4
MB3
Efficiency (%)
Efficiency (%)
95 95
90 90
85 85
W-2 W-1 W0 W1 W2 W-2 W-1 W0 W1 W2
80 80
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
Figure 3.35: Efficiency - 4DSegment reconstruction efficiency for all CMS top sector
chambers for a 3.8 T run.
In general, these drops in the efficiency are due to cases where the muon crosses
the test station near the borders. Due to the extrapolation error, a 4Dsegment may
be looked for in a chamber border when the muon has not actually gone through it.
In Fig. 3.36 the DTSegment efficiency versus the chamber position of the extrapolated
crossing point is depicted, where the drop in the efficiency near the chamber borders is
clearly seen. The non-sensitive part between the first cell and the physical limit of the
69
chamber (which should not be considered as an intrinsic inefficiency but of a geometrical

nature) and also some occasional electrical noise may add to this inefficiency.
MB1
MB2
1
1
0.8
0.8
Efficiency
Efficiency
0.6
0.6
0.4
0.4
0.2 0.2
0 0
100 100
Y 50 Y 50
loc 0 150 loc 0
al
(cm 50
100 al 100 150
-50 (cm
-50 0 50 )
) -100 -50 0 l (cm) ) -100 -50 m
-100 a -100 l (c
-150 X loc -150 X loca
MB3 MB4
1 1
0.8 0.8
Efficiency
Efficiency
0.6 0.6
0.4 0.4
0.2 0.2
0 0
100 100
Y 50 Y 50
loc 0 loc 0
al 100 150 al 100 150
(cm
-50 0 50 ) (cm
-50 0 50 )
) -100 -50 m ) -100 -50 m
-150 -100 X local (c -150 -100 X local
(c
Figure 3.36: Efficiency versus the chamber position - Efficiency versus the muon
crossing point position in the test chamber for (top-left) MB1, (top-right) MB2, (bottom-
left) MB3 and (bottom-right) MB4 using the STA based method.
The observed efficiency differences among stations in Fig. 3.35 can be explained as
a combination of this border effect, the requirements imposed in the sample selection
and the chamber geometrical disposition in CMS (see Fig.3.4). The requirement of
having three other 4Dsegments in the same wheel and sector affects unavoidably the
muon crossing point spatial distribution in the test chamber. Since MB2 and MB3
are both internal chambers, and all the four stations have the same length in the z
coordinate, it is unlikely to have a muon crossing MB2 or MB3 z border once the
previous requirement is imposed. Something similar happens in the r-φ plane border
70
region. In the selected sample it is more likely to have muons crossing the border of
the chambers in case of the external ones. In particular, the high size of MB3 in the
r-φ plane “protects” this station against the border effect. The different magnitude in
which the border effects are present in each station can be directly seen in Fig. 3.36.
The extrapolated position for muons in which the test chamber is not efficient is shown
in Fig. 3.37. It is interesting to point out from this figure the efficiency drop due to
the physical gap between MB4 S4 and MB4 S13. As a concluding remark, geometrical
border effects affect the total efficiency with a value that varies from 5% inefficiency
in MB4 S4 to the 0.1% in MB3 S4. Taking into account these considerations, the
differences in efficiency between MB3 stations on one side, and MB4 and MB1 ones
(the external stations) can be understood.
6 6
100 5
100 5
Y local position (cm)
Y local position (cm) Entries 219
50 Entries 367 50
4 4
0 3 0 3
MB1 MB2
2 2
-50 -50
1 1
-100 -100
0 0
-100 -50 0 50 100 -100 -50 0 50 100
X local position (cm) X local position (cm)
2 6
1.8
100 100 5
Y lopal position (cm)
Y local position (cm)
1.6
Entries 4
1.4
50 50 Entries 661 4
1.2
0 MB3 1 0 MB4 3
0.8
2
-50 0.6 -50
0.4
1
-100 0.2 -100
0 0
-150 -100 -50 0 50 100 150 -150 -100 -50 0 50 100 150
X local position (cm) X local position (cm)
Figure 3.37: Inefficiency - Extrapolated position in the test chamber for those muons
where the test station fails to reconstruct a 4Dsegment for MB1 (top-left), MB2 (top-right),
MB3 (bottom-left) and MB4 (bottom-right) in a 3.8T run.
Concerning the efficiency pattern correlated with the sector number, seen in Fig. 3.35,
it is, again, due to the combination of the border effect, the requirements imposed in
71
the sample selection and the chamber disposition in CMS, not designed to receive cos-
mic muons, but muons from the interaction point. In particular, the last two elements
make the muon impact position distribution in a station to be rather different depend-
ing on the sector considered. As a result, not every sector will be affected by the border
effect in the same magnitude. In Fig. 3.38 the muon crossing position distribution in
the test chamber, as given by the extrapolation, is shown for MB1 W-1 in sectors 3, 4
and 5. The population of muons arriving close to the border is higher for sector 3 and
therefore the border effect in this sector is more relevant than in sector 5.
0.15
MB1 W-1 S3
MB1 W-1 S4
0.1
MB1 W-1 S5
0.05
0 -150 -100 -50 0 50 100 150

X local position (cm)
Figure 3.38: Extrapolated position - Extrapolated local x position distribution of the
incident muon in MB1 station Wheel -1, for sectors S3 (red solid line), S4 (blue dashed
line), S5 (green dotted line).
As we are interested in an intrinsic efficiency computation, a fiducial volume is

defined in each test chamber, maximizing its volume to include the whole test station
except a frame of 8 cm width at the borders. It is important to remark that in proton-
proton collision none of this considerations are necessary, since all the muons come from
the interaction point, from which all sectors are identical due to rotational symmetry
of the CMS detector.
72
In the next section, the 4D-Segment reconstruction efficiency in CMS DT chambers

will be quantified, inside this fiducial volume. For that purpose, the muons where the
4Dsegment is reconstructed in the test chamber within a certain margin around the
extrapolated position are computed. It will be applied to both r-φ and r-z planes and
then, to a combination of both.
3.4.4.1 Results for top sectors
As a first step, the efficiency is computed for CMS top sectors. In Fig. 3.39 the efficiency
for all CMS top sectors in the r-φ plane (top) and r-z plane (bottom) is shown for MB1,
MB2, MB3 and MB4 stations for the run taken at 0T, and using the linear method.
In Fig. 3.40 the same quantities are plotted for the STA method, for a 3.8T run.
Comparing both methods we can appreciate the compatibility between both results,
and the high efficiency values for most stations (over 97% in the r-φ view). The r-θ view
presents smaller values for the efficiency (2%-3% lower values) related to the smaller
number of hits in this view, and thus, the less precise extrapolation. Moreover, the
smaller number of hits per station in the r-θ view, 4 hits, compared with the 8 hits in
the r-φ plane makes the loss of a θ segment more likely than the loss of a φ segment.
3.4.5 Results for bottom sectors
The efficiency for bottom sectors for a 3.8T run, using the same analysis procedure,
is shown in Fig. 3.41. The bottom sectors efficiency appears to be lower in the MB3
station, when compared to top sectors. Also the efficiency depends on the sector studied
(being lower for S8 and S12).
The origin of this efficiency drop lies in the lower transverse momenta distribution
of muons reaching the bottom sectors when compared to that in the top sectors. In
Fig. 3.42, the muon transverse momentum distribution is plotted for MB1, MB2, MB3
and MB4 stations for sector 4 (left) in the upper hemisphere and sector 10 (right) in the
lower one. A shift towards lower momentum values can be observed when going from
MB4 S4 to MB1 S4 and then from MB1 S10 to MB4 S10. This shift corresponds to the
muon energy loss when traversing CMS. For instance, in the MB3 case, sector 4, only
0.7% of muons have transverse momenta below 5 GeV/c, while up to 8.4% of muons
going through MB3, sector 10, have less than 5 GeV/c. As most of the muons follow
a path top-down in the detector, it is expected to have lower transverse momentum
73
Efficiency (%)
sectors in (top) r-φ and (bottom) r-θ planes for the linear method (no magnetic field run).
Figure 3.39: Efficiency (linear case) - Efficiency for all DT chambers in CMS top
100
94
96
98
S3
W-2
S6
S9 Efficiency (%) Efficiency (%)
S12
100
100
S3
97
98
99
97
98
99
W-1
S6
S9
S12 S3 S3
W-2
W-2
S3 S6 S6
MB1
W0
S6 S9 S9
S9 S12 S12
S12
S3 S3 S3
W-1
W-1
S6 S6
W1
S6
S9 S9 S9
S12 S12 S12
S3 S3 S3
MB3
MB1
W2
S6
W0
W0
S6 S6
S9
S12 S9 S9
S12 S12
S3 S3
S6 S6
W1
W1
Efficiency (%) S9 S9
S12 S12
100
S3 S3
94
96
98
W2
W2
S6 S6
S3 S9 S9
W-2
S6 S12 S12
S9
S12
74
S3
W-1
S9
S12
100
100
S3
97
98
99
97
98
99
MB2
W0
S6
S9
S12 S3 S3
W-2
W-2
S3 S6 S6
S9 S9
W1
S6
S9 S12 S12
S12 S3 S3
S3
W-1
W-1
S6 S6
W2
S6 S9 S9
S9
S12 S12 S12
S3 S3
MB2
MB4
W0
W0
S6 S6
S9 S9
S3 S3
100
S6 S6
W1
W1
94
96
98
S9 S9
S3 S12 S12
W-2
S6 S3 S3
W2
W2
S9 S6 S6
S12 S9 S9
S3 S12 S12
W-1
S6
S9
S12
S3
MB3
W0
S6
S9
S12
S3
W1
S6
S9
S12
S3
W2
S6
S9
S12
Efficiency (%)
sectors in (top) r-φ and (bottom) r-θ planes for the STA based method (3.8T run).
Figure 3.40: Efficiency (3.8 T run case) - Efficiency for all DT chambers in CMS top
100
94
96
98
S3
W-2
S6
S12
100
100
S3
97
98
99
97
98
99
W-1
S6
S9
S12 S3 S3
S3
W-2
W-2
S6 S6
MB1
W0
S6 S9 S9
S9
S12 S12 S12
S3 S3 S3
W-1
W-1
S6 S6 S6
W1
S9 S9 S9
S12 S12 S12
S3 S3 S3
MB3
MB1
W2
S6
W0
W0
S9 S6 S6
S12 S9 S9

S12 S12
S3 S3
S6 S6
W1
W1
S12 S12
100
94 S3 S3
96
98
W2
W2
S6 S6
S3 S9 S9
W-2
S6 S12 S12
S9
S12
75
S3
W-1

S9
S12
100
100
S3
97
98
99
97
98
99
MB2
W0
S6
S9
S12 S3 S3
W-2
W-2
S3 S6 S6
S6 S9 S9
W1
S9 S12 S12
S12 S3 S3
S3 W-1
W-1
S6 S6
W2
S6
S9 S9 S9
S12 S12 S12
S3 S3
MB4
MB2
W0
W0
S6 S6
S9 S9
S3 S3
100
S6 S6
W1
W1
94
96
98
S9 S9
S3 S12 S12
W-2
S6 S3 S3
S9
W2
W2
S6 S6
S12 S9 S9
S3 S12 S12
W-1
S6
S9
S12
S3
MB3
W0
S6
S9
S12
S3
W1
S6
S9
S12
S3
W2
S6
S9
S12
100 100
MB1 MB2
Efficiency (%)
Efficiency (%)
95 95
90 90
85 85
W-2 W-1 W0 W1 W2 W-2 W-1 W0 W1 W2
80 80
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
100 100
MB3 MB4
Efficiency (%)
Efficiency (%)
95 95
90 90
85 85
W-2 W-1 W0 W1 W2 W-2 W-1 W0 W1 W2
80 80
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
Figure 3.41: Efficiency - Efficiency for all DT chambers in CMS top and bottom sectors
in the r-φ plane for the STA based method(3.8T run).
76
distributions in the CMS bottom part. The unlikely low transverse momentum muons
(pt less than 5 GeV/c) in the top sectors, turn into more numerous in the bottom ones.
Low momentum muons undergo higher multiple scattering when going through the
CMS barrel iron, being more difficult to extrapolate. In Fig. 3.43, the propagation
error in the r-φ plane versus the muon transverse momentum is shown for W-1 S10
MB3. The plot on the right is an inset of the small extrapolation error region. It can
be concluded that low transverse momentum muons present higher propagation errors.
In Fig. 3.44 the 4D-Segment reconstruction efficiency is plotted as a function of muon
transverse momentum for MB2 and MB3.
400
Number of muons / 0.5 GeV

MB1 W-1 S4 MB1 W-1 S10
MB2 W-1 S4 MB2 W-1 S10
300 200
MB3 W-1 S4 MB3 W-1 S10
MB4 W-1 S4 MB4 W-1 S10
200
100
100
0 0
0 10 20 30 40 50 0 10 20 30 40 50
p [GeV] p [Gev]
T T
Figure 3.42: Muon transverse momentum distribution - Muon transverse momen-

tum distributions for MB1 (black solid line), MB2 (red dashed line), MB3 (green dotted
line) and MB4 (blue dashed-dotted line) in (left) an upper sector (S4) and (right) a lower
sector (S10).
The muon propagation through the barrel chambers is performed such that it starts
either from the innermost station (MB1) or the outermost one (MB4). In the case of
computing the extrapolated position in any of these two stations (MB1 and MB4) the
propagation is minimal, taking place inside the chamber, and thus the propagation
error is also small. In the case of the inner stations MB2 and, mostly, MB3 the amount
of iron traversed in the propagation is maximal, so being the propagation error. In
these terms, it is comprehensible to have a wider distribution of the difference between
the extrapolation point and the reconstructed one, leading to a higher rate of cases
where the reconstructed segment lies beyond the 5 cm window imposed. Within these
conditions, it is reasonable that efficiency (in the way it has been defined) in these
77
100 50
102
80 40 10
PT [GeV]
PT [GeV]
60 10 30
1
40 20
1
20 10
10-1
0 0
0 2 4 6 8 10 12 14 0 0.1 0.2 0.3 0.4 0.5
Extrapolation error [cm] Extrapolation error [cm]
Figure 3.43: Extrapolation error versus muon pT - Extrapolation error versus the
incoming muon transverse momentum in the W-1 S10 MB3. The right plot is a zoomed
version of the left one, in the low extrapolation error region.
chambers for low momentum muons is lower. MB1 and MB4, chambers which do not
require any propagation to obtain the extrapolated position, are not sensitive to muon
momentum.
Concerning the differences in the efficiency values among sectors for the MB3 sta-
tion, shown in Fig. 3.41, the situation is illustrated in Fig. 3.45. Here the muon mo-
mentum distribution in MB3 for different bottom sectors is plotted. The differences
between S8 and S12 with respect to S9, S10 and S11 can be easily appreciated. Low
momentum muons, when entering CMS vertically from MB4 S4, are bent by CMS
solenoid in many cases towards S12 or S8. This fact makes these sectors to have a
remarkable amount of low momentum (below 5 GeV/c) muons. Consequently, these
sectors will undergo the previously analyzed effect (high extrapolation error) in a bigger
proportion.
To prevent this effect, a maximum value of the propagation error is set at 1.5 cm
for muons entering the efficiency calculation, which ensures a good extrapolation point.
Fig. 3.46 shows the final efficiencies for all CMS MB1, MB2, MB3 and MB4 stations
for the r-φ view. Fig. 3.47 present the efficiencies now for the r-z view. In these plots,
data from the 4 runs in study have been used (67818, 67838, 68021 and 68141). The
same conclusions pointed out before can be also drawn when observing these plots:
high efficiency in all CMS chambers (above 99% in r-φ), 2% lower efficiencies in r-z and
high uniformity in the efficiency values for different sectors and wheels.
78
100 100
80 80
Efficiency [%]
Efficiency [%]
60 MB2 60 MB3
40 40
20 20
0 0
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
PT [GeV] PT [GeV]
Figure 3.44: Efficiency vs pT - DT 4D-Segment reconstruction efficiency versus the

incoming muon transverse momentum for (left) MB2 and (right) MB3 in S10.
79
1200
Sector 8
Sector 9
1000 Sector 10
Sector 11
Sector 12
800
600
400
200
0 2 4 6 8 10 12 14
PT [GeV]
Figure 3.45: Muon pT distribution - Muon transverse momentum distribution in MB3

station for S8 (blue dotted line), S9 (red dashed line), S10 (dark blue solid line), S11 (green
dotted-short dashed line) and S12 (purple dotted-long dashed line).
80
3.5 Validity tests
In Fig. 3.48 a bi-dimensional plot showing the combined 4D-Segment efficiency for
all sectors (except the vertical ones, S1 and S7) and all wheels and stations is shown.
This efficiency is determined as a combination of both r-φ and r-z efficiencies. Low
efficiency values can be spotted quickly with the help of the colour code, looking for
detector problems or features (e.g. chambers partially off). In the particular case
shown here, the MB4 chamber in sector 9 wheel 2 looks inefficient. Fig. 3.49 shows the
number of hits in each cell (occupancy) of every chamber in a given sector and wheel
(in this case sector 9 and wheel 2). It is shown that MB4 was inactive during part of
the run, giving rise to the inefficiency observed in the previous plot.
100 100
99 99
Efficiency [%]
Efficiency [%]
99.6±0.5 99.9±0.1
98 98
MB1 MB2
97 97
W-2 W-1 W0 W1 W2 W-2 W-1 W0 W1 W2
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
100 100
99 99
Efficiency [%]
Efficiency [%]
98 99.7±0.3 98 99.2±0.9
MB3 MB4
97 97
W-2 W-1 W0 W1 W2 W-2 W-1 W0 W1 W2
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
Figure 3.46: Final efficiency - CMS DT 4DSegment reconstruction efficiency in the r-φ
plane.
3.5 Validity tests
When computing the efficiency with the STA based method, a potential bias arises
since hits from the test chamber are included in the STA track fit procedure. This
is not the case in the linear method where hits from the test chamber are explicitly
81
100 100 100

MB1 MB2 MB3
98 98 98
Efficiency [%]
Efficiency [%]
Efficiency [%]
96 96 96
96.6±1.4 96.5±1.2 97.4±1.4
94 94 94
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S12
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
S3
S6
S9
Figure 3.47: Final efficiency - CMS DT 4DSegment reconstruction efficiency in the r-θ
plane.
Wh2 MB4 100

Wh2 MB3
Wh2 MB2 90
Wh2 MB1
Wh1 MB4 80
Wh1 MB3
Wh1 MB2 70
Wh1 MB1
Wh0 MB4 60
Wh0 MB3
Wh0 MB2 50
Wh0 MB1
Wh-1 MB4 40
Wh-1 MB3
Wh-1 MB2 30
Wh-1 MB1
Wh-2 MB4 20
Wh-2 MB3
Wh-2 MB2 10
Wh-2 MB1
0
S2 S3 S4 S5 S6 S8 S9 S10 S11S12
Figure 3.48: Final efficiency - Combined r-φ and r-z CMS DT 4DSegment reconstruc-
tion efficiency.
82
3.5 Validity tests
80 90
70 80
60 70
60
50
50
40
40
30
30
20
20
10 10
0 0
0 10 20 30 40 50 60 70 80 90 100
# cell
Figure 3.49: Occupancy distribution - Occupancy distribution for Wheel 2 Sector 9.

MB4 chamber was inactive during part of the run (low occupancy).
excluded from the fit and extrapolation. A possible bias may arise in two different
ways: first when performing the sample selection, second when propagating to the test
chamber from the inner or outermost position.
When producing the STA track, hits from the test chamber may help the Producer
to build the track. Efficient cases will then be in higher proportion, biasing the sample.
However, the restrictive cut set in the minimum number of hits in the STA track, in
addition to the requirement of having a 4Dsegment in each of the other 3 chambers
in the same wheel and sector as the one being tested, allows only good tracks to be
computed. These two requirements guarantee that in most cases the STA producer does
not depend on hits in the test chamber. For this clean sample the STA is produced
independently whether a 4Dsegment in the test chamber exists or not. In order to
check this assumption eight 3.8T runs, 4 runs in which one of the stations was inactive
(MB2 W-2 S11) and 4 ones in which all the 4 stations in W-2 S11 were active, were
analyzed (see Tab. 3.3).
If the study is biased, the proportion of events after the sample selection should
be higher for runs having 4 active chambers (as a consequence of the test chamber
contribution to the STA producer). However, this is not what it is observed. Fig. 3.50
83
shows the proportion of events with at least one STA track in two cases: before any
requirement and after asking for a minimum number of hits and a maximum propaga-
tion error, for the 8 runs. In the first case there is a bias effect of around 1% difference
bewteen both kind of runs. This can be understood since the test chamber is helping
the STA producer to build the STA track.
However, in the second case, the quality cuts are somehow allowing only good tracks
to become part of the selected sample, thus rejecting those cases in which the test
chamber is essential to build the STA track. As seen in this figure, the bias introduced
in this study by the use of the STA track is more conceptual than practical. In order
to crosscheck this conclusions, a 0T run has been analyzed independently, with the
non-biased linear method and with the STA based method. Results in both cases are
identical.
The second possible bias source in the study is the contribution of the test chamber
hits to the extrapolation. These hits will make the residual distribution to be narrower,
but will not make inefficient cases become efficient. Bad reconstructed segments (e.g.
with a different slope with respect to the STA track) will not take part in the STA
track, thus avoiding any kind of bias in the extrapolation.
Table 3.3: Runs used to test the possible bias in the study.
Run Active stations in W-2 S11 Total number of events
analyzed (Mev)
67818 MB1-MB2-MB3-MB4 1.9
67838 MB1-MB2-MB3-MB4 1.8
68021 MB1-MB2-MB3-MB4 1.8
68141 MB1-MB2-MB3-MB4 2
69396 MB1-MB3-MB4 1.8
69797 MB1-MB3-MB4 1.9
69594 MB1-MB3-MB4 1.9
69522 MB1-MB3-MB4 2
84
3.5 Validity tests
Graph
1
99.06%±0.24% 97.94%±0.36%
0.95
STA cut
4 chambers
0.9
3 chambers
0.85 Nhits + properror cut
81.6%±1%
0.8 81.3%±1%
1 1.5 2 2.5 3 3.5 4
Figure 3.50: Proportion of events with at least one STA track for the runs listed in
Tab. 3.3 in two cases: no requirement at all or some quality cuts imposed. In red runs with
four active chambers, in black with three chambers. Data are fitted to a constant straight
line.
85
86
4
Muon momentum measurement

in CMS
In Chapter 3 it was shown that CMS aims to obtain numerous physics results through
the analysis of physics events with muons in the final state. As explained there, the
muon transverse momentum is determined in CMS by measuring the curvature of the
particle inside the 3.8 T magnetic field created by the solenoid.
The path followed by the muon inside the solenoid is reconstructed by the inner
tracker. The muon spectrometer located outside the solenoid and inserted in the iron
return yoke of the magnetic field provides a measurement of the muon trajectory in
that region.
Thus, two independent measurements of the muon momentum are available (19).
Optimal resolution is reached when information from both subdetectors is combined
to reconstruct the so-called global muon track (36), from which the muon momentum
is derived. For muons with a transverse momentum . 200 GeV, the resolution is,
nevertheless, dominated by the resolution of the inner tracker measurement.
The muon momentum measurement can be affected by different factors (variation
of the magnetic field with respect to the expected value, bad alignment,...). Important
physical analyses (e.g. those based on the reconstructed dimuon invariant mass) depend
on these deviations, imposing the necessity of knowing them as best as possible.
An accurate simulated MC description of the CMS detector is crucial for the analysis
of many physics topics. Comparison between observed data distributions and predicted
ones, in the framework of different physics models will allow to confirm or discard some
87
4. MUON MOMENTUM MEASUREMENT IN CMS
of them. To validate MC predictions, agreement data-MC, already for the very basic
quantities, such as single muon distributions, should be optimal.
Extensive MC validation studies during the commissioning of the detector and in
particular analyzing cosmic muon data allowed to reach a very good level of agreement
data-MC, already before the startup of the LHC running.
In proton-proton collisions the channel pp → Z+X → µ+ µ− +X allows the selection
of very clean muon samples, once the dimuon invariant mass constraint is applied. This
channel is thus an excellent candidate to improve and define the ultimate performance
of CMS for muon momentum measurement, as well as to tune the MC description.
In this Chapter we explore the possibility to improve the MC muon momentum
description determining additional effects, not accounted for in the MC, regarding muon
pT scale and resolution. The method proposed (hereafter called SIDRA, SImulation
DRiven Analysis) is to fit the MC reconstructed Z mass distribution to the experimental
one including “ad-hoc” functions that modify the MC reconstructed momentum. The
parameters of these functions are the parameters to be fitted. The validity of the
method is tested first on MC data. Afterwards it is applied on data collected during
√
run 2010. Initial studies were driven for a s = 10 TeV scenario (MC tests). However,
data taken at LHC used bunches of protons at 7 TeV. Data analyzed at 7 TeV confirmed
the predictions given by the tests at 10 TeV.
4.1 Muon momentum measurement
A charged particle moving in a magnetic field is subjected to a force q~v × B ~ where

~ the magnetic
q is the charge of the particle, ~v is the velocity of the particle and B
field (38). In modern detectors, the momentum of charged particles is measured using
high magnetic fields. The charged particles created in the collision are bent due to
the effect of a magnetic field. In a solenoidal magnetic field, as the one CMS uses,
the trajectory of these particles is circular in the transverse plane. The radius of the
circular trajectory depends on the velocity of the particle
pT
R= (4.1)
0.3 · q · B
with pT being the transverse momentum (in GeV) of the particle, B the magnetic field
(in Teslas), R the radious (in meters) and q the charge of the particle in proton charge
units.
88
By measuring the radius of curvature of the circular trajectory we can measure the
momentum of the charged particle in this plane (transverse momentum) knowing the
strength of the magnetic field (B).
The accuracy of transverse momentum measurement of a charged particle inside a
magnetic field is influenced by a number of different effects.
The intrinsic uncertainty is due to the finite track position resolution of the device.
The size of this effect is constant in 1/pT ,
∆(1/pT ) = κ
and thus, the relative resolution increases linearly with the momentum of the particle,
∆pT
= κ · pT
pT
Other effects contribute to degrade the momentum measurement. Multiple scattering
effects or magnetic field uncertainties induce an additional relative uncertainty that is
essentially constant with the momentum,
∆pT
=β
pT
Finally, if the particle traverses a large quantity of material, it may suffer strong energy
losses, essentially independent of momentum. The effect on the uncertainty will be:
∆pT γ
=
pT pT
In the most general case, all these effects will have an impact on the measurement, and
thus the general expression for the uncertainty will be,
∆pT γ
= κ · pT ⊕ β ⊕
pT pT
The expected muon momentum resolution in CMS is presented in Fig. 4.1, both
for the central part of the detector (left panel) and for the forward/backward regions
(right panel). Predictions are shown for muons reconstructed only with the inner tracker
detector, with the muon system and when both are combined. The resolution is . 1%,
in the whole η interval, for the low and medium momentum range (p < 100 GeV) (it
is limited to ∼ 0.8% by multiple scattering effects at low momentum), increasing to
∼ 10% for high momentum (p ≃ 1 TeV).
89
∆ p/p
∆p/p
0.0<η<0.2 1.8<η<2.0
10-1 10-1
10-2 10-2
Full system Full system

Muon system only Muon system only
Inner Tracker only Inner Tracker only
10-3 10
-3
3
10 102 103 10 102 10
p[GeV/c] p[GeV/c]
Figure 4.1: Expected muon momentum resolution in CMS. (Left) Barrel region. (Right)
Endcap region.
The resolution for muon momentum measurement only in the muon system varies
from 10 to 20 − 30%, depending on the momentum and pseudorapidity region.
Fig. 4.2 presents the reconstructed dimuon invariant mass distribution, obtained
with a large pp → Z + X → µ+ µ− + X MC sample and it constitutes the expected
experimental distribution once a Z-optimized selection process is applied on the simu-
lation. This simulation is composed of a MC generator to emulate the collision process
(hard interaction), the parton shower and hadronization, and a detector simulator. The
expected performance of CMS as a muon spectrometer is reproduced with the GEANT4
package (39). The complete CMS geometry, including construction tolerances and the
detailed description of the materials, together with the magnetic field map are included
in the program.
Any inaccuracies in the simulation, such as alignment errors, discrepancies in the
magnetic field value, incorrect evaluation of the material budget, etc. will give rise to
a discrepancy between the simulated and the experimental distributions. Agreement
data-MC can be retrieved if the reconstructed momentum in the simulation is further
modified according to the expressions that describe the physical effects addressed before.
90
# events / 0.2 GeV

6000
4000
2000
0
80 85 90 95 100
Mµµ [GeV]
Figure 4.2: Reconstructed invariant mass distribution from a high statistics (1M event)
simulated pp → Z + X → µ+ µ− + X sample.
Muon momentum tuning will reflect in the simulated Z invariant mass distribution.
In this Chapter we have considered two possible effects:
• Scaling factors in the muon momentum. They induce a shift in the peak position
of the reconstructed versus real momenta distribution.
• Extra resolution effects. They produce a widening of the reconstructed versus

real momenta distribution.
Moreover, depending on the origin of these effects, they will impact on the mea-
surement either of pT or of 1/pT , thus leading to different parametrization in terms of
the muon momentum:
1. Global scaling factor in the muon momentum, due for instance, to a global scaling
factor in the magnetic field,
pT ′ = (1 + δ) · pT (4.2)
2. An extra smearing term constant with momentum, as the one due to local differ-
ences in the magnetic field map or in the multiple scattering effect,
∆pT
= σM S (4.3)
pT
91
3. A constant shift in 1/pT ,

1 ′ 1
= + δa
pT pT
that translates into a pT -scaling effect linearly increasing with the muon momen-
tum,
pT ′ = (1 + δa · pT ) · pT (4.4)
4. An extra resolution effect, following the same parametrization as an intrinsic

resolution term, i.e. constant in 1/pT ,
∆(1/pT ) = σintrinsic
It translates into a smearing factor linearly depending on the momentum,
∆pT
= σintrinsic · pT (4.5)
pT
Studies performed with low mass resonances samples have shown that the effects
constant in pT are negligible (40, 41). For this reason, only effects linear in pT will be
considered hereafter (they will be referred as δ(1/pT ) and σ(1/pT ) in this Chapter).
Figs. 4.3 and 4.4 illustrate how the reconstructed dimuon-invariant mass distri-
bution is affected by these two effects presented above. In the case of Fig. 4.3 the
reconstructed dimuon invariant mass distribution is shown, where the transverse mo-
mentum of every muon is scaled with two different linearly dependent scaling factors
(δ(1/pT ) = 0.03 TeV−1 and 0.3 TeV−1 ). Fig. 4.4 illustrates the case where the re-
constructed pT of every muon is smeared according to an intrinsic resolution-like term
(σ(1/pT ) = 0.6 TeV−1 ).
The main effect of a scaling factor is a displacement of the peak position with
respect to the original one. In addition to this shift, the result of a pT -dependent scale
factor is that the distribution is somewhat widened in an asymmetric way, more visible
in the case of δ(1/pT ) = 0.3 TeV−1 .
The net effect of the additional smearing term is a visible widening of the recon-
structed dimuon mass distribution, remaining the peak almost in the same position.
In general, all these effects may affect the muon momentum measurement at the same
time, being the peak interval (88 GeV < mµµ < 94 GeV) the most sensitive region to
them.
92
7000 δ (1/pT) = 0
δ (1/p ) = 0.03 TeV -1
# events / 0.2 GeV

6000 T
δ (1/p ) = 0.3 TeV -1
T
5000
4000
3000
2000
1000
0
80 82 84 86 88 90 92 94 96 98 100
Mµµ [GeV]
Figure 4.3: Reconstructed invariant mass distribution from a high statistics (1M events)
simulated pp → Z + X → µ+ µ− + X sample. The pT of every muon is scaled with a
scaling factor δ linearly dependent on the muon pT . Two values of δ are shown: δ(1/pT ) =
0.03 TeV−1 (red, dashed histogram) and δ(1/pT ) = 0.3 TeV−1 (blue, dotted histogram).
The original distribution (δ(1/pT ) = 0) is shown as a solid black line.
σ (1/pT) = 0
σ (1/pT) = 0.6 TeV-1
6000
# events / 0.2 GeV
4000
2000
0
80 85 90 95 100
Mµµ [GeV]
Figure 4.4: Reconstructed invariant mass distribution from a high statistics (1M events)
simulated pp → Z + X → µ+ µ− + X sample. The pT of every muon is smeared with an
additional Gaussian function of width σ(1/pT ) = 0.6 TeV−1 (red, dashed histogram). The
original distribution (σ(1/pT ) = 0) is shown as a solid black line.
93
4.2 Method
At generator level the Z-dimuon invariant mass distribution is well described by the
different MC generators and hardly any differences are appreciated in the Z peak neither
between LO and NLO predictions, nor due to PDF effects. Several MC studies have
shown the robustness of the simulated Z mass distribution, showing very little variations
due to the different theoretical assumptions to generate it. Therefore, experimental
effects will be the only ones significantly distorting the Z mass shape. Among them,
the most relevant ones will be muon scale and resolution and they will have a direct
impact in the reconstructed Z-mass distribution.
The method proposed is to fit the MC reconstructed Z mass distribution to the
experimental one: the muon description in the MC is allowed to vary according to
certain functions and the following binned likelihood function, arising from a Poissonian
distribution, is minimized:
i=nbins
X
− log L = [NiMC (δ(1/pT ), σ(1/pT )) − ndata
i × log NiMC (δ(1/pT ), σ(1/pT ))] (4.6)
i=1
where the index i runs over all bins in the invariant mass distribution between 70 GeV
and 110 GeV, in 1 GeV bins. The variables ndata
i and NiMC (δ(1/pT ), σ(1/pT )) are the
number of entries in the i mass bin of the experimental and MC distributions, respec-
tively.
The pT of the muon is distorted using the previous expressions. The MC Z-mass
distribution is consequently modified and for every mµµ bin, N MC is a function of the
parameters modifying the reconstructed momentum, N MC = N MC (δ(1/pT ), σ(1/pT )).
δ(1/pT ) and σ(1/pT ) are the parameters of the minimization.
Fig. 4.5 illustrates the method. A small subset of the MC Z-sample is taken to
play the role of the data (“fake data” sample), in this case of a 1 pb−1 equivalent
luminosity. Fig. 4.5 (left) shows the Z reconstructed distribution from this “fake data”
sample, where the pT of the muons is distorted simulating a possible experimental case.
In this example a smearing effect has been simulated. The disagreement between the
“fake data” and MC can be clearly noticed.
The pT of the muons in the MC sample is now described in terms of the functional
forms defined before, and expression (4.6) is minimized in order to recover the distortion
introduced.
94
4.3 Validation of the method with MC samples
Fig. 4.5 (right) shows the same “fake data” distribution and the MC distribu-
tion resulting from the minimization. The agreement between fitted MC and “fake
data” is significantly improved. The extra smearing introduced in the fake data sample
(σ(1/pT ) = 1.5 TeV−1 ) is recovered by the fit, as σ fitted (1/pT ) = 1.53 TeV−1 .
200
Fake Data (L=1 pb-1, σ=0.05)
Fake Data (L=1 pb-1, σ=0.05) 100
180 -1 Fitted MC (L=545 pb-1, σ=0.051)
Reference MC (L=545 pb )
160
# events / 1.2 GeV

# events / 1.2 GeV
80
140
120
60
100
80
40
60
40 20
20
0 0
70 80 90 100 70 80 90 100
Mµµ [GeV] Mµµ [GeV]
Figure 4.5: “Fake data” Z invariant mass distribution (black dots) and MC Z invariant
mass distribution (red histogram). (Left) The MC prediction is produced with ideal de-
tector conditions. (Right) The MC prediction is fitted to the “fake data” sample including
terms distorting the pT of the muons.
4.3 Validation of the method with MC samples
The method was established with a high statistics MC sample of pp → Z+X → µ+ µ− +

√
X events at s = 10 TeV, generated with Pythia (42, 43). From this sample, smaller
sub-samples were selected to play the role of fake data sets. The data process, generator,
number of events generated, cross section and equivalent integrated luminosity of the
process are summarized in Table 4.1.
Z-boson candidates are selected from those events passing the single muon trigger
path. Events are required to have two opposite charge, high momentum (pT > 20 GeV)
muons in the pseudorapidity range |η| < 2.
Certain quality criteria are imposed to reconstructed muons (46): the muon is
required to have been reconstructed with two different reconstruction algorithms. The
χ2 /ndf of the global muon track fit should be less than 10 and the number of hits used
to fit the track in the inner part of the detector should be greater than 10. The impact
95
Table 4.1: MC data sample used in this study. Cross section quoted include the Branching
Ratio of Z→ µ+ µ− . The Z cross section is computed at NLO with MCFM (44) program
and MSTW08 (45) NLO Central PDF.
Process Generator # events σ Lint
(pb) ( pb−1 )
pp → Z + X → µ+ µ− + X PYTHIA 1.3 M 2403 545
parameter in the transverse plane is required to be below 0.2 cm. In addition, muons
must be isolated. A relative isolation definition is used; it is defined as the sum of the
p
pT of the tracks, in a cone of ∆R = ∆η 2 + ∆φ2 = 0.3, centered in the muon track,
and normalized to the muon pT . A muon is considered to be isolated if the isolation
variable is smaller than 0.1. These requirements are set in order to get a clean Z boson
sample. More details about this set of cuts can be seen in Chapter 6. In Table 4.2
the number of events remaining in the MC samples, after the different requirements is
shown. The last column of the table indicates the expected number of selected events
for an integrated luminosity of 1 pb−1 .
Table 4.2: Number of events after the different steps in the selection process. The
Acceptance column indicates the number of events with both muons in the acceptance
region |η| < 2, and the Two sel. muons column refers to the number of events containing
two muons of opposite charge and fulfilling the quality and selection requirements (high pT
and isolation for Z → µ+ µ− selection). Last column gives the expected number of selected
events for an integrated luminosity of 1 pb−1 at 10 TeV.
Process # events Trigger Acceptance 2 sel. µs #ev./pb−1
pp → Z + X → µµ + X 1309714 622940 351795 227890 418
With the presented selection requirements, the residual backgrounds are due to
other electroweak processes (such as W → µν or Z → τ τ ) and tt̄. However, the total
background contribution estimated using MC samples is negligible (amounts to less
than 1% of the final sample) and can be ignored.
96
4.4 Results on MC using fake data sets
We have established the performance of the method, from a quantitative point of view,
with the MC sample. Muons in the “fake data” sample have been distorted according
to the effects presented before (with different, known, input parameters) and their
dimuon invariant-mass distribution computed. The reconstructed dimuon distribution
from the remaining MC sample is fitted to this “experimental” one and the output
parameters are compared to the input ones. In this part, two luminosity scenarios are
explored. This method was initially developed for the beginning of the data taking, in
a low statistics scenario. A sample of Lint = 1 pb−1 serves to emulate a low statistics
scenario, for the very beginning of the data taking. However, most of the analyses will
benefit from a bigger amount of data. A second scenario (with a luminosity of Lint = 10
pb−1 ) is explored to check the capabilities of the method in these conditions.
The initial multidimensional problem reduces to a one-dimensional one if only one
of the effects distorting the sample is considered. Fig. 4.6 shows the −(log-likelihood)
curves for two particular cases of the one-dimensional problems: a pT -dependent scale
factor term with δ(1/pT ) = 1. × 10−2 TeV−1 and an additional intrinsic-like term with
an input σ(1/pT ) = 2. × 10−2 TeV−1 .
In all cases the −(log-likelihood) functions are parabolic around the minimum. For
the smearing effect, it exhibits a double parabola structure symmetrically located with
respect to a null distortion parameter1 .
Both minima will get closer when dealing with distortion values close to zero, making
a rather broad minimum in the limit of a non-distorted sample (σ ∼ 0).
To test the method for a 1 pb−1 luminosity scenario, and given the simulated MC
statistics (equivalent to 545 pb−1 ), 545 independent tests with 1 pb−1 luminosity
subsets behaving as “fake data” can be made, taking in each case the remaining of the
Z sample as the MC sample to be fitted. To test the method for a 10 pb−1 luminosity
scenario, the available MC statistics only affords 54 independent tests.
1
Due to the nature of the fit, positive and negative values of the sigma parameter are equally
probable.
97
-30840
-805 -30860
- log-likelihood
- log-likelihood
-810 -30880
-30900
-815
-30920
-820
-30940
-825 -30960
0.005 0.01 0.015 -0.04 -0.02 0 0.02 0.04

δ (1/p ) [TeV-1] σ (1/p ) [TeV-1]
T T
Figure 4.6: −(Log-likelihood) curves for: (left) a scale factor linearly dependent with the
momentum δ(1/pT ) = 1. × 10−2 TeV−1 and (right) an additional intrinsic-like term with
σ(1/pT ) = 2. × 10−2 TeV−1 .
4.4.1 Constant shift in 1/pT
The pT of the muons is modified with a constant shift in the inverse of the momentum.
This effect is equivalent to a linearly pT -dependent scale factor:
pT ′ = [1 + δ(1/pT ) · pT ] · pT (4.7)
This modification is applied to muons in “fake data” samples of 1 and 10 pb−1 in-
tegrated luminosity, and the full MC is fitted to them, trying to retrieve the input
distortion parameters.
The fitted scale factors when dealing with 1 pb−1 luminosity data sets are plotted
in Fig. 4.7 (left) and Fig. 4.8 (left), once they are distorted with scale factors δ(1/pT ) =
0.3 TeV−1 and δ(1/pT ) = 0.03 TeV−1 , respectively.
The case of no distortion (δ(1/pT ) = 0) is presented in Fig. 4.9. The right plots in
the same Figs. show the distribution of the fitted parameters. The mean values are in
agreement with the input value in all three cases.
The width of these distributions are of the order of 0.041 TeV−1 . They are in
agreement with the expected error values from the −(log-likelihood) curves, like the
one plotted in Fig. 4.10, shown as an example.
Table 4.3 summarizes the recovered values (mean and width of the distributions of
fitted parameters) for three different input values of the scale factor, and for the two
luminosity scenarios considered. We can conclude that an overall linearly pT -dependent
98
0.45
χ2 / ndf 6.732 / 10
100 Constant 104.8 ± 5.7
0.4
# events / 0.02 TeV-1

Mean 0.3004 ± 0.0018
Sigma 0.04105 ± 0.00138
80
δ (1/p ) [TeV-1]
0.35
60
0.3
T
40
0.25
20
0.2
0
0 100 200 300 400 500 0 0.1 0.2 0.3 0.4 0.5 0.6
# Luminosity set δ (1/p ) [TeV-1]
T
Figure 4.7: (Left) Scale factor recovered for each 1 pb−1 “fake data” luminosity set.
(Right) Histogram of fitted values. The input value is δ(1/pT ) = 0.3 TeV−1 .
0.15 120
χ2 / ndf 11.79 / 11
Constant 103.4 ± 5.6
100 0.03023 ± 0.00183
Mean
0.1 Sigma 0.04117 ± 0.00137
δ (1/p ) [TeV-1]
80
0.05
60
T
0
40
-0.05
20
-0.1
0
0 100 200 300 400 500 -0.3 -0.2 -0.1 0 0.1 0.2 0.3
T
99
0.1 χ2 / ndf 11.66 / 9
100 Constant 104.1 ± 5.7

Mean 0.0003945 ± 0.0018440
0.05 Sigma 0.04108 ± 0.00145

δ (1/p ) [TeV-1]
80
0 60
T
-0.05 40
20
-0.1
0
0 100 200 300 400 500 -0.3 -0.2 -0.1 0 0.1 0.2 0.3
T
(Right) Histogram of fitted values. The input value is δ(1/pT ) = 0.
-827
-827.2
- log-likelihood
-827.4
-827.6
-827.8
-828
-828.2
0.25 0.3 0.35

δ (1/p ) [TeV-1]
T
Figure 4.10: −(Log-likelihood) shape for the pT -dependent scale factor term for a par-
ticular 1 pb−1 “fake data” luminosity set when the “fake data” is distorted with an input
scale factor δ(1/pT ) = 0.3 TeV−1 . The error interval defined by the increase of the −log-
likelihood in 0.5 from the minimum units is [0.264, 0.33] TeV−1 , from which an average
error of ∆δ(1/pT ) = 0.033 TeV−1 is derived.
100
25
0.33 χ2 / ndf 1.218 / 1
Constant 26.07 ± 4.72

0.32 20 Mean 0.3 ± 0.0
Sigma 0.0163 ± 0.0021
δ (1/p ) [TeV-1]
0.31
0.3 15
T
0.29
10
0.28
0.27 5
0.26
0
0 10 20 30 40 50 0 0.1 0.2 0.3 0.4 0.5 0.6
T
0.08
χ2 / ndf 0.04221 / 2
0.07 25 Constant 27.95 ± 4.72
0.06 Mean 0.02924 ± 0.00211

0.01541 ± 0.00155
δ (1/p ) [TeV-1]
Sigma
20
0.05
0.04 15
T
0.03
0.02 10
0.01
5
0
0
0 10 20 30 40 50 -0.3 -0.2 -0.1 0 0.1 0.2 0.3
T
101
22
0.04 χ2 / ndf 1.268 / 3
20
Constant 25.18 ± 4.70

18
Mean -0.0005359 ± 0.0023204
0.02
δ (1/p ) [TeV-1]
16
Sigma 0.01671 ± 0.00211
14
0 12
T
10
-0.02 8
6
4
-0.04
2
0
0 10 20 30 40 50 -0.3 -0.2 -0.1 0 0.1 0.2 0.3
T
(Right) Histogram of fitted values. The input value is δ(1/pT ) = 0.
-16641
- log-likelihood
-16641.5
-16642
-16642.5
0.26 0.28 0.3 0.32
δ (1/p ) [TeV-1]
T
Figure 4.14: −(Log-likelihood) shape for the pT -dependent scale factor term for a par-
ticular 10 pb−1 “fake data” luminosity set for an input scale factor δ(1/pT ) = 0.3 TeV−1 .
The error interval defined by the increase of the −log-likelihood in 0.5 units from the min-
imum is [0.267, 0.303] TeV−1 , from which an average error of ∆δ(1/pT ) = 0.018 TeV−1 is
derived.
102
scale factor can be determined, if it is larger than ∼ 0.041 TeV−1 , even with very low
collected luminosity (Lint ∼ 1 pb−1 ).
Table 4.3: Mean and width of the distributions of fitted scale factor parameter (all
magnitudes are in TeV−1 ).
Lint = 1 pb−1 Lint = 10 pb−1

δinput (1/pT ) < δfitted (1/pT ) > σ(δfitted (1/pT )) < δfitted (1/pT ) > σ(δfitted (1/pT ))
0.3 0.3 0.041 0.3 0.016
0.03 0.03 0.041 0.03 0.015
0. 3.94 × 10−4 0.041 −5.36 × 10−4 0.017
4.4.2 Intrinsic resolution-like term
Muon pT in the “fake-data” sample is now distorted with a term of the form:
∆pT /pT = σ(1/pT ) · pT
The σ(1/pT ) parameters obtained when the MC is fitted to the different 1 pb−1 “fake-
data” luminosity sets are presented in Fig. 4.15, Fig. 4.16 and Fig. 4.17. Input values
are σ(1/pT ) = 1.5 TeV−1 , σ(1/pT ) = 0.6 TeV−1 and no distortion. The histogram of
the fitted values is also shown in the same Figs. The mean values agree well with the
input ones, with a precision of ∼ 0.1 TeV−1 .
In all cases there is an agreement between the expected precision and the behavior of
the likelihood function (Fig. 4.18 shows a particular example). Here again, the fact that
the two parabolas of the −(log-likelihood curve) approach to each other (see Fig. 4.18)
when trying to fit in the non distortion case leads to a slightly higher uncertainty
(∼ 0.15 TeV−1 ).
The same exercise is repeated with 10 pb−1 “fake-data” luminosity samples. The
results are presented in Fig. 4.19, Fig. 4.20 and Fig. 4.21. Mean values of the fitted
parameters distributions agree well with the input values. Concerning precision, a
factor of 2 to 3 of improvement is observed, depending on the case. Now, there is
also an agreement between the width of the histogram and the behavior of the −(log-
likelihood) curve (like the one shown in Fig. 4.22).
103
2 90
χ2 / ndf 21.74 / 18
1.9 80
Constant 65.29 ± 3.86

1.8 70 Mean 1.499 ± 0.006
σ (1/p ) [TeV-1]
Sigma 0.1277 ± 0.0052

1.7 60
1.6 50
T
1.5 40
1.4 30
1.3 20
1.2 10
1.1 0
0 100 200 300 400 500 1 1.2 1.4 1.6 1.8 2
# Luminosity set σ (1/p ) [TeV-1]
T
Figure 4.15: (Left) Additional intrinsic-like term recovered for each 1 pb−1 luminosity
set. (Right) Histogram of fitted values. The input value is σ(1/pT ) = 1.5 TeV−1 .
0.9 100
χ2 / ndf 20.07 / 12
Constant 89.55 ± 5.26
0.8 0.6007 ± 0.0042

80 Mean
σ (1/p ) [TeV-1]
Sigma 0.0936 ± 0.0038

0.7
60
0.6
T
0.5 40
0.4 20
0.3
0
0 100 200 300 400 500 0 0.2 0.4 0.6 0.8 1 1.2
T
104
90 χ2 / ndf 116.9 / 21
0.4
80 Constant 44.42 ± 3.14

Mean 0.002768 ± 0.007639
0.2 70 0.1542 ± 0.0082

σ (1/p ) [TeV-1]
Sigma
60
0 50
T
40
-0.2 30
20
-0.4
10
0
0 100 200 300 400 500 -0.6 -0.4 -0.2 0 0.2 0.4 0.6
T
set. (Right) Histogram of fitted values. The input value is σ(1/pT ) = 0.
-789.2
- log-likelihood
-789.4
-789.6
-789.8
-790
-0.4 -0.2 0 0.2 0.4

σ (1/p ) [TeV-1]
T
Figure 4.18: −(Log-likelihood) shape for the intrinsic-like smearing term for a particular
1 pb−1 luminosity set for a σ(1/pT ) = 0 input. As the −log-likelihood function has
two minima, the error in the parameter is estimated checking the parameter value where
the −log-likelihood function increases in 0.5 units from the minimum, evaluated to the
right of the minimum (at σ(1/pT ) ∼ 0.21 TeV−1 ). This increase is reached at σ(1/pT ) =
0.36 TeV−1 , and an average error of ∆σ(1/pT ) = 0.15 TeV−1 is derived.
105
1.7
χ2 / ndf 3.015 / 4
16
1.65 Constant 14.31 ± 2.61

14 Mean 1.488 ± 0.011
1.6
σ (1/p ) [TeV-1]
Sigma 0.04948 ± 0.00924

12
1.55
10
1.5
T
8
1.45
6
1.4
4
1.35
2
1.3 0
0 10 20 30 40 50 1.2 1.3 1.4 1.5 1.6 1.7 1.8
T
0.7 22 χ2 / ndf 3.279 / 3

20 Constant 19.18 ± 3.34
0.68
18 Mean 0.6019 ± 0.0045

0.66 Sigma 0.03111± 0.00325
σ (1/p ) [TeV-1]
16
0.64
14
0.62
12
T
0.6 10
0.58 8
0.56 6
0.54 4
0.52 2
0
0 10 20 30 40 50 0.3 0.4 0.5 0.6 0.7 0.8 0.9
T
106
0.2 12 χ2 / ndf 11.28 / 10

Constant 5.442 ± 1.246
0.15 Mean 0.03001 ± 0.01947

10
0.1 Sigma 0.1028 ± 0.0250
σ (1/p ) [TeV-1]
0.05 8
0
6
T
-0.05
-0.1 4
-0.15
2
-0.2
-0.25 0
0 10 20 30 40 50 -0.3 -0.2 -0.1 0 0.1 0.2 0.3
T
set. (Right) Histogram of fitted values. The input value is σ(1/pT ) = 0.
-16641
- log-likelihood
-16641.5
-16642
0.55 0.6 0.65

σ (1/p ) [TeV-1]
T
Figure 4.22: −(Log-likelihood) shape for the additional intrinsic-like term for a particular
10 pb−1 luminosity set with a σ(1/pT ) = 0.6 TeV−1 input. The error interval defined by
the increase of the −log-likelihood in 0.5 units is [0.571, 0.642] TeV−1 , from which an
average error of ∆σ(1/pT ) = 0.035 TeV−1 is derived.
107
Table 4.4 summarizes the recovered values (mean and width of the distributions
of fitted parameters) for three different input values of an additional smearing term,
with the dependence on the muon momentum like an intrinsic resolution term, and
for the two luminosity scenarios considered. We can conclude that an overall intrinsic
resolution-like term larger than ∼ 0.12 TeV−1 , can be well recovered, even with very low
collected luminosity (Lint ∼ 1 pb−1 ). It means that an additional intrinsic resolution
contribution, larger than 15% TeV−1 , can be determined.
Table 4.4: Mean and width of the distributions of fitted parameter σ(1/pT ) (all magni-
tudes are in TeV−1 .)
Lint = 1 pb−1 Lint = 10 pb−1

σinput (1/pT ) < σfitted > (1/pT ) σ(σfitted )(1/pT ) < σfitted > (1/pT ) σ(σfitted )(1/pT )
1.5 1.5 0.129 1.49 0.049
0.6 0.6 0.093 0.602 0.031
0.0 27.7 × 10−3 0.153 0.03 0.1
4.5 Results on collision data
Once the performance of the method was established in MC, it was applied to data
already with the very first data collected. The final data sample used corresponds to
√
36 pb−1 of proton-proton collisions at s = 7 TeV (all the data recorded during 2010).
After the full selection process 12252 Z bosons are selected.
The MC used in this section corresponds to a sample generated with POWHEG (42).
When applying the method to real data, the selection slightly differs from the one
applied to the validity tests with MC to comply with the standard selection procedure
set in the context of electroweak studies in CMS. In this case, no trigger is required,
the pseudorapidity region considered is set to 2.1 and the isolation variable threshold
is set to 0.15. Several tests confirmed the independence of the results with respect to
these variations.
108
4.5.1 Misalignment scenarios on MC samples
Simulated events are reconstructed with the same programs used for real data. Special-
ized studies carried out in parallel with the physics studies, to assess the performance
of the detector, provide a set of parameters that define the working conditions of the
detector. These parameters are stored in a database and are incorporated into the
reconstruction programs. A new reconstruction of both, real data and simulated events
may be needed in case new effects are discovered or a new set of running conditions
becomes available.
For muons coming from a Z decay (medium pT range, where the resolution from the
tracker dominates) the residual misalignments of the different elements that compose
the tracker are specially relevant. These misalignment conditions are evaluated by
software and included in the database for general use in the track reconstruction. The
method consists basically of determining with a selected collection of tracks a set of
new positions of the tracker modules such that it minimizes the residuals of the tracker
hits with respect to the reconstructed track.
Even if the minimization converges and the pattern recognition performs correctly,
possible remaining misalignment could still be present, biasing the measurements of the
track parameters. In particular, those effects that keep invariant the χ2 minimized by
the alignment algorithm will not be detected by the procedure. These transformations
of the geometry are the so called “weak modes” (47).
The simplest example of a weak mode is a global translation or rotation of the
whole tracker. Any of these transformations will keep the χ2 of the track invariant,
not being spotted by the alignment algorithm. This trivial example is easily removed
by other means, but several non-trivial transformations affecting the geometry of the
tracker, could still be present after the track based alignment. Systematic biases in a
physics measurement could arise as a consequence: a systematic rotation of the layers of
the tracker would give rise to a charge-dependent momentum asymmetry in the muon
tracks. In a similar way, the radial expansion or compression or a Z-scale distortion
would change the distance scale measured, and, as a consequence, would distort any
lifetime measurements.
For the lowest-order modes, 9 systematic misalignments for a cylindrical geometry
are identified. They can be classified by the 3 different coordinates in which the shift
109
Table 4.5: Classification of weak modes by the way the tracker geometry is modified.
∆R ∆φ ∆Z
R Radial expansion Curl Telescope
φ Elliptical Clamshell Skew
Z Bowing Twist Z expansion
can take place (∆R, ∆φ and ∆Z) and the dependence of this shift (in R, φ and Z) as
can be seen in Tab. 4.5 and Fig. 4.23.
Figure 4.23: Weak modes - Graphical representation of the weak modes and their
classification by the way the tracker geometry is modified.
In Fig. 4.24 the reconstructed momentum minus the generated momentum is plot-
ted for the nine studied weak modes. With them, the global effect included in the
startup geometry (last row, right plot, which is the one applied to MC) and a reduced
offset in the transverse curvature κ = q/pT of 0.5 TeV−1 are shown. In this Fig. the
misalignments are applied over a sample of simulated W bosons, which are selected
using an optimized selection process. Several conclusions can be drawn from these
plots: none of these modes shifts the muon momentum distribution, except the curl
mode (this mode can be removed in most of the cases); all these modes make the muon
momentum distribution wider; the reduced mode induces a sinusoidal behavior in φ.
In order to make the MC resemble the data, a misalignment scenario is simulated.
110
Figure 4.24: Impact of weak modes on muon pT - Impact of the 9 studied weak modes
on the muon transverse momentum of a simulated sample of W+ bosons decaying into a
muon and a neutrino (upper three rows). The lower two plots illustrate two additional
geometry scenarios: the effect due to a reduced offset in the transverse curvature κ = q/pT
of 0.5 TeV−1 , and the actual geometry used for the simulation and reconstruction of the
data sample.
111
Due to their properties, previously described, the weak modes induced by this scenario
are not fully removed after the alignment procedure. Among the possible modes that
could still be present in the final simulated sample, the reduced one stands out. It
produces a shift with sinusoidal shape in φ. In this mode, shifts in one hemisphere of
CMS would be positive, and negative in the opposite hemisphere. Furthermore, they
are opposite in sign for both charges. Since the muon distribution must be flat in φ,
the net shift effect of this mode is null. The peak of the invariant mass distribution
remains in the same region. However, a widening of this distribution can be observed
when this mode is present, inducing a similar effect as the resolution term does.
In Fig. 4.25 the invariant mass distribution for the MC Z → µ+ µ− sample versus
the negative muon φ can be seen. A φ-dependent peak position is observed. This φ
dependent shift makes the reconstructed dimuon mass peak wider.
105
Mµµ [GeV]
100
95
90
85
80
-3 -2 -1 0 1 2 3
φ (µ)
Figure 4.25: Z invariant mass distribution versus negative muon φ - A φ depen-
dent shift is clearly observed in the Z invariant mass distribution. The reduced weak mode
shown before is present in the simulated tracker geometry.
We have then followed a two-step procedure: First, we have applied the method over
112
the MC sample (using the generator information as the reference and the reconstructed
momentum as “fake data”) to remove these weak modes already present in the simulated
samples. Afterwards, we have repeated the process but now comparing data with the
already corrected MC. In the first case, we use a sinusoidal form to parametrize the
shift in φ. After the fit we get:
δ(1/pT ) = 0.00107 − 0.278 · q · sin(φ − 1.517) [TeV−1 ]
In Fig. 4.26 the result of this fit is plotted. In dots the Z boson peak using data
is shown. In green the reference MC and in red the corrected MC are shown. After
correcting the weak mode, the MC distribution is narrower than before, as expected
following the previous reasoning. In this situation we can apply the method over the
data.
As shown in this subsection, a parametrization of the scale factor in φ becomes
necessary to remove the weak modes. In addition, and to reach a better agreement
data-MC and a better description of the effects, a dependence of the scale factor in η
is introduced. The resolution term is allowed to vary as a function of η.
4.5.2 Scale factor and resolution parametrization in η and φ
In this subsection we will derive a full parametrization of the shifts and resolution terms
as a function of η and φ. We will assume that the resolution term does not depend on
φ.
4.5.2.1 Scale factor dependence on φ and η
The scale factor will be studied first independently as a function of φ and η as a prior
step to the full parametrization in the two variables. The full phase space is divided
into several regions in φ and η and a single parameter, describing the scale factor
will be determined for each of them, and for each muon charge. Six regions in the
azimuthal angle are defined covering each of them 60 degrees in φ. In η, four regions
are identified: barrel (|η| < 0.9), overlap (0.9 < |η| < 1.2), near endcap (1.2 < |η| < 2.1)
and far endcap (|η| > 2.1) and also distinguishing between positive and negative η. The
total number of parameters in the fit are 12 (for φ) plus 16 (for η).
In Fig. 4.27 the fitted scale factor in every φ region is plotted. From this plot we can
conclude that: the shifts in φ follow a sinusoidal shape for both positive and negative
113
1800
# events / 1 GeV
1600 Data
1400 Corrected MC
Reference MC
1200
1000
800
600
400
200
0
70 75 80 85 90 95 100 105 110
Mµµ [GeV]
Figure 4.26: Corrected MC - MC distribution after the shift correction. The invariant
mass distribution in MC before the correction (green) and MC after the correction (red)
are plotted. Experimental points (black dots) are also included for comparison. Note that
before the correction the MC was wider than the data. Now it is no longer the case.
114
muons; the amplitude of the sinusoidal is similar for both charges; the phase of the
sinus for positive and negative muons differs in π. All these features already appeared
when studying the reduced weak mode. For this reason a parametrization of the form
δ(1/pT ) = constant + A · q · sin(φ + φ0 ) (4.8)
is appropriate.
δ (1/p ) [TeV-1]
positive µ
0.3 negative µ
0.2
T
0.1
-0.1
-0.2
-0.3
-0.4
-200 -100 0 100 200
φ (µ) [deg.]
Figure 4.27: Scale factor dependence on φ - Scale factor as a function of φ for positive
and negative muons where the 12 parameters of the fit are left free.
Concerning the dependence of the scale factor on the muon pseudorapidity we first
look for a possible dependence on charge. We let the scale factor in η take different
values for positive and negative muons. In Fig. 4.28 the fitted values of the scale factor
in the different regions in η are plotted. No dependence on charge is observed.
In addition, a possible asymmetry between the positive and negative parts of CMS
as a function of the z coordinate is considered. In Fig. 4.29 the variation of the scale
factor with η for the whole pseudorapidity range is presented. In this case the most
appropriate functional form to parametrize the scale factor is not obvious. Long tails
observed in high η regions (the performance of the detector degrades at high η values,
due partially to higher track density, giving rise to high residuals) prevent the result
115
δ (1/p ) [TeV-1]
0.2
0.1
T
-0.1
-0.2
positive muons
-0.3 negative muons
-0.4
0 0.5 1 1.5 2 2.5 3
η (µ)
Figure 4.28: Scale factor dependence on η - Study of the dependence of the scale
factor term on η and charge. The values of the scale factor for negative and positive muons
are similar. The scale factor parameters in each region are allowed to vary freely.
from being highly accurate in that region. Additionally, the asymmetry in the location
of the CMS shaft (a vertical tunnel built to descend the different detector components
during CMS installation in the cavern) makes one half of CMS to be better aligned
that the other one (the region where the shaft receives a bigger amount of cosmic
muons which are used for the alignment of the detector). However, a parabolic shape
independent with charge seems to be a good compromise between simplicity in the
parametrization and a good modeling of the detector behavior. A parametrization of
the form
δ(1/pT ) = constant + A · η 2 (4.9)
is chosen for the variation of the scale factor in η.
4.5.2.2 Resolution dependence in η
In this subsection we analyze the possible dependence of the resolution term on η. The
same procedure applied to the scale factor is used in this case. We distinguish 4 regions
116
δ (1/p ) [TeV-1]
0.4
positive µ
0.2 negative µ
T
0
-0.2
-0.4
-0.6
-0.8
-2 -1 0 1 2
η (µ)
Figure 4.29: Scale factor dependence on η - Scale factor form as a function η for
positive and negative muons where the 16 parameters of the fit are free.
in η as before: barrel, overlap region, near endcap and far endcap. By symmetry
reasons, we will assume that the resolution term does not depend on φ.
In Fig. 4.30 the variation of the resolution with η for the whole CMS is presented.
The behavior is symmetric in η. A parabolic shape seems again to be a good choice.
We use the form:
σ(1/pT ) = constant + A · η 2 (4.10)
for the resolution term.
4.5.3 Scale factor and resolution results after the parametrization
As it has been presented before, the parameters δ(1/pT ) and σ(1/pT ) that control
momentum scale and resolution effects, depend on the muon charge, its azimuthal angle
φ and its pseudo-rapidity η. Once we have defined what is the general dependence of
these effects on charge, φ and η we will determine the precise value of these effects.
The basis of the method is to modify the reconstructed transverse momentum, pT ,
according to the following expression:
1 1
= + δ(q, φ, η) + σ(q, φ, η) Gauss(0, 1) (4.11)
pT pT,sim
117
σ (1/pt) [TeV -1] 0.01
0.0095
0.009
-2 -1 0 1 2
η (µ)
Figure 4.30: Resolution in η - Dependence of the resolution term on η for the whole
pseudorapidity range. A symmetric behavior is observed in η. A parabolic function seems
to be an appropriate description of the resolution in η.
where pT,sim is the simulated value of the muon pT , Gauss(0, 1) denotes a sampling
following a Gaussian function of zero mean and unit variance and δ and σ the param-
eters controlling the additional effects we want to model. This ansatz is adequate for
cases presenting slightly worse resolutions in data than in simulation, and therefore
well suited for the present study. According to the results of the previous sections, we
can assume angular dependences of the type:
δ(q, φ, η) = A + Bη 2 + q C sin(φ − φ0 ) (4.12)
σ(q, φ, η) = A′ + B ′ η 2 (4.13)
where A, φ0 , B, A′ and B ′ are the final parameters to be determined.

The result is shown in Fig. 4.31. An excellent agreement for the reconstructed Z
invariant mass distribution between data and corrected simulation is reached.
The resulting shifts δ as a function of φ and η are shown in Fig. 4.32. Although
both data and simulation have an azimuthal bias the phases and amplitudes of the two
are significantly different. There is also a significant difference in the η biases where
118
data shows a bigger effect. Fig. 4.33 shows the transverse momentum resolution versus
η after the biases are corrected.
The resolution is defined as:
1/pT,rec − 1/pT,gen
σ=
1/pT,gen
The resolution in the reference MC (STARTUP MC) is shown in black dots and the
values from the MC, once it is adjusted to reproduce the data, is shown as red squares.
A general good agreement is observed. We can also appreciate that the reference MC
reproduces well the behavior observed in data for the barrel region (η < 1). However,
in the endcap region, the reference MC overestimates the resolution, and in fact, data
exhibit better muon pT resolution.
1800
1600
1400 Data
# events / 1 GeV
Corrected MC
1200
Reference MC
1000
800
600
400
200
0
70 75 80 85 90 95 100 105 110
Mµµ [GeV]
2
χ
0
-2
70 80 90 100 110
Figure 4.31: Reconstructed Z invariant mass distributions in data and Monte Carlo. The
original Monte Carlo prediction (green line) has been corrected (red line) by the parameters
determined in the fit.
The fitted values of the parameters are shown in Tab. 4.6.
119
0.12
δ (1/p ) [TeV-1]
δ (1/p ) [TeV-1]
0.3
0.1
0.2
T
T
0.08
0.1 0.06
positive µ
negative µ 0.04
0
0.02
-0.1
0
-0.2
-0.02
-0.3 -0.04
-3 -2 -1 0 1 2 3 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2
φ (µ) [rad.] η (µ)
Figure 4.32: Scale shifts obtained as a function of the azimuthal angle for positive and
negative charged muons (left) and of the pseudo-rapidity (right).
0.06
0.05
STARTUP MC
0.04
Corrected MC
T
σ(p )/p
0.03
T
0.02
0.01
0
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
η (µ)
Figure 4.33: Transverse momentum resolution as a function of the pseudo-rapidity of
the muon. The black dots represent the resolution from the reference MC and in red the
resolution once the MC has been corrected with additional scale and resolution terms.
120
Table 4.6: Fitted parameters that control muon momentum scale and resolution effects.
Effect Values
A = 0.016 ± 0.019 TeV−1
B = 0.005 ± 0.007 TeV−1
δ(q, φ, η) = A + Bη 2 + q C sin(φ − φ0 )
C = 0.282 ± 0.01 TeV−1
φ0 = −0.134 ± 0.03
A′ = 0.236 ± 0.01 TeV−1
σ(q, φ, η) = A′ + B ′ η 2
B ′ = −0.135 ± 0.005 TeV−1
4.5.4 Comparison with MuScleFit method
The results obtained in the previous section are now compared with those obtained
with an independent method (the MuScleFit method), also available in CMS (48).
The MuScleFit method is an alternative method to compute the resolution and
scale factors in data. It is based on a multidimensional likelihood fit of a functional
form of the Z boson peak given by MC at generator level.
The approach in this method can be summarized in the following steps: ansatz
functions are used to describe the muon pT for scale and resolution, of the form: pT =
F (xi ; aj ) · pT and σi (xi ) = Gi (xm ; aj ), with xi = η, φ, pT and aj being the parameters
to be computed. A Lorentzian convoluted with a gaussian is the ansatz function used
to model the Z boson mass peak. The background is described with a sum of linear and
exponential functions. This ansatz function is used to describe the Z boson mass peak.
The likelihood of these ansatz functions to reproduce the reconstructed dimuon mas
distribution is computed. Note that for this method the resolution and scale factor
terms are computed as a whole, and not as extra terms to be added to the already
simulated one (as SIDRA does).
A comparison of the result from both methods when applied to data for the scale
factor term in φ, in terms of the δ(1pT ) extra scale factor, is shown in Fig. 4.34. A
good agreement is observed between both methods. In η the agreement between both
methods is also remarkable (see Fig. 4.35).
In Fig. 4.36 a comparison of the results for the resolution term with SIDRA and
MuScleFit. Differences from both methods are within statistical errors.
121
This approach is shown to perform accurately to get the resolution and bias terms.
However the method proposed in this thesis presents several advantages: using the
simulation as starting point avoids ad-hoc parametrizations of the reconstructed mass
at generation level, which are necessarily cumbersome due to the presence of analysis
cuts and other theoretical effects (QED final state radiation, PDFs, . . . ); Monte Carlo
generators like PYTHIA (49), POWHEG, etc. are already precise enough in the pre-
dictions for the Z invariant mass shape, particularly around the Z pole, and this shape
is minimally modified by higher order QCD and QED effects; the simulation already
includes inhomogeneities and tails in the response of the detector, which are hard to
reproduce with a fully analytical ansatz. In addition, it gives directly the distortion
to be incorporated in the MC to fit the data. From a practical point of view this last
feature makes this method specially appealing for physics analysis.
0.3
0.2
δ(1/p ) [TeV-1]
0.1
negative µ (MuScleFit)
negative µ (SIDRA)
0 positive µ (MuScleFit)
T
positive µ (SIDRA)
-0.1
-0.2
-0.3
-3 -2 -1 0 1 2 3
φ (µ) [rad]
Figure 4.34: Sidra-MuScleFit comparison for the scale factor dependence on φ.
122
0.14
0.12
0.1 MuScleFit
δ (1/p ) [TeV-1]
0.08 SIDRA
0.06
T
0.04
0.02
0
-0.02
-0.04
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
η (µ)
Figure 4.35: Sidra-MuScleFit comparison for the scale factor dependence on η.
0.09
0.08
resolution from MuScleFit on MC
0.07
resolution from MuScleFit on Data
0.06
T
resolution from SIDRA on Data

σ(p )/p
0.05
T
resolution from SIDRA on MC

0.04
0.03
0.02
0.01
0
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
η (µ)
Figure 4.36: Sidra-MuScleFit comparison for the resolution term dependence on η.
123
4.6 Theoretical uncertainties and biases in the determi-

nation of muon resolution parameters
The method described in this Chapter is based on the assumption that the modeling
of the Z → µ+ µ− invariant mass distribution at the generator level is largely inde-
pendent on theoretical assumptions. In this Section we discuss how much theoretical
uncertainties affect the description of the Z mass peak and therefore propagate to the
muon momentum scale and resolution determination.
We will consider PDF uncertainties, ISR and FSR uncertainties, and weak and
QED interference effects.
We will obtain an estimate of: a) how much the average value of the invariant mass
shifts and b) how much the width of the invariant mass changes. The results from a)
can be interpreted as a systematic uncertainty on a scale shift at the Z peak, whereas
results from b) can be interpreted as systematic uncertainties on the smearing at the
Z peak. In order to ensure that results are meaningful, we will determine these shifts
and widths on a simulated and reconstructed Z → µ+ µ− sample with a reconstructed
dimuon invariant mass cut of Mµµ > 60 GeV.
Our Monte Carlo sample consists of 55003 Z → µ+ µ− selected events. They have
been generated with POWHEG (42), which should provide a much more reliable de-
scription of the process than PYTHIA, particularly for the transverse momentum dis-
tribution of the Z boson, as discussed below. All systematic effects are collected in
Table 4.7. From the technical point of view the values of the associated uncertainties
will be evaluated with a reweighting procedure. Every element is given a weight de-
pending on the ratio of the reconstructed distribution once one of the parameters of
the generation is modified according to its uncertainty, and the original distribution.
4.6.1 Parton density function (PDF) uncertainties
PDF uncertainties can obviously modify the relative fraction of on-peak and off-peak
events, and therefore affect the determined parameters. The logic employed to as-
sign uncertainties is the one reported in (50). For the CTEQ6M NLO PDFs (51)
used in the POWHEG generation, we observe a possible scale shift uncertainty of
+37±9
∆ =−32±5 MeV. Using MSTW2008 NLO PDFs (45) instead, we do not observe any
124
4.6 Theoretical uncertainties and biases in the determination of muon
resolution parameters
change in the central values with respect to CTEQ6M, while the shift uncertainty is
+26±16
reduced to ∆ =−20±15 MeV, i.e. a negligible relative shift of 0.03% at the Z peak.
+268±3
The change in the observed width of the resonance is however larger: σ =−215±3 MeV
+193±11
with CTEQ6M and σ =−131±11 MeV with MSTW2008, i.e. a relative smearing of
0.2 − 0.3% at the Z peak, to be compared with the typical ideal resolutions of order
1 − 2%. The CTEQ6M numbers are reported in Table 4.7. One has to notice that the
MC sample used in the 2010 analysis was simulated with the CT10 PDF set, so the
CTEQ6M numbers are very conservative upper values.
4.6.2 Initial State Radiation (ISR) uncertainties and higher order

QCD effects
ISR uncertainties are mostly affecting the pT distribution of the Z boson, and induce
an effect very similar to a smearing of the muon pT . It is evaluated by reweighting
the PYTHIA Z boson pT spectrum obtained with PARP(64)=0.2 (default) with the
PYTHIA Z boson pT spectrum using PARP(64)=0.1, which is known to give a better
agreement to the predictions from programs like RESBOS (52) or POWHEG itself. This
leads to a scale shift of ∆ = 62 ± 41 MeV, and an extra smearing of σ = 293 ± 29 MeV.
Note however that, since we are using POWHEG, this extra smearing is largely
overestimated, and we should expect values that are much smaller than the original
estimate, i.e. something probably below 100 MeV for σ. Nevertheless, and in order to
cover possible higher order QCD effects, ISR numbers reported in Table 4.7 correspond
to the most pessimistic case in which we fully rely on a classic PYTHIA generation.
4.6.3 Weak and QED interference effects
Weak corrections are estimated by using a similar reweighting strategy. The essential
effect at the Z peak with respect to the Born-improved approach employed in PYTHIA
is an increase of the pure Z exchange diagram by 0.5% relative to the photonic one.
The result is a negligible additional scale shift and a decrease of the Z width of σ =
−18 ± 1 MeV. QED interference effects are totally negligible at the Z resonance, and
have just some minimal effect at the edge of the invariant mass region (< 70 GeV),
where the sensitivity of the method is almost negligible, since shifts and degradations
of resolution predominantly manifest very close to the maximum of the Z peak.
125
4.6.4 Final State Radiation (FSR) uncertainties
FSR QED effects significantly distort the invariant mass shape with respect to the orig-
inal generator distribution in the absence of QED radiation. However, the dominant
effects are reliably reproduced by the parton-shower approach used and PYTHIA (and
therefore in POWHEG), since it must correctly reproduce leading-order collinear ef-
fects. Possible missing effects are of order O(α/π), i.e. 0.2 − 0.3% and mostly affecting
non-collinear directions with respect to the final lepton. We have tried to estimate
missing effects of this order and higher orders by changing the scale of the QED radi-
ation from Q2 = 0 to Q2 = p2T , where pT is the relative photon pT with respect to the
final lepton. This change induces negligible effects: a scale shift of ∆ = −12 ± 2 MeV
and an extra smearing of σ = 8 ± 2 MeV.
4.6.5 Total systematics
All different sources of systematics and their (rounded) sum in quadrature are compiled
in Table 4.7. Overall, theoretical uncertainties can at most produce scale shifts below
0.1% and extra smearings below 0.5%, relative to the value of the Z mass (91.2 GeV).
A relative shift of 0.1% in the mass induces a fake momentum shift of order 0.1% ·
pT /[91 GeV] if the effect is assumed to equally affect positive and negative muons. An
√
extra smearing of 0.5% can be interpreted as an extra smearing of 0.5%/ 2 for muons
with pT ∼ MZ /2 GeV, which is smaller or of the order of the ideal tracker resolution in
this range.
Note in any case that all these sources of uncertainty will equally affect any other
method trying to estimate resolution parameters, since in general they will be hardly
distinguishable from authentic resolution effects.
126
4.6 Theoretical uncertainties and biases in the determination of muon
resolution parameters
Source ∆[MeV] σ[MeV]

PDFs 37 ± 9 268 ± 3
ISR and NLO effects 62 ± 41 293 ± 29
WEAK effects 0 −18 ± 1
FSR effects −12 ± 2 8±2
TOTAL 73 ± 37 398 ± 22
Table 4.7: Summary of systematic uncertainties considered for resolution parameters. ∆

denotes a systematic scale shift, whereas σ denotes and additional increase (or decrease
if sign is negative) in the observed width of the dimuon invariant mass. Quoted PDF
uncertainties correspond to the size of the maximum observed deviations in the CTEQ6M
case, irrespective of the sign (CTEQ uncertainties are asymmetric). The total uncertainty
corresponds to a rounded sum in quadrature of all effects.
127
128
5
Measurement of the inclusive W

boson production cross section in
the muon channel
W and Z bosons are copiously produced at LHC proton-proton collisions, being one
of the dominant process (53). Although being particles widely studied in past experi-
ments, their importance in the initial phases of LHC operations is clear. When decaying
to leptons, they are relevant processes for several reasons: it is a benchmark for lepton
reconstruction and identification necessary for other analysis, a precision test of per-
turbative QCD and the parton distribution functions (PDFs) of the proton, a possible
estimator of the integrated luminosity at LHC collisions and a dominant background
for many interesting physical searches.
Lepton reconstruction and identification is capital for many analyses. Since all
analyses based on muons are clean they are the golden channel of many searches.
A correct muon reconstruction and identification is, as a consequence, of paramount
importance. They are also used to test and calibrate the detector response in terms of
trigger, efficiencies and resolution. The measurement of muon momentum resolution
presented in the previous Chapter computed using the Z boson resonance is an example
of how the study of electroweak processes can influence the understanding of results in
other analyses.
Several electroweak channels give direct access to the proton PDFs. Predictions
from different PDFs are sometimes in slight disagreement (54), or their knowledge
129
5. MEASUREMENT OF THE INCLUSIVE W BOSON PRODUCTION
CROSS SECTION IN THE MUON CHANNEL
is insufficient. A better understanding of certain electroweak processes could help to

solve these contradictions and shed light on the quark composition of protons. As an
example, the W + c channel that is described in Chapter 6 gives direct access to the
strange quark content. The strange quark composition in protons is not well understood
and the different PDF groups would benefit for such a study. It will also have influence
on important measurements as the W boson mass (55), decreasing the systematic error
associated to such a measurement.
The total luminosity delivered by the accelerator is the dominant uncertainty of
many analyses (of the order of 4%). The number of W or Z reconstructed bosons is
a reliable and stable estimator of the delivered luminosity that could be used in the
future.
Electroweak processes can help to understand the signal of important physical
searches (for example the Higgs decay in two W bosons) and are background of many
others (being even the dominant background in certain cases). A better understanding
of electroweak processes allows us to improve our simulations and as a consequence to
tune the analyses based on these channels. Take for instance the case of the W’→ eν
search shown if Fig. 5.1. As it can be seen in this figure, the dominant background
corresponds to electroweak processes (W+jets).
CMS deploys a variety of different studies centered on the production of electroweak
bosons in proton-proton collisions. Several electroweak studies are already published
(or accepted for publication). These studies are classified:
1. Production of W and Z bosons in pp collisions.
2. Diboson production cross section and constraints on anomalous triple (and quar-
tic) gauge couplings
3. Measurement of electroweak parameters: sin2 θW and the W mass
The measurement of the inclusive W and Z boson cross section and cross
sections ratios was the first electroweak measurement performed in CMS (57). It
serves as a test of the Standard Model at this new energy. The measurement of the W
boson production cross section in the muonic channel is the subject of this Chapter.
W charge asymmetries are also studied (58). W+ are created via a ud̄ current
interaction, while in the case of W− we need a d quark in the initial state (dū current
130
Figure 5.1: - Transverse mass distribution for the W’ search in CMS. The W → eν
process is the dominant background in this search (56).
interaction). The relative content of u quarks in the proton respect to d quarks is

2:1, so a W charge asymmetry of around 2 is expected in LHC collisions at 7 TeV.
Precision measurement on the W charge asymmetry as a function of the boson rapidity
can provide new insights on the u/d ratio and sea antiquark densities in the ranges of
the Bjorken parameter x (59) explored at the LHC.
The differential cross-sections of Drell-Yan pairs measurement (60) in mass
also helps to better understand the parton PDF. This process is theoretically well
understood and differential cross sections predictions at the level of NNLO with respect
to the inclusive production are available. Differential cross section measurements are
thus sensitive to variations due to the imperfect knowledge of the parton PDFs, due to
uncertainties in αS , and the choice of QCD renormalization and factorization scales.
The study of the polarization of W and Z bosons produced with high trans-
verse momentum (61) provides a stringent test of perturbative QCD calculations.
It is also an important prerequisite to searches for new physics. In this case we look for
W and Z boson produced with hard jets. In this situation the W boson will exhibit a
high transverse momentum. Due to the V - A nature of the coupling of the W boson
to fermions and the enhancement of the gluon-quark contribution to the W+jet pro-
131
duction with respect to Tevatron collisions we expect W boson with high pT to exhibit
a sizable transverse polarization.
The measurement of the production of jets in association with W and Z
bosons (62) provides a stringent and important test of perturbative QCD calculations.
Predictions at the Next-to-leading order (NLO) level respect to the inclusive production
are available for V + n jets, with n up to four for the W and three for the Z (63, 64, 65,
66). The precision of the prediction varies from 10% up to 30% due to uncertainties
on parton distribution functions and on the perturbative nature of the calculations.
This study is specially relevant since, as shown in the introduction, they constitute an
important background in searches for new physics and for studies of the top quark. As
a consequence, precise measurement of the cross section and an understanding of the
jet and lepton kinematics is essential.
The production of heavy quark jets in association with W and Z bosons is
the subject of other interesting studies at CMS (67, 68). The production of the Z boson
in association with b-quarks is an important measurement at the LHC. It represents a
benchmark channel to the production of the Higgs boson in association with b-quarks.
At the same time it is an important Standard Model background to many new physics
searches and Higgs in final states with leptons and b-jets. The production of W bosons
in association with a c-quark gives direct access to the s-quark content of the protons.
A better understanding of this process can improve the knowledge of the proton PDFs
on the s-quark. This study is the subject of next Chapter.
Electroweak boson production with tau leptons in the final state are an
important probe for many new physics processes in proton-proton collisions at the
LHC. Both electroweak boson decays into taus (Z → τ + τ − and W → τ ν) are studied
within the CMS collaboration (69, 70). Tau decays can be purely leptonic (in electrons,
muons and neutrinos) or semileptonic (neutrino plus hadronic system). Both possible
decays are studied by the CMS electroweak group. Tau channels can be signal of many
interesting studies. The study of the decay in the semileptonic channel is essential for
searches for new physics based on tau leptons. Tau leptons can also be an important
feature in signals for supersymmetry and extra dimensions, and in searches for extra
gauge bosons. Taus are also an important background of relevant processes. As an
example, the W → τ ν production is the major background in the search for the charged
132
Higgs boson in the τ ν final state. The W → τ ν production has to be well understood
as a test of the Standard Model.
Another electroweak study in CMS is the forward-backward asymmetry of
Drell-Yan pairs (71). This study is interesting for a host of reasons. The pro-
cess qq̄ → Zγ ∗ → l+ l− involves both vector and axial-vector couplings of electroweak
bosons to fermions, resulting in a forward-backward asymmetry in the number of lep-
ton pairs. This asymmetry depends on the dilepton invariant mass, quark flavour and
the electroweak mixing angle. So deviations from the Standard Model prediction for
the asymmetry may indicate the existence of a new neutral gauge boson, quark-lepton
compositeness, supersymmetric particles, or extra dimensions. This measurement can
also contribute to improve QCD measurements with high order corrections and to con-
straint parton PDFs. The electroweak mixing angle can also be measured using this
asymmetry. Finally, the asymmetry as a function of the mass also provides information
on the u and d quarks separately.
Diboson processes are also studied at CMS. The self-interaction between gauge
bosons occurs via the trilinear gauge boson couplings (TGCs) and it is well understood
within the Standard Model. Boson self-interactions are a consequence of the non-
abelian SU (2)×SU (1) group which describes the electroweak force within the Standard
Model. Their couplings are predicted by the gauge structure of the Lagrangian. The
high sensitivity of the production of Zγ, Wγ, WW, WZ and ZZ bosons in proton-
proton collisions to these TGCs makes these channels to be of special interest to test
the Standard Model (72, 73). Any deviation observed from the SM prediction would
be a hint of new physics (new particles decaying to a diboson state or a new interaction
that would modify the strength of the couplings).
The sin2 θW is the only free parameter in the Standard Model that fixes the relative
couplings of all fermions to the neutral weak bosons (74). The weak-mixing angle
θW describes the rotation that transform the original W0 and B0 vector bosons states
into the observed γ and Z bosons. This measurement is of special interest to test
the universality of the fermion-gauge boson couplings and predictions of the Standard
Model. Precision measurements at LHC aim to reach the level of 1% accuracy in this
parameter.
The measurement of the mass of the W boson presents several technical difficul-
ties that can’t be addressed until the integrated luminosity recorded in CMS is of the
133
order of 10 fb−1 . The mass of the W boson is known at the level of 20 MeV by the
combination of measurements in different experiments. With the current precision, the
combined measurement top mass-W boson mass is one sigma away from the Standard
Model prediction.
5.1 W and Z boson production in CMS
At high energy, a proton-proton collision can be interpreted at the hard scattering level
as a collision of either quark-quark, quark-antiquark, gluon-quark or gluon-antiquark
partons. The higher the energy of the primary protons, the bigger the radiation and
hence the influence of gluons and sea quarks. At the LHC energy, W and Z bosons
will be produced by interaction of all flavour of quarks and antiquarks. Z bosons are
dominantly produced via u ū or d d̄ annihilation. For the W production, the dominant
processes are ud̄ → W+ and dū → W− . The LHC is a W and Z factory, with over 100
bosons produced per second in nominal conditions.
√
The total cross-section of a certain process pp → X at a center of mass energy s
can be written (75):
XZ Z 1Z 1
σ= σ̂ij (ŝ, µf , µr ) fi (x1 , µ2f )fj (x2 , µ2f )δ(ŝ − x1 x2 s)dx1 dx2 dŝ (5.1)
i,j 0 0
√
where σ̂ij are the parton-parton cross-section i + j → X, ŝ the reduced center-of-
mass energy of the parton reaction, x1 the energy fraction of parton i, x2 the energy
fraction of parton j, fi (x) the parton distribution function for parton i (same for j),
µf the factorization scale and µr the renormalization scale. The partons i and j may
be the valence quarks of the proton (u,d), the gluons interchanged by the quarks (g)
or any of the other sea quarks present in the proton (ū,d̄,s,c,b,t,s̄,c̄,b̄,t̄). The delta
function considers the energy conservation of the process.
In a very simplified case of the production of a narrow resonance, the previous
cross-section can be written as
σ̂ij (ŝ) = σij δ(ŝ − M 2 )M 2 (5.2)
with M being the mass of the resonance and σij the constant cross-section of the reaction
i + j → X at the peak of the resonance. Once applying this simplification, the total
134
5.2 W and Z boson decays
cross-section of the process pp → X can be written as

X
σ= σij M 2 × Lij (5.3)
i,j
being Lij the parton-parton luminosity

1
1 1
Z
Lij = fi (x)fj (M 2 /xs)dx (5.4)
s M2 x
s
This simplified picture can be improved by taking W and Z width effects into
account, as well as QCD and QED radiative corrections.
√
Collision data were recorded at center-of-mass energies of s 7 TeV and 2.76 TeV.
W bosons are copiously produced in LHC events due to the high energy available in
the collisions. With NNLO cross-sections (including branching fractions to leptons) of
18456 pb for W+ → l+ νl and 12858 pb for W− → l− νl (where l considers the whole
contribution to leptons e, µ or τ ) at 7 TeV, the production of W boson decaying into
leptons are one of the most important processes that occur at LHC collisions. The
production of Z bosons is about 3 times smaller than that of the W bosons.
5.2 W and Z boson decays
W and Z bosons are massive particles that decay rapidly (half-life is of the order of
10−25 s). They can only be detected through their final states. Their decays into lepton
and quarks are accurately predicted by the gauge theory of electroweak interactions and
strong forces.
W bosons can decay either to leptons or quarks. Their leptonic decay can be either
W± → e± νe , W± → µ± νµ or W± → τ ± ντ via the weak force. W bosons can also decay
into quarks following the scheme W+ → ud̄ or W− → dū where the d quark can be
replaced by any d-quark type (s or b quarks). W bosons can not decay into t quarks
since the t quark mass is bigger than the W boson mass.
Theoretical predictions for the results shown in the following Chapters are computed
at NNLO using FEWZ (Fully Exclusive W and Z production) (63, 76) and the CT10 set
of PDFs. With uncertainties of the order of 2% to be reached at LHC, NLO predictions
in the strong coupling (order of 10% accurate) are insufficient for a precise comparison
with data. A NNLO is necessary in this case.
135
W boson reconstruction using its decay in quark-antiquark pair, even if it is domi-

nant with respect to the leptonic channels, is humped by the high background coming
from QCD interactions. Within the leptonic channels, the muon channel is specially
clean. This is the reason why the analysis presented in this thesis are based on the
reconstruction of W bosons when decaying into muons.
5.3 W signature in the muon channel
W bosons are produced at LHC almost at rest, with a typical pT of the order of 4-5
GeV. Therefore, muon and neutrino from its decay will be produced back-to-back in
the transverse plane, and the transverse momentum of the muon and neutrino will be
relatively high, peaking typically at one half of the W boson mass. The neutrino will not
be detected but its presence can be inferred as an apparent lack of energy-momentum
balance in the event. This energy-momentum balance can only be evaluated in the
transverse plane as most of the particles produced in the collision escape through the
beam pipe. By the same reason, the invariant mass of the muon-neutrino system can
not be computed since the longitudinal momentum of the neutrino can not be measured,
but a partial reconstruction in the transverse plane can be performed, making use of
the neutrino and muon variables in this plane.
In Fig. 5.2 a typical W → µν event as recorded by the CMS detector is shown.
The characteristics pointed out before are clearly seen in this Fig.: an isolated, high pT
muon and essentially nothing else. The green arrow in the event display indicates the
direction of the missing transverse energy (MET) obtained as the negative vector sum
of the transverse momentum of all particles in the event. As it can be appreciated in
the figure, the direction of the muon and the MET are almost back-to-back.
Several other physical processes occurring on high energy proton-proton collisions
also produce events with a similar signature: a big amount of MET and a high pT muon.
The main one are QCD processes. Other electroweak processes like W → τ ν with the τ
promptly decaying into a muon will have a similar signature, although with a softer pT
spectrum. Also Z boson production with the boson decaying into a muon pair should
be considered. If one of the muons from the Z boson decay escapes detection either
because it lies beyond the active region of the detector (|η| > 2.4) or due to detector
inefficiencies, the event will look like a typical W boson event. Other processes like top
136
Figure 5.2: W → µν event - W candidate as detected with CMS detector. The presence
of a W boson is inferred by the detection of a high pT muon and a big amount of MET
production (tt̄ or single top) and diboson production (WW, WZ, ZZ) also give rise to
events with a similar signature. Compared to the inclusive W production, their cross
section is rather small and will not contribute experimentally in a significant amount.
However, for the W+jet studies with smaller production cross section, top events (like
tt̄ and single top events) will be an important background. The top, decaying into a
W plus a b quark, has a similar signature to that of the signal.
In Fig. 5.3 the pT distribution of the muon of a W − → µ− ν simulated sample by
MC is shown. In the same figure, the pT distribution of muons coming from QCD
simulated processes is also plotted. Muons from W bosons range from 20 to 60 GeV in
transverse momentum, with a peak at ≈ MW /2. QCD events instead, have a muon pT
spectrum peaking at low values, pT . 20 GeV. Therefore, setting a threshold at about
20-30 GeV will allow to discriminate between events originated in these two processes.
In Fig.5.4 the isolation variable used in the analysis is plotted for the W → µν sample
and for the QCD events. This variable is defined as the quantity of energy in a cone of
0.3 in ∆R around the muon direction over the pT of the reconstructed muon,
X
rel
Icomb = (ET (ECAL) + ET (HCAL) + pT (tracks))/pT (µ)
with ∆R defined as
p
∆R = ∆η 2 + ∆φ2
137
0.2
Arbitrary units / 1 GeV

0.15
W → µν
QCD
0.1
0.05
0
0 20 40 60 80
muon p [GeV]
T
Figure 5.3: pT distribution - Muon pT distribution for the W → µν channel and QCD
processes. The sharp fall at low muon pT for the QCD sample is due to the generated
phase space (pµT > 15 GeV). The decrease for the signal (at ≈ 10 GeV) is a consequence of
the trigger requirement. The distributions are after the selection detailed in Section 5.4.3.
Arbitrary units
0.6
0.4
W → µν
QCD
0.2
0
0 0.1 0.2 0.3 0.4 0.5
Iso variable
Figure 5.4: Isolation variable distribution - Isolation variable of the muon defined as
the quantity of energy (computed using information from the tracker and the calorimeters)
in a cone of 0.3 in ∆R around the muon direction over the pT of the reconstructed muon,
for QCD and W → µν signal. The distributions are after the selection detailed in Section
5.4.3.
138
As seen in this Fig.5.4 muons from W boson exhibit a low value of the isolation variable.
Muons from QCD events show a more even distribution, spanning a large range of
isolation variable. An isolation value of 0.1-0.15 clearly divides a signal region (isolation
< 0.1-0.15) and a QCD region (isolation > 0.1-0.15).
The MT distribution is defined as:
q
MT = 2pT (µ)MET(1 − cos(∆φµMET )) (5.5)
Fig. 5.5 shows the MT distribution for W → µν and QCD events as expected from
MC. The signal is found at high MT values (bigger than 40 GeV). QCD events are
concentrated at a lower value of MT .
0.05
Arbitrary units
0.04 W → µν
QCD
0.03
0.02
0.01
0
0 20 40 60 80
MT [GeV]
Figure 5.5: MT distribution - MT distribution of the muon-MET system for W → µν

and QCD events. A requirement of 50 GeV on the transverse mass may be applied to get
a pure sample of W bosons. The distributions are after the selection detailed in Section
5.4.3.
In Fig. 5.6 the acoplanarity of the muon-neutrino system defined as:
Acoplanarity = π − |φ1 − φ2 | (5.6)
is plotted for the signal and the QCD background. Signal shows a distribution that
corresponds to a system muon-neutrino preferentially back-to-back. QCD is flatter.
In the following sections we will address the precise selection criteria established to
extract a clean sample of W boson candidates, based on the variables just presented.
139
Arbitrary units
0.3
0.25 W→µν
0.2 QCD
0.15
0.1
0.05
0
0 0.5 1 1.5 2 2.5 3
Acoplanarity
Figure 5.6: Acoplanarity distribution for the signal (yellow) and the QCD contribution
(purple) in arbitrary units. Signal events are distributed preferentially at low acoplanarity.
5.4 Inclusive W boson production cross section
CMS published a first measurement (57) of the W inclusive cross section with the first
data collected in 2010, corresponding to an integrated luminosity of 3 pb−1 . The preci-
sion of the measurement was already better than 5% (excluding luminosity uncertain-
ties) being dominated by systematic effects (4%) versus the statistical uncertainty (1%).
With the statistics available at the end of 2010 run (10 times bigger) new techniques
can be envisaged to reduce the uncertainty associated to some of the ingredients of the
measurement and make a high precision determination. In this Section, the strategies
followed for a high precision measurement and the results with the data available at the
end of 2010 run are shown. The ratio W+ over W− is also computed. Many systematic
uncertainties are cancelled in the ratio (e.g. the luminosity uncertainty that amounts
to some 4% is completly cancelled when computing the ratios), thus the ratio will allow
a strict test of the SM prediction.
The number of events produced through a certain physical process observed in the
detector after a collision is related to the luminosity provided by the machine and the
cross section of the process as:
N = σLAǫ (5.7)
where σ is the process cross section, L the integrated luminosity, A the detector ac-
140
ceptance (fraction of events actually observed in the detectable phase space) and ǫ the
detector efficiency (number of candidates after the selection process over the number
of total events in the detectable phase space). This expression will also allow us to
measure the production cross section of any process by identifying and counting the
number of events of that kind registered once we know the luminosity, the acceptance
and efficiencies.
What we measure is the total number of events that are reconstructed as W can-
didates. This sample contains, in general, events coming from other processes with a
similar signature as the signal, that is, background events. The acceptance takes into
account the fraction of events that escape detection because their final product lies
beyond the detector sensitive region. Events that are not reconstructed due to detec-
tor inefficiencies or due to the selection requirements are considered with the efficiency
term. The cross section is measured as
(Nsignal − Nbackground )
σ= . (5.8)
AǫL
The signal we are interested in consists of W bosons when they decay on the muonic
channel. Thus the signal signature, as explained before, consists of a high pT , isolated
muon plus a high imbalance of the energy measured in the transverse plane. However,
events produced by other processes will show a similar signature. They constitute a
background for the measurement. As we already introduced in the previous section,
we will rely on the signal characteristics to maximize the discrimination between signal
and to evaluate the remaining background in the final sample. An effort has been made
to minimize reliance on MC modeling, trying to develop methods based on data control
sample to control the background, at least the most critical ones. QCD processes with
a muon in the final state represent the dominant background as they are copiously
produced in proton-proton collisions. Several techniques to reduce and control this
background are shown in this Chapter. Events produced by other electroweak processes
also present a similar signature: W → τ ν and Z → µµ are the most important ones.
Additionally, minor contributions to our background are the diboson production (like
WW or WZ events) and tt̄ production. To evaluate their contribution we will rely
on the MC simulation for the shape of key distributions as well as the cross section
prediction.
141
In the following sections we will show the data and MC used in the study, the
event selection to optimize the W cross section measurement and we will cover the
measurement of the different terms contributing to the production cross section.
5.4.1 Data and Monte Carlo samples
In this analysis the full 2010 data set, 36 pb−1 of collision data are used. They were
collected on-line by the lowest non-prescaled single muon trigger. In order to keep
constant the bandwidth of the trigger selection, the trigger threshold was raised every
time the instant luminosity increase significantly. At the beginning of data taking, the
single muon trigger accepted events with a muon with a pT bigger than 9 GeV. This
threshold was increased and in the end of 2010 run a trigger on muons with pT bigger
than 15 GeV was implemented. In Tab. 5.4.1 the different triggers with their threshold
depending on the period of data taking is shown.
Sample Run range Int. Lumi pb−1 Trigger Path

Run A 136033-144114 3.07 HLT Mu9
Run B 146239-148058 4.41 HLT Mu9
Run B 148059-149442 28.61 HLT Mu15
Run A+B 136033-149442 36.09
Table 5.1: Trigger path used for the different periods of data taking.
Once the event is triggered it is centrally reconstructed, and streams of different

objects are created to easier the physical analysis. This streams are devoted to a certain
physical object (i.e. muons) and are stored in different computing sites, available for
the physical analysis.
Even if the information from data is maximally exploited in all the analyses steps,
large samples of events simulated with MC are also used in the analysis. They are
employed in several steps of the analysis, for cross checks and to asses the validity of
the methods deployed.
Simulated samples of signal events and from other electroweak processes (Z bo-
son production and decay, and other W boson decays) are produced using a NLO
MC generator (POWHEG) (42, 43) interfaced with PYTHIA (49) for showering and
hadronization. tt̄ events are generated with PYTHIA. QCD processes with muons in
142
the final state are also simulated. For both signal and the background several mini-
mum bias events are superimposed to the hard interaction to simulate the effect of the
“pile-up” of several proton-proton collisions occurring at the same bunch crossing.
In Tab. 5.2, a detailed table with the MC samples used in the analysis is shown.
Generator Process Kinematic cuts (in GeV, c = 1) σ (pb) Events PDF set
+ +
POWHEG W →µ ν no cuts 5825 ∼2M CT10
POWHEG W − → µ− ν no cuts 3954 ∼2M CT10
POWHEG W + → τ +ν no cuts 5825 ∼2M CT10
POWHEG W + → τ −ν no cuts 3954 ∼2M CT10
POWHEG Z → µµ mµµ > 20 1631 ∼ 2M CT10
POWHEG Z → µµ 20 > mµµ > 10 3216 ∼ 2M CT10
POWHEG Z → ττ mτ + τ − > 20 1631 ∼ 2M CT10
PYTHIA tt̄ no cuts 162 ∼1M CTEQ6L1
PYTHIA Inclusive µ QCD p̂T > 20, pt (µ) > 15, |ηµ | < 2.5 84679 25M CTEQ6L1
PYTHIA WW no cuts 42.9 ∼1M CTEQ6L1
PYTHIA WZ no cuts 18.3 ∼1M CTEQ6L1
PYTHIA ZZ no cuts 5.9 ∼1M CTEQ6L1
Table 5.2: Summary of analyzed Monte Carlo samples for the various signal and back-
ground processes. In the first column the generator used to simulate the sample in the
second column is shown. Third column shows the generator cuts for the sample produc-
tion. In forth column the cross section computed at NLO for each process is presented
with the number of total events in the fifth column. Last column shows the PDF set used
in the generation of the sample.
5.4.2 Muon reconstruction and identification
Typical W boson events consist of a high pT , isolated muon, with a big amount of MET
consistent with the existence of a high ET neutrino. A good muon reconstruction is re-
quired to suppress instrumental background from punch-through, meson decay in flight,
cosmic muons, etc. and to ensure an accurate transverse momentum measurement.
Muons are reconstructed separately in the tracker and in the muon chambers. In
order to guarantee an excellent reconstruction, the muon has to be reconstructed as a
global muon and as a tracker muon (36). For the typical momentum of muons coming
from a W boson, the resolution in the transverse momentum of the muon is dominated
by the tracker measurement. Thus a minimum number of hits in the tracker when
building the muon trajectory is required to guarantee an accurate muon transverse
143
momentum measurement. At least 10 hits in the tracker and 1 hit in the pixels should
take part in the track reconstruction. Also to avoid pathological bad reconstructed
muons we reject events with a poor χ2 of the global muon fit. Events with the muon
reconstructed with a χ2 over the number of degrees of freedom bigger than 10 are not
considered.
In order to avoid missidentification of muons (punch-through in our muon chambers)
we require the events to be detected by at least two muon chambers. Punch-through
particles will lose energy in the iron of CMS wheels and will not reach the second station
most of the times. At the same time, this requirement makes the selection to match
the requirements used in the trigger. It is also required that at least one hit from the
muon chambers participate in the final muon track fit.
Besides the background originating from other physical processes produced in the
proton-proton collisions additional contamination may come from cosmic muons cross-
ing the detector. Background from on time cosmic muons (those synchronous with
CMS data taking) is reduced by requiring the track to pass within 2 mm to the pri-
mary interaction point in terms of the impact parameter in the transverse plane. When
applying both requirements (on time muon and small impact parameter) the contami-
nation from cosmic events is negligible. In Fig. 5.7 the impact parameter distribution is
shown. All the collision events are present in the central bin. To estimate the remaining
number of cosmic muons entering this bin, we extrapolate from the region dominated
with cosmics. We assume the rate of cosmic muons arriving to CMS is constant in the
impact parameter. The high range of the impact parameter distribution is fitted to a
flat distribution. This rate is then extrapolated to the low impact parameter region.
It gives an estimation of the number of this kind of events in the sample to be of the
order of 1 in 104 .
5.4.3 W → µν selection
We already anticipated the muon criteria to select a pure sample of W boson events:
muon pT bigger than 20 or 25 GeV, isolation variable below 0.1 or 0.15. The other
characteristic of this type of events, the presence of a large ET imbalance, will serve as
the final discriminant variable to distinguish between signal and background.
The aim of the selection process is to remove as much background as possible
keeping our signal as much as we can. The dominant background are the inclusive
144
Figure 5.7: Transverse impact parameter distribution. - Distribution of the muon

track impact parameter in the transverse plane (referred to the beam spot) of the selected
muon, before cutting in this variable. The vertical lines indicate the threshold that define
the signal region |dxy | < 2 mm.
muons produced as a consequence of QCD interactions. In the first analysis performed

in CMS (57) the selection requirements were set rather loose: a muon pT threshold
at 20 GeV and an isolation threshold at 0.15. The compromise was to keep the signal
efficiency as high as possible, given the limited size of the sample, at the cost of a relative
large QCD background. One of the main contributors to the systematic uncertainty
budget was precisely the modeling of the QCD background in terms of the discriminant
variable.
The full 2010 data set, ten times larger than the one used in (57), offers the pos-
sibility to review the selection process with the aim to reduce mainly the systematic
associated to the background subtraction.
The selection optimization cannot focus only on the background subtraction un-
certainty, but one has to evaluate the whole picture. Any selection process provoke
also a loss of signal efficiency that will have to be corrected later in the analysis. The
evaluation of these correction factors induce as well certain systematic uncertainties
that will have an impact in the final measurement. In general, the larger the correction
factors will be, the larger the systematic uncertainties will be.
145
To optimize the selection process we will play with the two main variables to dis-
criminate between signal and background: the muon pT and the isolation variable. The
figure of merit we will use in the optimization process is the systematical uncertainty
on the cross section measurement.
The sources of systematic uncertainties we will consider are the background model-
ing, the initial state radiation (ISR) effect and the PDF uncertainty in the evaluation
of the signal acceptance. As we remove the QCD background component we expect
the background modeling to be less important in the cross section measurement. As a
consequence, when performing a more restrictive cut we expect the background related
systematical uncertainty to decrease. The ISR effect related systematic uncertainty is
expected to increase when making the selection process more restrictive. The typical
muon pT in a W boson decay is between 20 and 50 GeV, with a maximum between
30 and 40 GeV. When moving the muon pT requirement to higher values, entering the
signal region, any change in the modeling of the boson pT (and the muon one) affects
directly the measurement. The uncertainty due to the ISR effect affects the muon pT ,
and, as a consequence, the increase of the systematical uncertainty with the increase
of the muon pT requirement is expected.
The complete list of systematical uncertainties considered and the way they are
computed will be explained in the following sections. However, a brief description of
the way they are computed is necessary to understand this optimization process.
QCD background subtraction. Background subtraction is performed evaluating
the MET spectrum of the candidate sample. No reliable description of QCD processes
is available in the MC and thus an effort was made to derive this shape directly from
control samples in data. Several options were considered. First, the MET shape was
taken from a QCD-enriched control sample, taking the candidate sample of events with
a high-pT muon, but non-isolated (unlike the signal characteristics). This MET tem-
plate can be improved, as it will be explained later, to account for observed correlations
with the isolation variable. Finally, the MET shape is predicted by the MC, although
known not to be fully reliable, is also used. The maximal difference in the W cross
section obtained using the different templates is taken as an estimate of the systematic
uncertainty associated to the method.
The other two sources of uncertainty considered lie more on the theoretical side as
they affect the determination of the correction factor to extrapolate the measurement
146
in a reduced fiducial volume to the whole phase-space.

In the first case we study how this correction factor varies with the different param-
eters that govern the modeling in the MC of Initial State Radiation (ISR) effects. This
variation is studied for different values of the parameters that define the measurement
phase-space: varying the pT threshold from 20 GeV to 30 GeV, and for two values of
the isolation threshold, 0.1 and 0.15.
Similarly we evaluate the impact of the uncertainty on the knowledge of the Par-
ton Distribution Functions (PDFs) in the determination of the acceptance factor, for
different values of the pT threshold, from 20 GeV to 30 GeV, defining the measure-
ment phase-space. For that purpose, several PDF sets are used and variation in the
acceptance factor evaluated.
In Fig. 5.8 the three systematic uncertainties studied in this section and their de-
pendence with the pT requirement is shown for an isolation variable maximum of 0.10
(black dots are used for the QCD background modeling systematic uncertainty, the red
squares for the ISR effect one, green triangles for the PDF one and blue triangles to
represent the sum of the three components). The same is plotted in Fig. 5.9 in which
now the isolation variable maximum is 0.15. The behaviour of the systematic uncer-
tainty studied is very similar in both cases (isolation variable maximum of 0.1 or 0.15).
The systematic uncertainty due to the QCD background subtraction decreases with the
pT threshold. It ranges from a 2% to 0.4% for pT threshold of 20 GeV and 30 GeV, for
an isolation threshold of 0.1, and from 2.8% to 0.6% for an isolation threshold of 0.15.
The uncertainty due to the modeling of ISR effect varies from ≈ 1% to ≈ 2.8%, not
depending significantly on the isolation threshold. Finally, the uncertainty due to the
PDFs diminishes softly from 1.8% to 1.4% with no dependence on isolation. The total
error for the isolation threshold at 0.1 varies from ≈ 5% to ≈ 4.8% when changing the
pT threshold from 20 GeV to 30 GeV, with a minimum at ≈ 4.2% for a pT threshold
of 25 GeV. The corresponding numbers for iso < 0.15 are a variation from 5.8% to 5%
with a minimum of 4.4% at 25 GeV.
From these plots we can conclude that we get a minimum of the systematic un-
certainty when the pT requirement is set to approximately 25 GeV. We also observe a
reduction in the systematic uncertainties when the isolation cut is set to 0.1.
The selection process to get an optimal sample of W candidates to compute the
cross section production can be summarized as follows:
147
Error (%)
5
0
20 22 24 26 28 30
ptCut (GeV)
Figure 5.8: - Variation of the systematic uncertainty studied in this section with the pT
requirement applied. Black dots are used for the QCD background modeling systematic
uncertainty, the red squares for the ISR effect one, green triangles for the PDF one and blue
triangles to represent the sum of the three components. An isolation variable maximum of
0.1 is applied in this plot.
6
Error (%)
0
20 22 24 26 28 30
ptCut (GeV)
Figure 5.9: - Variation of the systematic uncertainty studied in this section with the pT
requirement applied. Black dots are used for the QCD background modeling systematic
uncertainty, the red squares for the ISR effect one, green triangles for the PDF one and blue
triangles to represent the sum of the three components. An isolation variable maximum of
0.15 is applied in this plot.
148
1. The event has to be selected online by any of the HLT triggers seeded by a muon
identified in the muon spectrometer (HLT Mu9, HLT Mu11 or HLT Mu15).
2. We define a W candidates as an event with a reconstructed muon.
3. Contribution from Drell-Yan events is suppressed by removing events with two

high pT muons (20 GeV as threshold for the highest pT muon and 10 GeV for the
other one).
4. The reconstructed muon has to fulfill the quality criteria presented before (muon
reconstruction and identification requirements).
5. Pseudorapidity of the muon should be below 2.1, to match the region where the
muon trigger is more stable.
6. Finally, the selection requirements studied in this section are applied. Muon pT
should be bigger than 25 GeV and the isolation variable should be below 0.1.
This selection procedure is applied to the full 2010 data sample.

The breakdown of the data reduction at the different stages of the selection is
summarized in Table 5.3 both for the total sample of muon events, and splitted by the
muon charge.
Event Sample Events with µ± Events with µ+ Events with µ−

Candidates 3490239 1812890 1677349
Triggered 3174722 1644727 1529995
DY Rejection 3133420 1623994 1509425
Muon ID 2618199 1346297 1271902
|η| < 2.1 2527047 1298777 1228270
pT > 25 GeV 412266 222080 190186
rel
Icomb < 0.10 166457 97533 68924
Table 5.3: Data reduction at every step of the selection process. The number of events
is given for the whole muon data sample, and also separated by the muon charge.
After the selection process just described, 166 457 events are selected, 97 533 of
them with a positive charged muon and 68 924 with a negative charged muon. The
estimated number of events of the different processes (signal and background) to the
W → µν candidate sample are summarized in Tab. 5.4.
149
source Nbg in 36.1 pb−1 Nbg /(NW + Nbg ) Nbg /(NW + Nbg ) (MET> 20 GeV)
QCD multi-jet 8831 5.1% 0.6%
Z → µµ 6130 3.5% 2.9%
W → τν 4623 2.7% 2.9%
Z → ττ 910 0.5% 0.2%
W W +W Z+ZZ 205 0.1% 0.1%
tt̄ 592 0.3% 0.4%
EWK + tt̄ 12538 7.1% 6.4%
total 21434 12.2% 7.0%
W → µν signal 152676 87.8% 93.0%
Table 5.4: Estimates of signal and backgrounds in the W → µν candidate sample, based
on Monte Carlo simulations. The second column gives the number of events expected
from each type of processes in a 36 pb−1 sample, based on the theoretical predicted cross
sections. The third column shows the fraction of each of the processes that contributes to
the candidate sample. The last column indicates the fraction of events in a reduced phase
space MET > 20 GeV.
5.4.4 Acceptance and efficiencies computation
In this section we will compute the detector acceptance (A), fraction of the W phase
space accessible by the CMS detector, and the efficiency of the event selection (ǫ) for
W events within the detector acceptance.
The detector acceptance has to be evaluated necessarily by MC simulations. The
detector efficiency can be evaluated either by MC simulation or with control samples.
The approach we have followed in this thesis is to compute by MC a single factor,
Aw , accounting for the two effects. This Aw factor will be in fact equivalent to the
product A × ǫ. Any deviation of the MC description of the detector efficiency will be
later evaluated with independent control samples and appropriate correction terms are
derived to modify the Aw factor just computed.
However it is difficult to have a perfect description of the MET behaviour in the
MC. Thus, remaining discrepancies between data and MC (or remaining miscalibrations
in the MC) of the MET scale and resolution will be corrected with information from
Z → µµ samples.
The signal acceptance times efficiency (Aw ) term is computed with the W → µν
150
Acceptance × efficiency (Aw )

W± W+ W−
0.4071 ± 0.0003 0.4134 ± 0.0004 0.4021 ± 0.0004
Table 5.5: W signal acceptance and efficiency (Aw ) factors evaluated on the MC
POWHEG sample. The errors shown are the statistical error due to size of the MC sample.
POWHEG reference sample listed in Tab. 5.2. An event lies within the detector accep-
tance and is selected if the pseudorapidity of the reconstructed muon is below 2.1 (in
absolute value) and its transverse momentum is higher than 25 GeV and passes all the
selection requirements. This factor is computed as
Nsel
Aw = .
Ngen
The values of this factor are shown in Tab. 5.5.

Single muon efficiencies prediction from the MC is probed in Z → µµ samples.
Muon efficiencies are evaluated with a Tag and Probe method both in a Z → µµ MC
sample and in data, and a ρ factor is defined as the efficiency in data over the efficiency
in MC. This ρ factor will be used to correct the MC prediction of the Aw term. When
doing this we extract the efficiencies from a sample of muons kinematically very similar
to those from the W decay and exploits the relatively clear selection of Z candidates.
The possible presence of background processes in the selected event sample is also taken
into account. The method developed is explained in detail in (77, 78). ρ factors are
computed for different regions in pseudorapidity to reproduce the real structure of the
detector (lower efficiency in the transition regions between the barrel wheels and the
barrel and endcap regions) and split by charge to account for possible asymmetries
in the detection (selection of positive and negative muons). The overall ρ factors are
shown in Tab. 5.6.
A small fraction of muon events are lost because of L1 muon trigger prefiring, i.e.,
the assignment of a muon segment to an incorrect bunch crossing, occurring with a
probability of a few per mille per segment. The effect is only sizable in the drift-tube
system. Since this effect is not accounted for in the efficiency from tag and probe we
have to consider it separately (we correct the ρ factors by a factor 0.995).
151
µ+ µ− µ±
ǫdata (85.98 ± 0.38 ± 0.72)% (85.00 ± 0.36 ± 0.72)% (85.48 ± 0.27 ± 0.72)%
ǫM C (89.25 ± 0.05)% (89.38 ± 0.05)% (89.32 ± 0.04)%
ρ (96.33 ± 0.43 ± 0.81)% (95.09 ± 0.40 ± 0.81)% (95.70 ± 0.30 ± 0.81)%
Table 5.6: Efficiencies and corrections factors for positive, negative and full sample of
muons. The first uncertainty shown is statistical and the second systematical.
Acceptance
W± W+ W−
0.4638 ± 0.0003 0.4706 ± 0.0004 0.4570 ± 0.0004
Table 5.7: .
To have an idea of how much of the Aw term is due to detector acceptance and to
detector efficiency, we include also the values of the acceptance. We will define that an
event lies in the detector acceptance if it has a muon with a pT > 25 GeV in the region
|η| < 2.1, at generator level:
Ngen (pgen
T > 25 GeV, |η| < 2.1)
A= .
Ngen
The values of the acceptance are given Tab 5.7.
5.4.5 W signal extraction
We now focus on the numerator of the previous formula to compute the production
cross section:
(Nsignal − Nbackground )
σ= . (5.9)
AǫL
To evaluate the number of signal and background events in the selected sample,
we focus on a highly discriminant variable, the missing transverse energy (MET). This
variable allows to clearly discriminate signal from the dominant background QCD. The
strategy will be to perform a binned likelihood fit to the observed MET distribution of
a sum of the different contributions already discussed (signal, QCD background, EWK
backgrounds, dibosons and tt̄). The shape of the signal and background MET distri-
bution are modeled as individual templates and the amount of different contributions
152
is allowed to float in the fit. The way the MET templates are built for the different
contributions and the details in the fit will be shown in the following subsections.
MET signal template. MET template for signal can be drawn from MC simula-
tion. ElectroWeak processes are well known, and the description of the MET shape for
the signal contribution can be safely extracted from the MC prediction. To improve
the MET description, any possible miscalibration of the ET scale and resolution in the
MC, it is corrected for with information from Z → µµ event (79).
W and Z bosons have similar recoil models, they are produced at a similar Q2 and
their production and decay processes are similar. The idea is to extract information
from Z → µµ events where the whole event can be reconstructed in the transverse plane
and use it to model the behaviour of W events. The recoil to the Z, in the transverse
plane, i.e. all the energy of the event but the two muons, can be decomposed in
two orthogonal directions: parallel and perpendicular to the Z direction. The parallel
component of the recoil is due to initial state gluons radiated from the quarks that
produce the Z. This emission balances the pT of the boson, and therefore the average
value of METk is expected to increase with Z-pT . The second component (MET⊥ ),
is due to multiple interactions and remnants of the beam particles involved in the Z
production. The average value of MET⊥ is essentially zero with a certain resolution
effect.
Figure 5.10: Recoil distribution in Z → µµ events: the plots show the parallel (left) and
perpendicular (right) components, as a function of the pT of the boson.
153
Figure 5.11: MET distribution in Z events: Parallel (left) and perpendicular (right)
components, plotted for several intervals of the boson pT (from bottom to top), [0 − 20]
GeV, [20 − 30] GeV, [30 − 40] GeV and [40 − 60] GeV.
In Fig. 5.10 the parallel and perpendicular components of the Z boson recoil as a
function of the Z pT in real data is shown. The information is stored binned in the Z
boson pT , with a variable binning to have a uniform population in every bin. For every
MC simulated W with a given pT , we sort randomly a value for the METk from the
slice corresponding to the same pT of the Z boson. The perpendicular component, since
it does not depend on the boson pT , is sorted from the global distribution. In Fig. 5.11
the METk and MET⊥ projections for several bins in the Z boson pT are shown. The
MET so constructed can be combined with the rest of the information of the simulated
event (reconstructed muon, angle between the muon and the MET, etc.) to calculate
any derived quantity. The resulting W template is presented in Fig. 5.12 compared to
the prediction of the MC simulation. One can appreciate a slight broadening of the
shape towards higher values of MET.
QCD MET template
It is known that MC simulation for QCD related processes is not to be accurate
enough as to fully rely on them to draw conclusions about their characteristics. We
will define an independent control sample from which this shape can be extracted.
The distribution of the isolation variable is shown in Fig. 5.13 for the MC simulation.
The threshold in the isolation variable for the signal selection was set at 0.1. One can
appreciate that events with isolation variable bigger than 0.2 are mostly QCD events.
154
Figure 5.12: The plain prediction from the MC (dashed-line) is compared with the
template incorporating the Z recoil correction (solid line).
We can build a MET template for QCD background as the MET shape of the non-
isolated events. However, as shown in Fig. 5.14, the non-isolated QCD events do not
behave like the isolated QCD events. In particular, for the simulation, MET varies with
the isolation variable. This correlation can be removed applying a correction to MET
comb . Under these conditions a new corrected template is
which varies linearly with Irel
obtained applying this correction to the events in the non-isolated region. The physical
reason of this correlation is the dependence of the isolation quantity (energy in the
calorimeters and pT of the tracks) on the total activity of the event (ΣET ). Fig. 5.15
shows the corrected template that is now much closer to that shape of the isolated
region. The same kind of behaviour is observed in the data (Fig. 5.14). There is also
a linear dependence of the MET variable with the isolation as in MC, although with a
slightly different slope. Following the same spirit that in MC we will extract in data
the MET shape from the non isolated region, once the correlation is corrected. The
MET-isolation dependence was obtained as MET′ = MET(1 + αIcomb
rel ) with α ≈ 0.19.
The QCD templates obtained using this procedure are shown in Fig. 5.16. There it is
plotted the MET shape from the non isolated region (dashed line) and how it is modified
(dots) once the MET-iso correlation is removed. This will be the final template used
in the W cross section measurement. It is also compared with the MC prediction for
the isolated region. They do not fully agree, but, as it was already discussed, it is not
expected that QCD MC reproduces completely the behaviour of the data. Finally, the
155
red rectangles indicate the systematic uncertainty assigned to the QCD MET shape,
evaluated as discussed in Section 5.5.
QCD template is assumed to be the same in shape and magnitude for both charges.
This assumption is shown to be correct within the errors associated to the QCD shape.
Figure 5.13: Isolation variable distribution. - Isolation distribution of candidates

with a good quality muon of pT > 25 GeV in the fiducial region |η| < 2.1. Dots represent
the data and the solid histograms the contribution from the different SM processes, evalu-
comb
ated by MC and normalized to the theoretical cross sections. The blue arrow (Irel < 0.10)
indicates the threshold that defines the signal region. The green arrow defines a background
comb
control region, Irel > 0.20.
MET Template for the other ElectroWeak processes. We rely on the MC

prediction to derive the shape of the MET distribution of the candidates originating in
the other electroweak processes (W → τ ν, Z → τ τ and Z → µµ). A similar procedure
as the one followed with the signal can be envisaged. However the improvement in the
MET description in this case is found to be negligible and within the errors associated to
this modeling, with a negligible impact in the final result (the electroweak background is
of the order of 5% in the selected sample). The template used for the other electroweak
processes is shown in Fig. 5.17. The absolute level of background will be derived from
the fit.
MET templates for tt̄ and diboson events. Contribution from dibosons pro-
duction and tt̄ production to the final sample of candidates is negligible. The shape
is obtained from MC simulation and their absolute normalization from theoretical pre-
dictions.
156
Figure 5.14: - Correlation of MET with the isolation variable. Red circles are the
prediction from QCD events and the blue squares the data behaviour. A linear dependence
is observed and extracted to improve the template for the QCD sample. The first point
(low isolation variable) presents a high MET value. This is due to the presence of signal
events, entering specially in this bin.
5.4.6 Cross section measurement
Once we have modeled the behaviour of the MET for the different contributions we can
proceed to the W cross section evaluation. The total contribution to the number of
selected events is divided in the 4 previously described: signal, QCD, EWK and others.
Technically speaking, we perform a binned fit to minimize the likelihood between data
and the sum of the contributions. In each bin, the number of expected events is:
N (MET) = NW (MET) + NEWK (MET) + NQCD (MET) + NOthers (MET) (5.10)
The W and EWK terms are expressed in terms of their cross sections, their ac-
ceptances and efficiencies (AW and AEW K include both factors) and their probability
distribution functions (p.d.f) in the MET variable (from the templates), FW and FEW K .
The EWK contributions, Z → µµ, Z → τ τ and W → τ ν are normalized to the W → µν
signal, through their theoretical cross section ratio (K factor). The QCD contribution
is described as well in terms of a p.d.f. (template) on MET (FQCD (MET)) and a con-
stant (NQCD ) setting the absolute background level. The contribution given by the rest
157
Figure 5.15: Distribution of the corrected MET for selected events with a non isolated
muon (black points) superimposed on the distribution of uncorrected MET for the same
events (blue, dashed line) and MET for events with an isolated muon (black, solid his-
togram). All distributions are from simulated QCD events. The shaded area represents
the systematic uncertainty due to corrections with factors α ± ∆α, for ∆α = 0.08.
of the processes, as explained before, is computed from MC and theoretical predictions

and is constant through the fit process. We can compute the expected number of events
in a certain region of the MET spectrum as:
N (MET) = {σW ×[AW ·FW (MET)+K·AEWK ·FEWK(MET)]+NQCD ·FQCD (MET)}

× Lint + NOthers (5.11)
This fitting process is performed simultaneously for the positive and negative chan-
nels. The fitting function can be expressed in terms of two different sets of parameters:
1. the total W cross section (σ(W+ ) + σ(W− )) and the ratio R = σ(W+ )/σ(W− ),
together with the overall normalization of QCD events (NQCD ),
2. the individual σ(W+ ), σ(W− ) and the overall normalization of the background.
158
Figure 5.16: Distribution of the corrected MET for selected events with a non isolated
muon in data (black points) superimposed on the uncorrected MET distributions for data
(blue dashed line) and simulated QCD events (black, solid histogram, same as the black,
solid histogram in Fig. 5.15. The shaded area represents the systematic uncertainty due
to corrections with factors α ± ∆α, for ∆α = 0.08.
Arbitrary units
Arbitrary units
0.06 0.06
0.05 0.05
0.04 0.04
0.03 0.03
0.02 0.02
0.01 0.01
0 0
0 20 40 60 80 100 0 20 40 60 80 100
MET [GeV] MET [GeV]
Figure 5.17: Sum of all MET distribution of the different electroweak processes considered
(appropriately normalized, using the theoretical cross section for each electroweak process)
for minus (left) and plus (right) processes. In both figures the area is normalized to the
unity.
159
The fit is performed over the MET range [0, 200] GeV. Fitted W yields of 141226 ±
376 (total), 84315 ± 290 W+ and 56911 ± 239 W− , events are obtained.
The total W cross section, and the individual W+ and W− cross sections, as well
as their ratio, obtained from the fit are presented in Table 5.8. The errors shown are
only statistical.
In Fig. 5.18, the experimental distribution in the MET variable together with the
other contributions after the fit is shown, both in linear and logarithmic scale. A good
agreement is observed after the fit in all the MET range plotted.
The W-signal extraction procedure can also be formulated in terms of a fit to the
MT distribution, simply deriving the corresponding MT templates from those in MET
(equation 5.5). In Fig. 5.19 the results of the fit in the MT variable are plotted.
Numerical results from the MT fit are also given in Tab. 5.8. The two sets of results
are in full agreement. As we will see as one of the conclusions of this Chapter, a
requirement of MT > 50 GeV will guarantee the selection of an almost background free
W signal. We can notice in the Fig. 5.19 the decrease of the background when this
later requirement is set.
5.5 Systematical uncertainties
The measurement presented is affected by the uncertainties in all the different elements
that take part of it.
In this section we will show the estimation of these systematical uncertainties. They
are grouped as experimental uncertainties and theoretical ones, depending on the source
of the uncertainty. Luminosity uncertainty (dominant in the measurement) is consider
separately.
As experimental sources we will consider the uncertainties on the lepton reconstruc-
tion and identification, the trigger prefiring, the muon momentum scale and resolution
and the signal and background modeling.
The total uncertainty (statistical and systematic) on the correction factors (ρ) is
taken as the systematic uncertainty due to muon efficiency (reconstruction, identifica-
tion, selection, isolation and trigger).
The muon pT scale and resolution affects the measurement of the W boson cross
section production through the selection based on the muon pT as well as in the MET
160
Fit in MET distribution

σ(W) (nb) 10.101 ± 0.027
R = σ(W+ )/σ(W− ) 1.421 ± 0.008
QCD normalization factor 1.37
σ(W+ ) (nb) 5.929 ± 0.021
σ(W− ) (nb) 4.172 ± 0.018
Fit in MT distribution
σ(W) (nb) 10.150 ± 0.027
R = σ(W+ )/σ(W− ) 1.418 ± 0.008
σ(W+ ) (nb) 5.953 ± 0.020
σ(W− ) (nb) 4.197 ± 0.018
Table 5.8: Total W, W+ and W− production cross section (times the Branching Ratio of
the W decaying into a muon and a neutrino) and ratio between W+ and W− cross sections.
The QCD normalization factor is also presented. The upper set of numbers are obtained
from a fit to the MET distribution. The lower one was obtained from a fit to the MT
distribution.
shape. The SIDRA method presented in Chapter 4 is used to get the corresponding scale
and resolution factors binned in pseudorapidity and azimuthal angle. These factors are
applied to the muons coming from the MC simulation, obtaining a new pT , and thus
new templates and efficiencies. In these conditions, we repeat the cross section fit
previously exposed. The error associated to this source is computed as the difference
of this value with the reference one. A 0.3% variation is obtained.
The mismodeling of the MET shape for the signal is another source of uncertainty,
related with the MET scale and resolution. It is estimated as the difference in the cross
section measurement using the reference template for the signal and the one provided
by MC. This difference is shown to be of 0.2%.
The systematic uncertainty in the background subtraction takes into account the
effect of possible mismodeling of the MET shape of the QCD component on the cross
section measurement. MC studies indicated that the optimal value of α is 0.24 to
correct for the isolation-MET correlation. Fig. 5.20 shows the MC prediction for the
161
× 10
3 36 pb-1 at s = 7 TeV × 10
3 36 pb-1 at s = 7 TeV
10 10
number of events / 2 GeV

data data
8 8
-
+
W → µ +ν W → µ- ν
EWK+tt EWK+tt
6 QCD 6 QCD
4 4
2 2
5 5
χ
χ
0 0
-5 -5
0 20 40 60 80 0 20 40 60 80
ET [GeV] ET [GeV]
36 pb-1 at s = 7 TeV 36 pb-1 at s = 7 TeV
104 104

data data
3 3
10 W → µ +ν
+ 10 -
W → µ- ν
EWK EWK
102 tt 102 tt
QCD QCD
10 10
1 1
10-1 10-1
5 5
χ
0 0
-5 -5
0 50 100 150 200 0 50 100 150 200
ET [GeV] ET [GeV]
Figure 5.18: W+ (left plots) and W− (right plots) experimental distributions (black dots)
of MET, both in linear and logarithmic scale, together with the fitted contributions from
the different processes (shown stacked): W signal (red histogram), other EWK processes
(green histogram), QCD background (pink histogram) and tt̄ (blue histogram).
isolated QCD events (black solid line), the prediction derived from the non-isolated
region (red dashed line) and the prediction once the correlation has been corrected
(green line). The prediction in this last case is much closer to the prediction from the
isolated part. Predictions using other values of α (from 0.16 to 0.32) are also shown
in the Figure. A variation of ∆α = ±0.08 successfully covers the MC prediction for
the isolated region over all MET interval. To evaluate the uncertainty associated to
the method we fit the MET spectrum with two extreme MET shapes using modified α
correction factors according to this variation.
We consider a 0.5% to accout for the L1 muon trigger prefiring, i.e., the assignment
162
× 10
3 36 pb-1 at s = 7 TeV × 10
3 36 pb-1 at s = 7 TeV
15 15
data data

+ -
W → µ +ν W → µ-ν
10 EWK+tt 10 EWK+tt
QCD QCD
5 5
5 5
χ
χ
0 0
-5 -5
0 20 40 60 80 100 120 0 20 40 60 80 100 120
MT [GeV] MT [GeV]
36 pb-1 at s = 7 TeV 36 pb-1 at s = 7 TeV
104 104

data data
3 3
10 W → µ +ν
+ 10 W → µ- ν
-
EWK EWK
tt tt
102 QCD 102 QCD
10 10
1 1
5 5
χ
0 0
-5 -5
0 50 100 150 200 0 50 100 150 200
MT [GeV] MT [GeV]
Figure 5.19: W+ (left plots) and W− (right plots) experimental distributions (black dots)
of MT , both in linear and logarithmic scale, together with the fitted contributions from
the different processes (shown stacked): W signal (red histogram), other EWK processes
(green histogram), QCD background (pink histogram) and tt̄ (blue histogram).
of a muon segment to an incorrect bunch crossing. Since this effect is not consider in
the Tag and Probe method, the uncertainty due to this effect should be accounted for
separately. A 0.5% uncertainty is estimated for this effect.
As theoretical uncertainties we consider the PDF uncertainty and other theoretical
effects such as ISR (initial state radiation) and FSR (final state radiation) effects. The
MC generator used for the signal simulation is POWHEG interfaced with PYTHIA
for parton showering and hadronization and fragmentation. PYTHIA is “tuned” to
better describe the data by modifying a hundred of parameters related to not well
know processes in the collisions, such us multiple interactions, the underlying event,
163
Figure 5.20: MET template for the QCD MC sample (black). With other colors,
variations of this prediction obtained with different α values. The optimal value for α in
MC is computed to be 0.24. In this plot we observe how a variation of 0.08 around the α
central value completely covers the MC prediction. The MC truth refers to the isolated
MC.
etc. Several parameters control the ISR and FSR effects. Variation of standard tune of
PYTHIA in these parameters with respect to the official CMS tune is the way chosen
to estimate the uncertainty given by these effects.
The impact of these in the total W cross section and on the individual W+ and W−
as well as on the ratio is presented in Tab. 5.9.
The experimental error is of the order of 1.1% (total W cross section) and the
theoretical one 1.1%. The total error excluding luminosity is 1.6%. Individual cross
section theoretical uncertainties are slightly larger (the experimental are essentially the
same). Ratios are more affected by theoretical uncertainties compared to the total cross
sections. A 4% uncertainty in the recorded luminosity is assigned.
As a comparison, the 3.1% uncertainty of the previous measurement is now reduced
to 1.6%, thanks to the use of new techniques that profit from the big amount of data
available (ten times bigger than in the previous measurement of the W boson cross
section by CMS).
164
5.6 Kinematic distributions
Source W W+ W− W+ /W−
Lepton reconstruction & identification 0.9 0.9 0.9 1.3
Trigger prefiring 0.5 0.5 0.5 0
Momentum scale & resolution 0.3 0.3 0.3 0
MET scale & resolution 0.2 0.2 0.1 0.1
Background subtraction / modeling 0.4 0.4 0.5 0.1
Total experimental 1.1 1.1 1.2 1.3
PDF uncertainty for acceptance 0.8 0.9 1.5 1.9
Other theoretical uncertainties 0.8 0.9 0.8 0.8
Total theoretical 1.1 1.3 1.7 2.1
Total (excluding luminosity) 1.6 1.7 2.1 2.5
Table 5.9: Systematic uncertainties in percent for individual W cross sections and ratios
in the muon channel. A common luminosity uncertainty of 4% applies to all cross sections.
5.6 Kinematic distributions
In this Section we show several kinematic distributions, necessary to completely char-

acterize the W → µν events produced in the collisions. All the distributions are
normalized to the output of the fit.
In Fig. 5.21 the pT distribution of the W candidates is shown. The 25 GeV selection
requirement pT can be noticed here. The agreement data-MC in this plot is not perfect,
with visible disagreements in the whole range. However, the distribution for the “golden
candidates” (if we additionally require the transverse mass of the candidate to be
higher than 50 GeV) shows a good agreement with MC. In Fig. 5.22 we show the pT
distribution of this subsample.
In Fig. 5.23 (left), the pseudorapidity distribution of the W candidates is presented.
It is possible to notice the efficiency drop in the inter-barrel regions and the over-
laps. Fig. 5.23 (right) shows the azimuthal angle distribution for all the W candidates.
This last distribution is flat in this angle, showing that there is not a preferred decay
directionality in the transverse plane of the W bosons in CMS.
The acoplanarity variable measures the angle between the muon and the MET of
the event. It was defined as (5.6). In W events the muon and the neutrino are usually
back-to-back (differ by π radians in φ). As a consequence the acoplanarity of the muon
and the MET for these events will be low. The acoplanarity distribution for the W
165
× 10
3 36 pb-1 at s = 7 TeV 36 pb-1 at s = 7 TeV
6
10

15 data data
5
10
W → µν W → µν
EWK+tt 104 EWK
QCD
10 tt
3
10 QCD
102
5 10
10-1
5 5
χ
χ
0 0
-5 -5
20 30 40 50 60 70 80 20 40 60 80 100
pT(µ) [GeV] pT(µ) [GeV]
Figure 5.21: Muon pT distribution for the Wµν sample.
× 10
6
10
15 data data
5
10
W → µν W → µν
EWK+tt 104 EWK
QCD
10 tt
3
10 QCD
102
5 10
10-1
5 5
χ
0 0
-5 -5
20 30 40 50 60 70 80 20 40 60 80 100
pT(µ) [GeV] pT(µ) [GeV]
Figure 5.22: Muon pT distribution for the Wµν sample. Only Wµν candidates with a
MT > 50 GeV are included in the plot.
166
5.7 Discussion of the results
× 10
3 36 pb-1 at s = 7 TeV × 10
3 36 pb-1 at s = 7 TeV
data data
3
4 W → µν W → µν
EWK+tt EWK+tt
number of events
number of events
QCD QCD
3 2
1
1
5 5
χ
χ
0 0
-5 -5
-3 -2 -1 0 1 2 3 -2 0 2
η (µ) φ (µ)
Figure 5.23: Muon η and φ distributions for the Wµν sample.
candidates is shown if Fig. 5.24. The signal peaks at low acoplanarity (events are
back-to-back) as expected.
In Fig. 5.25 the pT distribution of the W candidate is shown. In Fig. 5.26 the same
variable is plotted for the W golden candidates (those with MT higher than 50 GeV).
A general disagreement is observed in these plot, specially in the low pT region. Low
pT region is dominated by non-perturbative QCD effects and MC tuning.
The results are summarized in Fig. 5.27 where the total cross section, individual
cross sections and ratio computed in this thesis are compared with the theoretical ones
and with the published results from CMS.
The results on the W boson production cross section in the muonic channel obtained
in this analysis (10.1 ± 0.03 (stat.) ± 0.16 (sys.) ± 0.4 (lumi.) nb) are in agreement
with the SM (FEWZ and the MSTW 2008 PDF predict 10.44 ± 0.27 nb). The charged
cross sections given in this study (5.93 ± 0.02 (stat.) ± 0.1 (sys.) ± 0.24 (lumi.) nb
for the positive and 4.17 ± 0.02 (stat.) ± 0.09 (sys.) ± 0.17 (lumi.) nb) are also in
agreement with NNLO SM prediction (6.15 ± 0.17 nb and 4.29 ± 0.11± nb for the
positive and negative production respectivily), as well as the ratio measured 1.421 ±
167
× 10
3 36 pb-1 at s = 7 TeV
14
12 data
W → µν
number of events
10
EWK+tt
QCD
8
5
χ
0
-5
0 1 2 3
acop
Figure 5.24: Acoplanarity between the muon and the MET distribution for the Wµν
sample.
× 10
6
10
15 data
data
5
10 W → µν
W → µν
EWK
EWK+tt 104
QCD tt
10 QCD
3
10
102
5
10
10-1
5 5
χ
0 0
-5 -5
0 20 40 60 80 0 20 40 60 80 100
pT(W) [GeV] pT(W) [GeV]
Figure 5.25: Reconstructed W-pT distribution for the Wµν sample.
168
× 10
6
10
15 data

data
5
10 W → µν
W → µν
EWK+tt 4 EWK
10
QCD tt
10 QCD
3
10
102
5
10
1
-1
10
5 5
χ
χ
0 0
-5 -5
0 20 40 60 80 0 20 40 60 80 100
pT(W) [GeV] pT(W) [GeV]
Figure 5.26: Reconstructed W-pT distribution for the Wµν sample. Only Wµν candidates
with a MT > 50 GeV are included in the plot.
0.008 (stat.) ± 0.04 (sys.) (to be compared with 1.43 ± 0.01). The value quoted by
the CMS official measurement 10.18 ± 0.03 (stat.) ± 0.16 (sys.) ± 0.4 (lumi.) nb is
compatible with the one presented in this analysis. The individual cross sections given
in this paper 5.98 ± 0.02 (stat.) ± 0.11 (sys.) ± 0.24 (lumi.) for the positive and 4.20 ±
0.02 (stat.) ± 0.09 (sys.) ± 0.17 (lumi.) for the negative, as well as the ratio positive-
negative 1.423 ± 0.008 (stat.) ± 0.04 (sys) are compatible with the results presented in
this thesis.
The method proposed for this measurement is similar to the official CMS measure-
ment (80). The only difference between them is in the way the MET template for
the signal contribution is built. In the present analysis, a “sampling” of the Z recoil
is used to improve the signal MET shape, as previously described. The official CMS
measurement also uses the Z recoil to improve the signal MET description, but in this
case, instead of randomly select a value of the parallel and perpendicular components
following a certain p.d.f., the distributions of the recoil components (parallel and per-
pendicular to the boson pT direction) are fitted with a double Gaussian. The mean and
width of the Gaussian vary with the boson transverse momentum. For each sample,
polynomials are fitted to the extracted mean and width of the recoil distributions as
functions of the boson transverse momentum. The ratios of data to simulation fit-
169
36 pb-1 at s = 7 TeV
NNLO, FEWZ+MSTW08 prediction
[with PDF4LHC 68% CL uncertainty]
10.44 ± 0.27 nb
W → µν (CMS)
10.18 ± 0.03 stat ± 0.16 syst ± 0.40lumi nb
W → µν (thesis)
10.10 ± 0.03 stat ± 0.16 syst ± 0.40lumi nb
0 2 4 6 8 10 12
σ( pp → WX ) × B( W → lν ) [nb]
6.15 ± 0.17 nb
W + → µ + ν (CMS)
5.98 ± 0.02 stat ± 0.11 syst ± 0.24lumi nb
W + → µ +ν (thesis)
5.93 ± 0.02 stat ± 0.10 syst ± 0.24lumi nb
0 2 4 6
σ( pp → W + X ) × B( W + → l + ν ) [nb]
4.29 ± 0.11 nb
W - → µ - ν (CMS)
4.20 ± 0.02 stat ± 0.09 syst ± 0.17lumi nb
W - → µ - ν (thesis)
4.17 ± 0.02 stat ± 0.09 syst ± 0.17lumi nb
0 2 4 6
-
σ( pp → W - X ) × B( W → l - ν ) [nb]
1.43 ± 0.01
W → µν (CMS)
1.42 ± 0.01 stat ± 0.04 syst
W → µν (thesis)
1.42 ± 0.01 stat ± 0.04 syst
0 0.5 1 1.5
R+/- = [ σ × B ](W + ) / [ σ × B ](W - )
Figure 5.27: Summary of the results shown in this Chapter. Total cross section (upper
plot), positive and negative W boson production cross section (medium plots) and the
positive-negative ratio (lower plot) are compared with the results from the official CMS
measurement and the theoretical prediction (yellow band).
170
parameters from the Z samples are used as scale factors to correct the polynomials
parameters of the W simulated recoil curves. For each W simulated event, the recoil
is replaced with a value drawn from the distribution obtained with the corrected pa-
rameters corresponding to the W pT . The MET value is calculated by adding back the
energy of the W lepton.
In this analysis we preferred to sample the real Z distribution instead of parametriz-
ing and fitting it so that all effects not accounted for in the Gaussian description are
included.
CMS published a first measurement with the first 3 pb−1 (57). The method in that
analysis to compute the cross section is similar to the one presented in this Chapter,
with several differences. The selection requirement were not the same. In order to
maximize the number of final events in this measurement it was preferred to set looser
requirements on the muon pT and on the isolation variable (20 GeV and 0.15 respec-
tively). Efficiencies were also computed with the tag and probe method. However,
due to the limited statistics available, it was not possible to split by charge, nor to
compute it for different regions in pseudorapidity. As a consequence we expect a reduc-
tion of the systematic uncertainties in this analysis with respect to the previous one in
the background modeling (we are using an optimal requirement based on this uncer-
tainty), and in the muon reconstruction and identification. The muon reconstruction
and identification is reduced from 1.5% to 0.9%. The biggest improvement observed is
in systematic uncertainty due to background substraction, that amounts now to 0.4%.
The good level of understanding of the data also allowed to drop any remaining de-
pendence on the QCD MC for the evaluation of the associated systematic uncertainty
that are now evaluated in a more realistic way, based only on data information. In the
previous measurement a 2% systematic uncertainty was assigned for the backgound
modelling error. A better measurement of the delivered luminosity by the experiment
made possible to reduce the 11% luminosity uncertainty of the previous measurement
to the 4% that is considered in this measurement. The measurement published with 3
pb−1 quoted a value for the cross section of 9.92±0.09 (stat.)±0.31 (syst.)±1.1 (lumi.)
nb, individual cross sections of 5.84 ± 0.07 (stat.) ± 0.18 (syst.) ± 0.64 (lumi.) for the
positive and 4.08 ± 0.06 (stat.) ± 0.15 (syst.) ± 0.45 (lumi.) for the negative. The ratio
quoted is 1.433 ± 0.026 (stat.) ± 0.054 (syst.). All these results are in agreement with
the updated study.
171
The other general purpose LHC experiment (ATLAS) also published the W boson
inclusive cross section in the muonic channel with 36 pb−1 (81). Their measurement was
in several aspects different to the one presented in this Chapter. The selection process
included a requirement on the MT of 40 GeV. The MC generators used by ATLAS
(MCatNLO for signal and Z → µµ, PYTHIA for other electroweak processes) differ
from the ones used by CMS. ATLAS measurement relies on the MC predictions (does
not improve the MET description of the signal) and does not perform a fit to extract the
normalization of the different contributions (a cut and count method is applied). The
results published in this measurement (10.21 ± 0.03 (stat.) ± 0.23 (syst.) ± 0.35 (lumi.)
nb) show a bigger uncertainty, with a leading systematic uncertainty of 1.5% from the
acceptance computation.
The measurement performed in this analysis shows a result with a precision at a
really challenging level of ≈ 2% as a systematic error. With the data collected by CMS
(5 fb−1 ) at the end of 2011 this measurement could be updated. However the limiting
luminosity uncertainty and the increasing prescale triggers make this measurement not
to be atractive from the physical point of view with 2011 data. With the LHC probably
running at 8 TeV in 2012, a new point in the collision energy-cross section could be
provided. A better description of the W boson transverse momentum and MT shape
are some of the goals for 2012 related to the W boson production in the collisions.
172
6
Measurement of associated
production of W bosons with
charmed jets
At tree level W bosons are produced via quark currents (qq̄) . The inclusive production
was the subject of the previous Chapter. With increasing recorded luminosity in the
experiment, the amount of W bosons produced in association with a jet (group of tracks
in the same region of the detector produced by the fragmentation and hadronization of
quarks) in the final state starts to be relevant. The study of the associated production
of W bosons with charmed jets is the subject of this Chapter.
The associated production of W bosons with charmed jets at LHC differs from
other previous experiments due to the high energy of the collisions and the nature
of the colliding particles. At very high energy, the presence of sea quarks inside the
proton starts to be relevant. At LHC, the associated charm production with W bosons
is mainly produced via the diagram shown if Fig. 6.1 (s̄g → W+ + c̄ and sg → W− + c
processes at the hard scattering level), thus directly probing the strange quark content
of the proton.
Strangeness composition in the proton is the worst known light quark composition.
Many physical studies would benefit from a better understanding of the proton compo-
sition at high energy. In this context, the relevance of the study of the W+c channel,
which gives direct access to the proton strange content, is remarkable. In particular, at
LHC, such a study would shed light on the strange quark composition over a wide range
173
6. MEASUREMENT OF ASSOCIATED PRODUCTION OF W BOSONS
WITH CHARMED JETS
Figure 6.1: Main diagrams at the hard scattering level for associated W - charm pro-
duction at the LHC.
174
Figure 6.2: Correlation coefficient for the W+c cross section measurement and each
of the quarks in the proton (82). The correlation is given for different values of the x
parameter.
of the x parameter (0.0001 ≤ x ≤ 0.1), specially at x ∼ 0.02. In Fig. 6.2 the correlation
coefficient of the W+c cross section with each of the parton PDFs is shown. It gives
an idea of the sensitivity of this measurement to constrain the different quark proton
densities. A high correlation coefficient implies a high sensitivity of this measurement
to that PDF. In particular, a big sensitivity to the strange content previously claimed
is observed (82).
One of the studies that would specially benefit from such a measurement is the
challenging determination of the W boson mass (83). The W boson mass is a key
parameter in the Standard Model. This model does not predict the W boson mass, but
it predicts the relation between the W boson mass and other experimental observables:
πα 1
r
MW = √ √
2GF sin(θW ) 1 − ∆r
where α is the constant coupling, GF Fermi’s constant, θW the weak angle and ∆r
the radiative corrections. These radiative corrections include loop diagrams in which
the Higgs boson and the top quark appear. As a consequence, ∆r depend on the top
quark mass and the Higgs boson mass. Being the current precision on the top quark
mass of 0.9 GeV (measured by Tevatron) a precision of 5 MeV on the W boson mass is
required in order to have both measurements contributing equally to the Higgs boson
mass uncertainty. The world average measurement has an uncertainty of 23 MeV, 5
175
WITH CHARMED JETS
Figure 6.3: World average measurement for the W boson mass and the top quark mass.
The dark region represents the allowed region for the SM depending on the Higgs mass.
times bigger than the desired one 1 . In Fig. 6.3 the constrain set by the combined
results on the top quark mass and the W boson mass is shown. In this plot the value
of the W boson mass versus the top mass is shown. The uncertainty in the top mass
and W boson mass measurement is shown as an ellipse centered in the world average
value. This average value in this Fig. is shown to be one sigma away from the limit on
the Higgs mass imposed by LEP measurements.
The leading theoretical uncertainties on the W boson mass measurements are pro-
ton PDFs. A better understanding of the proton PDFs is mandatory to reduce the
uncertainty of the W boson mass.
In Fig. 6.4 the differences beyond their current uncertainties on the strange PDF
from the different PDF sets, in the range of x 0.01-0.1 can be clearly noticed.
In this Chapter we present the measurement of the ratios Rc± = σ(W + +charm)/σ(W − +
charm) and Rc = σ(W + charm)/σ(W + ≥ 1 jet). The measurement of ratios instead
of cross sections has the advantage of being subjected to lower uncertainties (many
uncertainties are cancelled when performing ratios, like the luminosity uncertainty). It
is also easier since we do not have to deal with certain efficiencies that are also canceled
1
We don not consider recent results on MW and Higgs searches
176
Figure 6.4: Strange quark content (PDF) on the proton as a function of the x parameter
for the PDF sets used in this Chapter: MSTW, CT10 and NNPDF. A lack of agreement
is observed between the three groups in the intermediate region (0.01-0.1).
out when performing the ratio. At the same time, the strange proton composition
is also sensitive to the ratios, supporting this choice. At the LHC, analytical calcu-
lations from the MCFM program (44) at next-to-leading order (NLO) predict ratios
Rc± ≡ σ(W+ +charm)/σ(W− +charm) ≈ 1 and Rc ≡ σ(W+charm)/σ(W+jets) ≈ 0.1.
Other contributions to this final state are strongly suppressed. For example the pro-
duction of a W boson with a charm jet is also possible with a d quark in the initial state
(dg → W− + c). However this process is Cabibbo suppressed with the matrix element
kVdc k being small (≈ 0.2). This contribution is even lower for the W + (d̄g → W+ + c̄),
since the d̄ is not a valence quark. Both contributions account approximately for 15%
and 5% of the signal respectively. Gluon splitting of the type dū → W− + g → W− bb̄
or dū → W− + g → W− cc̄ will also contribute to our background and our signal (re-
spectively), but with a smaller impact (≈ 1%).
Charm quarks are the third most massive quarks of the Standard Model. They have
spin 1/2 and charge 2/3. Charmed-hadrons decay (due to conservation of the internal
charm quantum number by the strong interaction) via the weak force. In charmed
hadrons, the c quark is preferentially transformed into a s quark. Thus, charmed-
hadrons decay is mostly observed into kaons and pions. This decay process has a lifetime
of the order of 10−13 seconds. As a consequence, these hadrons usually travel several
mm before decaying and thus, producing a secondary vertex that can be observed in
the detector. The presence of the charmed-hadrons can be distinguished statistically
177
WITH CHARMED JETS
from a b-meson or a light meson for having a short flight distance compared to the b
hadrons and larger than the light hadrons. Other properties of the charmed-meson jets
that allow us to distinguish them from other kind of jets are their multiplicity, mass and
the opening of the jet cone. They are known to be intermediate between the b-hadrons
and the light quark hadrons in terms of mass of the jet, number of tracks in the jet and
angle of the jet cone. The techniques that allow us to distinguish a charmed-hadrons
from other hadrons are known as c-tagging techniques, and exploits the characteristics
shown before.
For this analysis we will consider as signal the presence of a charm-tagged jet in
addition to a W boson. We will call “W+b” in case the jet is coming from a b quark,
and “W+udsg” if we have a light quark, either from u quark, d quark, s quark or a
gluon.
6.1 Samples
The data sample used for this measurement is the sample employed for the W cross
section determination. It was already presented in the previous Chapter.
Large MC samples of the signal and main background processes are generated.
They are used for validation of the employed methods and various checks. They are
reported in Tab. 6.1.
W signal (W→ µν) as well as other EWK processes (such as Z→ µµ, W→ τ ν and
Z→ τ τ production) are generated with POWHEG (42). No cuts were applied in phase
space at the generator level, except for the Z/γ ∗ case (Mµµ > 20 GeV). The PDF set
used in these POWHEG productions was CT10.
The analysis requires at least one jet in the final state and sets a rather loose cut
on the maximum number of hard jets. Therefore a POWHEG approach, which should
correctly describe the inclusive production of W plus one hard jet (plus collinear/soft
QCD) is suitable for this analysis. The cc̄ and bb̄ contributions are small and originate
mostly from gluon splitting. There are also higher order processes like bq → bWq′
than compete with gluon splitting, but both contributions are included in PYTHIA
via parton evolution. An important advantage of the POWHEG approach (a NLO
generator) is the possibility to perform detailed comparisons with NLO PDFs, which
give a more accurate prediction.
178
6.1 Samples
Generator Process σ (pb) Events

+ +
POWHEG W → µ νµ 6152 ∼2M
POWHEG W− → µ− ν̄µ 4286 ∼2M
POWHEG W+ → τ + ντ 6152 ∼2M
POWHEG W− → τ − ν 4286 ∼2M
POWHEG Z → µ + µ− 1666 ∼ 2M
POWHEG Z → τ +τ − 1666 ∼ 2M
MADGRAPH tt̄ 162 ∼1.3M
MADGRAPH Single-top (t-channel) 21.5 ∼0.5M
PYTHIA Inclusive µ QCD 79688 28M
PYTHIA WW 42.9 ∼2M
PYTHIA WZ 18.3 ∼2M
PYTHIA ZZ 5.9 ∼2M
MADGRAPH W + jets 6152 ∼15M
Table 6.1: Summary of analyzed Monte Carlo samples for the various signal and back-
ground processes. Reported cross sections correspond to NNLO FEWZ MSTW08NNLO
for W and Drell-Yan cases and to NLO CTEQ66 cross sections for the remaining back-
grounds. No cuts have been applied at the generator level whenever is possible. Exceptions
(QCD, processes with photon exchange) are discussed in the text.
We also use an alternative MadGraph LO approach in the analysis. This allow us to

cross-check the POWHEG approach and evaluate explicitly any potential dependence
on the details of hard jet production or on the description of heavy-quark contributions
associated with higher jet multiplicities in the final state.
Important backgrounds in this analysis are tt̄ and single top (the latter being fully
dominated by the t-channel bq → tq′ component). Event samples of both are generated
with MadGraph (84) interfaced with PYTHIA. QCD (p̂T > 20 GeV, pµT > 15 GeV
and diboson backgrounds (WW, WZ, ZZ) are generated with PYTHIA. All PYTHIA
generations use the Z2 tune (85) to model the underlying event characteristics. All
PYTHIA LO generations use CTEQ6L PDFs.
For the signal (W → µν) and tt̄ backgrounds, we have considered simulations that
take into account the presence of pileup events (∼ 2 − 3 minimum bias interactions on
top of the hard process) to match the experimental conditions.
Details on the Monte Carlo samples used in this analysis are given in Tab. 6.1.
179
WITH CHARMED JETS
6.2 Selection
The selection of W+c events is aimed to select a pure sample of W candidates, contain-
ing at least a high pT jet in the event within detector acceptance, removing as much as
possible the background arising from other processes delivering events with a similar
topology in the final state.
The basic selection consist in finding a W candidate, using the selection described
in the previous Chapter. This set of requirements was shown to be optimal in terms of
muon reconstruction and signal extraction.
We require some minimal quality criteria to preselect muon candidates using tracks
traversing the muon and inner tracking volumes: the muon must be reconstructed by
two different reconstruction algorithms: the muon track must contain one or more pixel
hits; the number of hits used in the inner tracker (pixel+strips) must be greater than
10 to ensure an accurate pT computation; the transverse impact parameter of the muon
with respect to the beam should be smaller than 2 mm, to reject cosmic background;
the χ2 per degree of freedom of the global muon fit (system-inner tracker) must be less
than 10; at least one hit in the muon spectrometer must be included in the final refit
and at least two muon stations must be used in the global muon fit.
In addition, some specific criteria to select clean W → µν candidates are applied:
• the transverse momentum of the muon track must be larger than 25 GeV,
• and the muon lie in the region | ηµ |< 2.1.
• less than two muons with pµT > 10 GeV, in order to reduce the Drell-Yan con-
tamination,
• the relative combined isolation variable, built from the sum of ECAL, HCAL and
inner tracker transverse energies or momenta in a ∆R < 0.3 cone,
X
rel
Icomb = (ET (ECAL) + ET (HCAL) + pT (tracks))/pT (µ),
rel
must satisfy Icomb < 0.1,
• the missing transverse mass, MT , built from pµT and the missing transverse energy,
miss , must be larger than 50 GeV,
ET
q
MT ≡ 2 pµT ET miss (1 − cos(φ − φ
µ E miss )) > 50 GeV.
T
180
6.2 Selection
Compared with the optimized in the previous Chapter selection we have added the
MT > 50 GeV requirement, which strongly suppresses the contribution of the QCD
background. Given the nature of the present measurement - we are measuring ratios
- and the reasonable agreement between data and simulations for the QCD shape
background, this cut allows almost full suppression of the QCD background with a
negligible impact on the systematic uncertainty of the measurement.
Jets in the studies presented here are reconstructed using the anti-kT (86) clustering
algorithm with the size parameter R = 0.5. Jets in CMS are reconstructed using the
particle flow technique (87, 88).
In addition to the W → µν selection just described, a W + jets sample is selected
in the following way:
• at least one jet with pjet jet

T > 20 GeV and |η | < 2.1, separated from the muon
from the W decay by ∆R > 0.3,
• less than three jets with pjet jet

T > 40 GeV and |η | < 2.1, in order to reduce top
backgrounds.
After this selection, the most important processes that has a similar signature to
that of the signal are:
• tt̄ events. t quarks decay almost exclusively in a W boson plus a b quark. In the
case of a leptonic decay of the W, these events are characterized by two high pT
leptons and two b jets. In the hadronic decay, it could contain up to 6 jets of
different flavour. At least two of these jets contain a b-quark (long decay length,
high invariant mass, not as collimated as a light jet). tt̄ background is partially
removed by restricting to events with a reduced number of energetic jets.
• Single-top events. This channel is dominated by the t-channel production (bq → tq′ ).
Thus, it is characterized by a t quark accompanied by another quark in the final
state. The t quark decays into a b and a W (as shown before). As a consequence,
this channel is characterized by a W and two quark jets, one of them containing
a b quark. Setting strong constraints on the number of jets or in the mass of the
most energetic jet reduce the contribution coming from this process.
181
WITH CHARMED JETS
• W+b(b̄) events. This irreducible background has a similar signature to that of

the signal. The properties that we can exploit to distinguish it from the signal
were shown before and uses the properties of the massive b quark jet.
• W+light-jet events. This is the dominant background. Heavy flavour tagging

techniques, exploiting specially the flight distance, allow us to separate this pro-
cess from the signal.
• QCD multijets events. This huge background is largely removed by the W selec-
tion described in Chapter 6.Setting an additional requirement on the transverse
mass, MT , of the MET-muon system to be greater than 50 GeV reduces this
background to a negligible level.
• other electroweak processes (such as Drell-Yan or dibosons). The basic selection

(W candidate with a high pT jet) already reduces this contribution to reasonable
levels, not requiring dedicated selection requirements.
The two first backgrounds (tt̄ and single-top) are reduced by a loose cut on the
number of high energy jets. The remaining top background will be separated from the
signal using the long flight distance that b quark jets have. In Fig. 6.5 the number
of jets after the W selection for data, W+c, single-top and tt̄ from MC simulation is
shown. The high jet multiplicity of tt̄ events can be noticed in this plot. To remove
this background we veto events with three or more selected jets with energy over 40
GeV.
No additional criteria are applied to reduce the W+b background due to its simi-
larity to the signal. We will rely on the MC prediction and assume a 100% uncertainty
in the W+b normalization.
W+light jets background is the main contributor to the sample. In order to distin-
guish it from the W+c signal, b-tagging techniques are used, as it will be explained in
the section devoted to the method developed.
With this selection criteria, we select 31730 events. As evaluated by MC, the non
W+jets background amounts to 5.3%, and it is composed by Drell-Yan (2.1%), W → τ ν
(1.4%) and top events (1.4%). Subtracting these backgrounds according to the Monte
Carlo expectations we obtain a corrected W+jets yield of:
N (W + jets; pjet jet

T > 20 GeV, |η | < 2.1) = 30038.0 ± 178.1 (stat.)
182
6.2 Selection
107
Arbitrary units
Data
106 W+c
Single-top
105 tt
104
103
102
10
0 2 4 6 8 10 12 14
Number of jets
Figure 6.5: Number of jets in the event (within pseudorapidy and pT acceptance defined
in this analysis) after the W selection.
where the uncertainty is purely statistical.

Fig. 6.6 shows the MT distribution for W + ≥ 1 jet selected events, separately for
W+ and W− . Good agreement between total MC prediction and data is observed.
Fig. 6.7 shows the transverse momentum distribution of the most energetic jet in the
event compared with POWHEG and MadGraph references. They are in agreement
with expectations (MadGraph predictions are more accurate in the high-pT tail). Sim-
ilar conclusions can be extracted from the jet multiplicity distributions in Figs. 6.8.
POWHEG correctly predicts the kinematics in the events with 1 or 2 jets, failing at
high multiplicities.
6.2.1 Definition of W+c, W+b and W+ligh quark components at the

generator level
The definition of the jet flavour is an important part in the analysis to correctly define
the measured ratio. W plus jets events in CMS usually contain quarks of different
flavours at the same time. The tagging of jet as a certain flavour jet will depend on
our definition, but the result will be independent of such a definition provided we stay
coherent with this definition through the whole analysis.
The definition adopted in this analysis consider an event in Monte Carlo as “W+b”
183
WITH CHARMED JETS
1000
number of events/2 GeV

1200
900
1000 Data 800 Data
-
W++charm 700 W +charm
+ -
800 W +light W +light
600
top top
600 Other bckg. 500 Other bckg.
400
400 300
200
200
100
0 0
60 80 100 120 140 160 60 80 100 120 140 160
MT [GeV] MT [GeV]
Figure 6.6: Distribution of the transverse mass in selected W+ +jets and W− +jets events
compared with POWHEG predictions after the selection. In red the W+c signal, in green
the W+light quark contribution, in purple the top contribution and in blue the sum of the
contributions of the other backgrounds. Histograms are stacked.
POWHEG MADGRAPH
Data Data
103 W++charm 103 W++charm

W++light W++light
top top
Other bckg. Other bckg.
102 102
10 10
1 1
50 100 150 200 50 100 150 200
pT of most significant jet [GeV] pT of most significant jet [GeV]
Figure 6.7: Distribution of the transverse energy of the most significant jet (pjetT > 20
GeV) in selected W+jets events compared with POWHEG (left) and MadGraph predic-
tions (right) or the signal. In red the W+c signal, in green the W+light quark contribution,
in purple the top contribution and in blue the sum of the contribution of the other back-
grounds. Histograms are stacked.
184
6.2 Selection
POWHEG MADGRAPH
Number of events
Number of events
105 105
Data Data
W++charm W++charm
104 W++light 104 W++light
top top
103 Other bckg. 103 Other bckg.
102 102
10 10
1 1
0 5 10 0 5 10
Number of jets in the event Number of jets in the event
Figure 6.8: Distribution of the number of jets with pjet +

T > 20 GeV in selected W +jets
events compared with POWHEG (top) and MadGraph predictions (bottom). In red the
W+c signal, in green the W+light quark contribution, in purple the top contribution and
in blue the sum of the contribution of the other backgrounds. Histograms are stacked.
if it has a b (or b̄) entry at generator level with pT > 10 GeV and |η| < 2.5, “W+c” if
it has a c (or c̄) entry with pT > 10 GeV and |η| < 2.5, and “W+udsg” otherwise. This
is a way to ensure that tagged jets with pT > 20 GeV and |η| < 2.1 (the cuts that will
be used in the analysis described in the next sections) are taken into account properly
as heavy quarks despite fluctuations in energy and angle with respect to the generated
values. The definition also ensures that there is a reasonable transverse momentum for
a charm candidate to be considered for tagging purposes. We avoid for instance calling
W+c an event that has a charm entry with almost zero pT and one additional hard gluon
jet. Accepting an event like this after lifetime tagging cuts is unlikely, but considering
it as a charm event would be detrimental for the analysis because it would contribute
to the tagging region dominated by light quarks. Let us note that this convention is
mostly a matter of classification. Experimental cuts ensure that the selected events are
constrained kinematically to the region where jets have pT > 20 GeV and |η| < 2.1.
Changing the definition of the W components modifies minimally the results of the
fitted charm fractions and should not produce any bias in the final measurements,
simply because the efficiencies used in ratios is corrected according to our definition.
185
WITH CHARMED JETS
This classification is applied in Fig. 6.6, Fig. 6.7 and Fig. 6.8 to differenciate between
the contributions. A clear W+c component can be seen in these Figs.
6.2.2 W+charm extraction
Heavy flavour jets (with c or b quarks) are characterized by a long flight distance.
In order to enhance the W + charm content in the sample we require the presence
of a secondary vertex with two or more associated tracks, coming from the decay of
the charmed or b-hadron, significantly displaced from the primary interaction vertex,
following the “simple secondary vertex high-efficiency” (SSVHE) algorithm of CMS (89)
(in the HE, High Efficiency, version we require at least 2 tracks in the vertex, as opposed
to the HP, High Purity, version with a minimum of 3 tracks in the vertex). Only vertices
with decay length uncertainties smaller than 1.5 mm are used.
A lifetime tagging requirement significantly enhances the fraction of W+charm
events in the final state. The presence of W+b and top backgrounds is also enhanced
but, due to absence of a significant b component in W+jets subsample and the mod-
erate top backgrounds, the W+c signal is clearly dominant after this last requirement.
Instead of using directly the decay length, the following discriminant D is defined as
D = sign(S) log(1 + abs(S)), where S is the decay length significance. The logarithmic
function is used in order to have more sensitivity with lower values of the discriminator
(light contribution). Jets without a secondary vertex or not significant enough (an in-
trinsic requirement of ≈ 1 is set on the D value of the jet secondary vertex) do not enter
in the plot. Negative vertices are also included in the analysis. The (small) number
of events with negative vertices is due to the tails of tracker resolution (and possibly
jet angular resolution). The definition of a positive and negative vertex are shown in
Fig. 6.10 and Fig. 6.11 respectively. Negative vertices are built using the same criteria
and cuts as the positive ones, but for cases in which the vector connecting primary
and secondary vertex is opposite to the jet direction. Fig. 6.9 shows the distribution
of the lifetime discriminant variable for the most significant (the one with largest value
of the discriminant) jet in W+ → µ+ νµ ≥ +1 jet and W− → µ− ν̄µ ≥ +1 jet selected
samples. A clear W+charm component is observed, in good agreement with Monte
Carlo predictions.
186
6.2 Selection
Number of events
90 Data
80 W++charm
W++light
70
top
60 Other bckg.
50
40
30
20
10
0
-6 -4 -2 0 2 4 6
DSSVHE
Number of events
80 Data
-
W +charm
70 -
W +light
60 top
Other bckg.
50
40
30
20
10
0
-6 -4 -2 0 2 4 6
DSSVHE
Figure 6.9: SSVHE discriminator in selected W+ → µ+ νµ + ≥ 1 jet (top) and

W− → µ− ν̄µ + ≥ 1 jet (bottom). In red the W+c signal, in green the W+light quark
contribution, in purple the top contribution and in blue the sum of the contribution of the
other backgrounds. Histograms are stacked.
187
WITH CHARMED JETS
Figure 6.10: An example of a positive secondary vertex.
Figure 6.11: An example of a negative secondary vertex.
188
6.3 Measurement of the σ(W+ + charm)/σ(W− + charm) ratio
≥ 1 jet < 3 jets with b-tag ∆decay−length

Sample Muon MT > 50 GeV pjet
T > 20 GeV, pjet
T > 40 GeV SSVHE < 0.15 cm
& DY veto jet
|η | < 2.1
Events with µ± 195743 136008 32185 31735 1421 1199
Events with µ+ 112292 80259 18237 18010 768 636
Events with µ− 83451 55749 13948 13725 653 563
Table 6.2: Data reduction at every step of the selection process. The number of events is
given for the whole muon data sample, and also separated by the muon charge. The Muon
rel
requirement means a good quality and isolated (Icomb < 0.10) muon with pT > 25 GeV
and |η| < 2.1, according to the criteria described in the text.
Tabs. 6.2, 6.3 and 6.4 show the cumulative evolution of the number of selected
events as a function of the applied cuts in data and MC. The final sample composition
is detailed in 6.5.
6.3 Measurement of the σ(W+ +charm)/σ(W− +charm) ratio
The W+charm signal is extracted from a maximum likelihood fit to the different com-
ponents of the high-efficiency simple secondary vertex b-tagging variable (SSVHE)
distribution.
As reported in the previous section, after the selection process the dominant back-
grounds with positive discriminator to W+c are single-top and tt̄ with a small contribu-
tion from W plus light-quark/gluon jets. Consequently we perform a fit to determine
the normalization of the three components: 1) W+c signal, 2) top backgrounds, 3)
light-jets. Shapes are taken from MC. The contribution from the remaining compo-
nents, including the W+b, is taken from simulation and added to a fourth category
“Others”. Fig. 6.12 shows the DSSV HE distribution of the different components nor-
malized to the MC expectations. Their shapes are rather different, thus justifying the
approach of fitting these three main contributions.
In order to be rather insensitive to the light-jet component, the fit takes also into
account negative SSVHE discriminators. Including them in the fit is an effective way
to constrain the number of positive vertices from light-quark/gluon sources. We note
that the amount of negative light tags is slightly smaller than the amount of positive
189
WITH CHARMED JETS

T > 20 GeV, pjet
& DY veto jet
|η | < 2.1
W+ + c 4560 ± 22 3931 ± 20 2166 ± 15 2159 ± 15 237 ± 5 210 ± 5
W+ + udsg 81406 ± 73 73371 ± 71 14458 ± 38 14442 ± 38 187 ± 4 127 ± 4
W+ + b 295 ± 6 249 ± 5 137 ± 4 136 ± 4 23 ± 2 21 ± 1
tt̄ 382 ± 1 203 ± 1 202 ± 1 99 ± 1 69 ± 1 66 ± 1
single-t 90 ± 1 61 ± 1 57 ± 1 53 ± 0 32 ± 0 31 ± 1
W → τ + ντ
+ 2551 ± 16 1563 ± 13 249 ± 5 249 ± 5 7±1 6±1
Z → µ+ µ− 16546 ± 18 2205 ± 8 351 ± 3 350 ± 3 12 ± 1 9±1
Z → τ +τ − 492 ± 4 75 ± 1 24 ± 1 23 ± 1 1±1 1±1
QCD 4166 ± 20 72 ± 3 54 ± 2 52 ± 2 6±1 5±1
Total 110487 ± 82 81730 ± 76 17698 ± 41 17563 ± 41 573 ± 7 475 ± 6
Table 6.3: Expected number of events after every step of the selection criteria, for the
several physics processes producing a signal as defined in Section 4. Projection for an
integrated luminosity of Lint = 36 pb−1 .

T > 20 GeV, pjet
& DY veto jet
|η | < 2.1
W− + c 4583 ± 18 3949 ± 17 2191 ± 13 2187 ± 13 243 ± 4 212 ± 4
W− + udsg 53720 ± 50 49337 ± 49 10197 ± 26 10186 ± 26 136 ± 3 93 ± 3
W− + b 207 ± 4 178 ± 4 96 ± 3 95 ± 3 18 ± 1 17 ± 1
tt̄ 383 ± 1 205 ± 1 205 ± 1 100 ± 1 70 ± 1 67 ± 1
single-t 142 ± 1 97 ± 1 92 ± 1 84 ± 1 52 ± 1 49 ± 1
single-t 47 ± 1 32 ± 1 31 ± 1 28 ± 1 17 ± 1 16 ± 1
W− → τ − ν 1880 ± 12 1140 ± 9 188 ± 4 187 ± 4 8±1 6±1
Z → µ+ µ− 16175 ± 18 1924 ± 7 310 ± 3 309 ± 3 11 ± 1 8±1
Z → τ +τ − 467 ± 4 70 ± 1 23 ± 1 23 ± 1 1±1 1±1
QCD 3989 ± 20 65 ± 3 48 ± 2 47 ± 2 5±1 4±1
Total 81451 ± 61 56901 ± 53 13288 ± 30 13163 ± 30 509 ± 6 424 ± 5
Table 6.4: Expected number of events after every step of the selection criteria, for the
several physics processes producing a signal as defined in Section 4. Projection for an
integrated luminosity of Lint = 36 pb−1 .
190
Source Nbg /(Nsg + Nbg ) Nbg in 36 pb−1

W+ udsg 24.5 ± 0.6% 220 ± 4
W+ b 4.1 ± 0.2% 37 ± 2
tt̄ 14.8 ± 0.2% 133 ± 1
single-t 5.2 ± 0.1% 47 ± 1
W → τν 1.3 ± 0.1% 12 ± 1
Z → µ+ µ− 1.9 ± 0.1% 17 ± 1
Z → τ +τ − 0.1 ± 0.1% 1±1
QCD 1.1 ± 0.1% 10 ± 1
Total Bckg. 53.0 ± 0.7% 476 ± 5
W + c signal 47.0 ± 0.8% 423 ± 6
Table 6.5: Final sample composition from MC predictions. The second column shows
the fraction of each of the background processes that contributes to the candidate sample.
The third column gives the number of events expected of each type in a 36 pb−1 sample,
based on the theoretical predicted cross sections. The last row gives the prediction for the
signal process.
number of events/0.24
90 90
80 80
Data Data
70 MC (all) 70 MC (all)
-
W++charm W +charm
60 W++light 60 -
W +light
top top
50 50
40 40
30 30
20 20
10 10
0 0
-6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6
DSSVHE DSSVHE
Figure 6.12: SSVHE discriminant distributions of the main components of the enriched
W+c sample (separately for W+ , left, and W− , right), normalized to the expected lumi-
nosity and taking into account the cross sections predicted by Monte Carlo. In red the
in blue the sum of the contributions of the other backgrounds.
191
WITH CHARMED JETS
# events
16
14 tt
Single Top
12
10
0
-6 -4 -2 0 2 4 6
DSSVHE
Figure 6.13: Comparison between single-top and tt̄ templates in Monte Carlo. The
normalization in this plot is arbitrary.
light tags, due to remaining long-lived K0S and Λ decays.

The shape of the discriminant variable for single-top and tt̄ background events is
similar (Fig. 6.13), and it is taken from Monte Carlo. As a consequence, we can add
both contributions and take them in the fit as a single component (top background
contribution). Due to the dominance of single-top in t-channel, top backgrounds will
be larger in W+ + charm than in W− + charm (in absolute value).
The results of the fits to the W+ + charm and W− + charm distributions are shown
in Fig. 6.14. We obtain the following yields:
N (W+ + charm) = 247.4 ± 40.2 (stat.)

N (W− + charm) = 268.9 ± 38.5 (stat.)
Dividing both numbers, we get the following measurement of the W + charm charge
ratio:
Rc± (pjet jet

T > 20 GeV, |η | < 2.1) = 0.92 ± 0.20 (stat.)
Possible sources of systematic uncertainties have small contributions to this ratio. They
are discussed in Section 6.5.
192
90
80
Data
70 W++charm
60 W++light
top
50
Other bckg.
40
30
20
10
0
-6 -4 -2 0 2 4 6
DSSVHE
90
80
Data
70 -
W +charm
-
60 W +light
top
50
Other bckg.
40
30
20
10
0
-6 -4 -2 0 2 4 6
DSSVHE
Figure 6.14: Top: fit to the SSVHE discriminator of the W+ + charm selected sample.
Bottom: fit to the SSVHE discriminator of the W− + charm selected sample. In red the
in blue the sum of the other contributions of the backgrounds. Histograms are stacked.
193
WITH CHARMED JETS
6.4 Measurement of the σ(W + charm)/σ(W + jets) ratio
The yields obtained in the previous section and the overall W+jets yields obtained
before are combined to obtain a measurement of the W+charm over W+jets ratio for
pjet
T > 20 GeV. We use the expression:
N (W + + charm) + N (W − + charm)
Rc (pjet jet
T > 20 GeV, |η | < 2.1) =
ǫc N (W + jets)
where ǫc is a correction factor determined in Monte Carlo as:
ǫc ≡ Nselected and tagged jets /Ntotal selected jets ,
according to the phase space definition for the theoretical prediction introduced in
Section 6.2.1. This correction factor coincides with the SSVHE tagging efficiency in the
limit in which all other selection criteria have the same effect on W+c and W+jets and
both samples have similar kinematic properties. Effectively, ǫc also absorbs remaining
kinematic effects and also acceptance corrections due to the mismatch between the
reconstructed jet parameters and the properties of the parton that initiates the jet.
Given the reasonably accurate description of the tracker response in CMS simulations
the central value for this efficiency is taken from Monte Carlo with an uncertainty
determined from dedicated studies (section 6.5).
Using the Monte Carlo estimate ǫc = 0.1218 ± 0.0019 (stat.) we obtain:
Rc (pjet jet
T > 20 GeV, |η | < 2.1) = 0.141 ± 0.015 (stat.)
where the uncertainty is statistical.
6.5 Systematic uncertainties
All the systematic uncertainties are computed by comparing the results shown before
with the ratios computed by applying the modifications suggested in each of the system-
atic errors considered. The following sources of systematic uncertainty are considered
in the analysis:
• Difference in efficiency between µ+ and µ− . This source is only relevant

for the Rc± analysis. It has been computed in the CMS W → µν analysis and
amounts to ≤ 1%.
194
Figure 6.15: Distribution of the number of primary vertices in data and MC simulation
used.
• Muon momentum scale and resolution. This is computed by distorting the

muon resolution in Monte Carlo in order to match the performance found in data
following the procedure shown in Chapter 4 (SIDRA).
• Pileup. The simulated samples used do not fully reproduce the real pileup con-
ditions of the 2010 run. MC distributions are modified reweighting the events
according to the differences between data and pileup simulation for the distribu-
tion of the number of primary vertices (Fig. 6.15).
• Jet resolution and jet scale. We change the jet scale resolution by 3%, as
indicated by dedicated jet resolution studies (90). In the case of heavy quarks the
uncertainty is increased to 8% to account for unknown flavor-specific differences.
This should also cover possible uncertainties on the jet resolution (a fraction of
∼ 15% (91)) .
• Jet multiplicity. The POWHEG Monte Carlo is not expected to reproduce so

accurately the final states with high hard jet multiplicity (≥ 2). The analysis only
uses the lifetime tag of the most significant jet and signal events with two c or
b-jets (via gluon splitting) are also reasonably well described. Therefore a small
195
WITH CHARMED JETS
uncertainty due to this source is expected. It is anyhow estimated by reweighting

W+jet events in the POWHEG Monte Carlo in order to match the jet multiplicity
distribution observed in MadGraph (Fig. 6.8). This is done independently for light
and heavy quarks.
• Tracking resolution. Fig. 6.16 shows the distribution of the number of tracks
associated to the secondary vertex in the W+c selected sample. There is rea-
sonable agreement with the MC predictions. In order to check the sensitivity to
differences in tracking reconstruction between data and MC, we distort the MC
predictions in order to provide the optimal agreement with data. We define a
probability P for a track to be missed on top of the MC vertex reconstruction.
This probability is defined uniformly, i.e. we assume that the probability to miss
a track is independent of the number of charged particles originally produced at
the vertex, the decay length and the uncertainty on the decay length. With this
scheme there is a NV × P i × (1 − P )(NV −i) probability for a MC vertex with NV
to become a MC vertex with (NV − i) tracks. The resulting MC distribution is
normalized to the total number of vertex entries in data and a χ2 of the differences
between data and the distorted MC predictions (for the vertex track multiplicity
distribution) is defined. We found an optimal matching for P = 8% and use the
corresponding distorted MC to assign a systematic uncertainty due to tracking
resolution. Fig. 6.16 also shows a more extreme case (P = 14%) which is clearly
inconsistent with the observed distribution in data.
• Uncertainties in top templates. This is estimated as the difference in the

resulting ratios using a data-driven template and the top Monte Carlo template.
The data-driven template is built using a top-enriched control sample. We use the
same selection but inverting the requirement on the number of high energy jets
(i.e. the control sample is composed of events with three or more jets with pjet
T >
40 GeV). Fig. 6.17 shows a comparison of the data-driven templates obtained in
data and in the top Monte Carlo, together with the expected Monte Carlo shape
after the full W+c selection. While the MC control sample shows a slightly
larger tagging discriminator that the MC in average, due to the presence of more
b-jets in average, the data control samples has a softer discriminator spectrum,
due to a remaining W+jet background. The maximum difference between the
196
0.6 1
Data, L=36 pb-1 Data, L=36 pb-1
MC, p=0 (reference) MC, p=0 (reference)
0.5
MC, p=8% MC, p=8%
MC, p=14% MC, p=14%
0.4 10-1
0.3
10-2
0.2
0.1
10-3
0
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Number of tracks at vertex Number of tracks at vertex
Figure 6.16: Distribution of the number of tracks associated to the secondary vertex for
the W + charm selected events both in data (points) and MC (blue solid line), in linear
(left) and logarithmic (right) scales. The MC prediction is modified including an additional
probability to miss a track in the association to a vertex. The cases P = 8% and P = 14%
to miss a track are shown. The best agreement with data is reached with P = 8%.
results obtained with Monte Carlo and the results obtained with the data-driven
template is conservatively considered as the systematics due to the modeling of
top backgrounds.
• Uncertainty in the light-quark contribution. This is conservatively esti-

mated by assuming that the excess of positive light-quark tags with respective
to negative light-quark tags has a 20% uncertainty. This variation absorbs effec-
tively discrepancies in the number of K0S , Λ and conversions observed in dedicated
CMS studies (92).
• W+b contribution. The W+b contribution is changed by 50%, covering the

variations observed between PYTHIA and MadGraph predictions for that com-
ponent.
• Other backgrounds taken from MC. We estimate a 10% variation in the

Drell-Yan backgrounds and a 50% variation in the QCD background (1.5 is the
typical size of the k-factor that has to be applied to the LO predictions from
197
WITH CHARMED JETS
arbitrary units
0.18 Data, N
jets
≥3
0.16 MC (t+X,t +X), N
jets
≥3
jets
0.14 MC (t+X,t +X), N <3
0.12
0.1
0.08
0.06
0.04
0.02
0
0 1 2 3 4 5
DSSVHE
Figure 6.17: Comparison of data-driven templates for top background in data and Monte
Carlo. They are also compared with the Monte Carlo truth expectations after the selection
process. All templates are normalized to one.
PYTHIA in order to get agreement with the high-pT inclusive rate observed in
multijet events, as seen in the previous Chapter).
• PDF uncertainties. Since it is observed that the uncertainties related with

strange PDFs within the same PDF set are smaller or equal than the differ-
ences between the central values obtained with CT10 (93), MSTW08 (94) and
NNPDF21 (95), we assign as PDF systematic uncertainty due to the incomplete
knowledge of the PDFs the largest difference obtained using either MSTW08 or
CT10 in the case of the charged ratio Rc± , NNPDF21 or CT10 for Rc .
• Average energy fraction in charm fragmentation. We change the average

energy fraction carried by the D∗ (or D if no excited-D is present) by 2%, which
safely covers current estimates for this parameter from previous LEP experiments.
• Change in charm fragmentation fractions BR(c → D0 , D± , Ds , Λc ). We

change them according to the uncertainties reported in (96). The size of this
uncertainty is also consistent with the differences observed between the values
present in the CMS Monte Carlo and the central values from reference (96).
198
6.6 Additional tests
Table 6.6: Relative systematic uncertainties (%) in the measurement of Rc± .
Source Relative systematics (%)

Charge asymmetry in efficiency 1.0
Muon resolution <0.1
Pile-up effects 1.8
Jet scale/resolution 1.1
Jet multiplicity 0.7
Tracking resolution 0.3
Top templates in fit 1.4
Light-quark contribution in fit 1.1
W+b background 0.2
Other Monte Carlo backgrounds 1.4
PDF uncertainties 2.2
Charm fragmentation function <0.1
Charm fragmentation BRs 0.1
TOTAL 4.0
The breakdown of the different systematic contributions to the Rc and Rc± mea-
surements is shown in Tabs. 6.6 and 6.7, respectively.
The precision achieved for Rc± (a 4% uncertainty) is already at a low level. The
larger uncertainty comes from the PDFs. For the Rc value, the total uncertainty
amounts to 21%, being driven by the tracking uncertainties and the top background
modelling.
In order to tests the robustness of the measurement, we have repeated the analysis
in two different ways. First we have used a different tagging variable (track-counting
high-efficiency). Then we also compare the reference method with the results obtained
when using MadGraph for the W+jets simulation.
199
WITH CHARMED JETS
Table 6.7: Relative systematic uncertainties (%) in the measurement of Rc .
Source Relative systematics (%)

Muon resolution 0.7
Pile-up effects 2.5
Jet scale/resolution 2.3
Jet multiplicity 2.5
Tracking resolution 14.1
Top templates in fit 14.5
Light-quark contribution in fit 3.3
W+b background 2.4
Other Monte Carlo backgrounds 0.2
PDF uncertainties 0.2
Charm fragmentation function 0.2
Charm fragmentation BRs 0.2
TOTAL 21.1
6.6.1 Measurements with an alternative lifetime tagging variable
As a cross check of the analysis, the measurement is also done employing an alter-
native lifetime discriminator, the so called “track counting high-efficiency” (TCHE)
discriminant (89). In this case the discriminant variable is defined as the signed impact
parameter significance of the second most significant track in the jet. The W+ → µ+ νµ
and W− → µ− ν̄µ distributions for this variable are shown in Fig. 6.18.
The study follows exactly the same logic used in previous sections, but in this
case the fit is restricted to the region DT CHE > 3 and DT CHE < 20, in order to
suppress the light-quark contribution to almost negligible levels. Actually, the light-
quark contribution is fixed in the fit (so it does not contribute to increase the statistical
uncertainty of the fit). We obtain the following yields:
N (W+ + charm) = 290.8 ± 33.0 (stat.)

N (W− + charm) = 298.8 ± 30.2 (stat.)
200
Number of events
Number of events
Data Data
+ -
103 W +charm 103 W +charm
-
W++light W +light
top top
Other bckg. 2 Other bckg.
102 10
10 10
1 1
-20 -10 0 10 20 -20 -10 0 10 20

DTCHE DTCHE
Figure 6.18: TCHE discriminator in selected W+ → µ+ νµ + ≥ 1jet (left) and W− →

µ− ν̄µ + ≥ 1jet (right) events. In red the W+c signal, in green the W+light quark contri-
bution, in purple the top contribution and in blue the sum of the other contribution of the
other backgrounds. Histograms are stacked. The W+charm signal populates the region
1 < discriminant < 10.
201
WITH CHARMED JETS
ratio:
Rc± (pjet jet

T > 20 GeV, |η | < 2.1) = 0.97 ± 0.15 (stat.)
Similarly for the σ(W + charm)/σ(W + jets) ratio:
Rc (pjet jet
T > 20 GeV, |η | < 2.1) = 0.164 ± 0.012 (stat.)
where in this case the charm efficiency corresponding to the region 3 < DT CHE < 20
is ǫc = 0.1200 ± 0.0018 (stat.).
Both measurements are consistent with the measurements obtained using the SSVHE
discriminator, even if no dedicated evaluation of systematic uncertainties for the TCHE
method has been performed.
6.6.2 Measurements using a MadGraph W+jets Monte Carlo as ref-

erence
The POWHEG W inclusive sample should be a reliable NLO Monte Carlo reference
for this analysis, since it is accurate up to W + 1 hard jet plus additional jets in the
soft-collinear parton-shower approach. In addition, it contains the relevant cc̄ and bb̄
contributions from to gluon splitting. However, POWHEG is not expected to reproduce
accurately the jet multiplicity and kinematic properties of W + ≥ 2 hard jets in the
final state, but this limitation has a small impact in the analysis because it is based on
the b-tagging of the most significant jet. The jet multiplicity differences between data
and MC are also taken into account as systematic uncertainty.
Nevertheless we have repeated the analysis using a (LO) W+jets MadGraph Monte
Carlo as reference. The results of the fits to the W+ + charm and W− + charm
distributions are shown in Fig. 6.19. We obtain the following yields:
N (W+ + charm) = 254.5 ± 39.5 (stat.)

N (W− + charm) = 253.4 ± 36.6 (stat.)
ratio:
Rc± (pjet jet

T > 20 GeV, |η | < 2.1) = 1.00 ± 0.21 (stat.)
202
6.7 Final results and comparisons with theory
Number of events
Number of events
90 Data 80 Data
-
80 W++charm W +charm
70 -
W++light W +light
70
top 60 top
60 Other bckg. Other bckg.
50
50
40
40
30
30
20 20
10 10
0 0
-6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6
DSSVHE DSSVHE
Figure 6.19: Left: fit to the SSVHE discriminator of the W+ + charm selected sample,
using a W+jets MadGraph reference sample to derive. Right: similar fit to the SSVHE
discriminator of the W− + charm selected sample. In red the W+c signal, in green the
W+light quark contribution, in purple the top contribution and in blue the sum of the
contribution of the other backgrounds. Histograms are stacked.
Similarly for the σ(W + charm)/σ(W + jets) ratio:
Rc (pjet jet
T > 20 GeV, |η | < 2.1) = 0.129 ± 0.014 (stat.)
where the efficiency ǫc = 0.1320 ± 0.0018 (stat.) is found to be consistent - within

expected systematic uncertainties quoted before- with the efficiency derived from the
POWHEG simulation. The Rc and Rc± results thus obtained are consistent with the
POWHEG result within the expected uncertainties.
6.7 Final results and comparisons with theory
In summary, we have measured the ratios Rc± ≡ σ(W+ + charm)/σ(W− + charm) and
Rc ≡ σ(W + charm)/σ(W + jets) for leading jets with pT > 20 GeV and |η| < 2.1 using
36 pb−1 of data collected with the CMS detector in 2010. We obtain:
Rc± = 0.92 ± 0.20 (stat.) ± 0.04 (syst.)

Rc = 0.141 ± 0.015 (stat.) ± 0.030 (syst.)
203
WITH CHARMED JETS
The ratios are for muons from W decays satisfying pµT > 25 GeV and |η µ | < 2.1.
The results can be compared with detailed NLO predictions. Values for different
PDF sets in the kinematic region pjet jet ℓ ℓ
T > 20 GeV, |η | < 2.1, pT > 25 GeV, |η | < 2.1
are shown in Tab. 6.8, where ℓ is the lepton from the W decay. In order to match
a realistic experimental scenario and the CMS generator level conditions (W+charm
is generated at leading order), the MCFM calculation is performed by combining the
charm parton and the potential additional parton from the NLO process into jets
using an anti-kT algorithm (with R = 0.5 parameter). To have a theoretical estimate
closer to our experimental cuts, all MCFM calculations will use the phase space cuts
pjet jet µ µ
T > 20 GeV, |η | < 2.1, pT > 25 GeV, |η | < 2.1.
The theory uncertainties include just the PDF contributions (at 68% CL) and have
been calculated following the established recipes for each set of PDFs. CT10 (93) pre-
dictions for Rc± are larger than the ones from MSTW08 (94) or NNPDF21 (95), beyond
the expected uncertainties (at 68% CL). This is most likely due to the CT10 assumption
of equal PDFs for strange and anti-strange quarks. MSTW08 and NNPDF21 do not as-
sume this constraint and use experimental data (from NuTeV and others) to constrain
both PDFs independently. Concerning Rc , MSTW08 and NNPDF21 predictions have
much smaller uncertainties than CT10. CT10 has more relaxed criteria for the over-
all strange content of the proton, while MSTW08 strongly constrains the strange PDF
shapes to follow anti-u and anti-d shapes (thus constraining Rc in practice). NNPDF21
has many more experimental input measurements to constrain strange PDFs and also
employs a positiveness constraint that reduces the uncertainty further (also in a region
where no data are available, x < 10−2 ).
The results are in agreement with the theoretical prediction given in Tab. 6.8.
Fig. 6.20 shows a comparison of our data measurements with those predictions.
Ratio MCFM (CT10) MCFM (MSTW08) MCFM (NNPDF21)

Rc± 0.993 ± 0.008 0.926 ± 0.028 0.950 ± 0.007
Rc 0.125 ± 0.017 0.115 ± 0.002 0.104 ± 0.005
Table 6.8: Rc± and Rc predictions from MCFM at NLO. Kinematic cuts are: pjet T > 20
jet ℓ ℓ
GeV, |η | < 2.1, pT > 25 GeV, |η | < 2.1. Partons are joined using an anti-kT algorithm
with R = 0.5 parameter. The quoted values correspond to different PDF choices and only
PDF variations (at 68% CL) are considered for the total uncertainties.
204
6.8 Prospects for future improvements
Figure 6.20: Comparison of the CMS measured values for Rc± and Rc with MCFM NLO
predictions for different PDF sets (CT10, MSTW08, NNPDF21). Only PDF theoretical
uncertainties within each PDF set are shown.
As already advanced the discrepancy among the different predictions exceed their
uncertainties and a precise experimental measurement would help to clear the situation.
These measurements are a first step towards a precise understanding of the strange
and anti-strange parton distribution functions of the proton at the electroweak scale
probed by the LHC. With more data incoming, this analysis opens new oportunities to
improve our knowledge on the strangeness composition in the proton.
With 5 fb−1 recorded in 2011 run the amount of data is already large enough to
place strong constraints on the strange composition in the proton. A simulation study
performed on the differential measurement of the Rc parameter with jet pseudorapidity
with this amount of data indicates that it already gives valuable information on the
strange composition. In Fig. 6.21 we can appreciate the improvement on the strange
quark component of the NNPDF PDF set when the information coming from a W+c
measurement at LHC, with 5 fb−1 data is included. The information provided by
this extended analysis would not only increase our knowledge on the strange composi-
tion, but also would allow the NNPDF group to use only collision data (from HERA,
205
WITH CHARMED JETS
Figure 6.21: Left: Improvement on the strange quark component for different eta bins
for the NNPDF PDF set with and without data coming from LHC W+c study. Right:
Improvement on the strange quark densities in the proton for different x parameter values,
with current NNPDF PDF set and the new one (including the results from a W+c analysis
with 5 fb−1 LHC data) (82).
Tevatron and now LHC) to perform the fits. This latter advantage would be a major
improvement since older data coming from neutrino experiments would not be nec-
essary anymore. Including neutrino experiment data entails the modeling of nuclear
corrections for deuterium or heavy nuclei and bigger pQCD uncertainties from higher
scales are involved.
The improvements proposed previously imply a challenging analysis in the W+c
channel. The simulation shown in Fig. 6.21 requires the total experimental error on
the Rc parameter to be of the order of 5%. In the analysis presented in this Chapter
on the Rc parameter the error is estimated to be of the order of 20%, way too far
from the precision required by the PDF groups. With a bigger amount of data, new
techniques can be used to reduce the systematic errors associated to the measurement.
In particular, the tracking error and the top template, which amount each to 14%, can
be reduced.
The top template error can be decreased by reducing the top related backgrounds
to the minimum. More stringent requirements on the number of jets in the event (i.e.
setting to only one the number of jets with energy bigger than 40 GeV allowed in the
event) would strongly reduce the amount of top background in the sample. By imposing
a more restrictive requirement on the number of jets we reduce the tt̄ background by
a 75%, 30% the single-top background and only 5% the signal. MadGraph MC should
206
number of events/0.2 GeV

104 Data 104 Data
W+charm
W+charm
W+light
W+light
103 W+b
103 W+b
top
top
Other bckg.
Other bckg.
102 102
10 10
1 1
0 100 200 300 400 500 600 0 2 4 6 8 10

MHT [GeV] Secondary vertex mass [GeV]
Figure 6.22: MTH distribution (left) and secondary vertex mass distribution (right) for the
2011 dataset. Thresholds at 300 GeV and 2 GeV (respectively) are suggested to remove the
top related background. In red the W+c signal, in green the W+light quark contribution,
in purple the top contribution, in yellow the W+b and in blue the sum of the contribution
of the other backgrounds. Given the higher relevance of the W+b contribution, it is now
presented explicitly. Histograms are stacked.
be used in such case, since POWHEG does not correctly describe events with high jet
multiplicity. To further reduce the top related backgrounds, and profiting from the
high statistics available, stringent requirements on the invariant mass of the secondary
vertex or in the transverse mass of the jets plus W system (hereafter MTH ) can be set.
In Fig. 6.22 these distributions are shown for 2.2 fb−1 of 2011 data. Requiring the
secondary vertex mass to be smaller than 2 GeV and the MTH smaller than 300 GeV
removes partially the top background and consequently the systematic error associated
to it. These last requirements reduce tt̄ background by an additional 35%, the single-
top background by 23% and the signal by a 1%. With these new requirements, the
top related background is not anymore the dominant background, amounting to some
3% of the total data sample, being the dominant backgrounds W+light jets (30%) and
W+b jets (7%).
To reduce the tracking related systematic error the tagging efficiency should be
evaluated directly from the data. The proposed method, already checked successfully
207
WITH CHARMED JETS
on 2011 data, consists in computing this efficiency in a sample with dimuon events.
W+jet dimuon events are those events with a second muon coming from the decay of
a charm hadron in the jet. Signal events in this sample (as can be seen in Fig. 6.1) are
characterized by having two muons (one from the W boson, the other from the c quark
jet) with different sign (opposite sign, OS, muons).
With high statistics we can afford to select two subsamples: events having a sec-
ondary vertex characteristic of the decay of a heavy flavour quark (or c-quark) and
events with two muons of opposite charge in the final state, one of them characteristic
of the semileptonic decay of the heavy flavour quark (or c-quark). In fact we are using
an independent tagging method (a lepton tag). The normalization of these two samples
would give us information about the normalization scale of the signal and the efficiency
for each of the sets, i.e. the signal normalization and the c-tagging efficiency from the
SSVHE set and the signal normalization and the c-lepton tagging from the other set.
The intersection of these two sets (those events b-tagged and with two muons of oppo-
site sign) would give information of the signal normalization, the b-tagging efficiency
and the lepton tagging efficiency. The system:
tagged
NSSV HE = S · ǫV
tagged
Nlepton = S · ǫl
tagged
NSSV HE+lepton = S · ǫV · ǫl
can be solved by means of a set of discriminant distributions from the three plots for
each of the 3 subsamples (where S is the signal normalization, ǫV the vertex efficiency
and ǫl the lepton efficiency). To reinforce the W+c-W+b separation, we can use a
variable with high discriminant power for the dimuon subsample and the intersection
of it with the b-tagged one. The variable chosen is the pT,rel , defined as the pT of the
muon relative to the jet direction. b quark jets are less collimated, and a higher value
on this pT,rel variable is expected for this contribution. The pT,rel distribution for the
OS dimuon events subsample and the intersection b-tag dimuon OS sample is shown
Fig. 6.23.
We can further exploit the possibilities of this large recorded statistics and use the
set of same sign dimuon events to control the light background (see Fig. 6.24). The fit
is then performed over 5 histograms for each W charge: the SSVHE discriminant for
the b-tagged events and the pT,rel distribution for the dimuon tagged events and for
208

103
Data Data
+
W +charm W++charm
103 W++light W++light
W++b +
W +b
top 102 top
102
10
10
1
1
0 1 2 3 4 5 6 0 1 2 3 4 5 6
p for OS dimuon events [GeV] pT,rel for b-tagged OS dimuon events [GeV]
T,rel
Figure 6.23: pT,rel distribution for opposite sign dimuon events for the lepton tagged set
(left) and for the vertex and lepton tagged set (right). In red the W+c signal, in green
the W+light quark contribution, in purple the top contribution, in yellow the W+b and
in blue the sum of the contribution of the other backgrounds. Given the higher relevance
of the W+b contribution, it is now presented explicitly. Histograms are stacked.
209
WITH CHARMED JETS

Data Data
+
103 W +charm W++charm
W++light W++light
+
W +b
102 W++b
top top
102 Other bckg. Other bckg.
10
10
1
1
0 1 2 3 4 5 6 0 1 2 3 4 5 6
p for SS dimuon events [GeV] pT,rel for b-tagged SS dimuon events [GeV]
T,rel
Figure 6.24: pT,rel distribution for same sign dimuon events for the lepton tagged set
(left) and for the vertex and lepton tagged set (right). In red the W+c signal, in green
the W+light quark contribution, in purple the top contribution, in yellow the W+b and
in blue the sum of the contribution of the other backgrounds. Given the higher relevance
of the W+b contribution, it is now presented explicitly. Histograms are stacked.
the b-tagged dimuon events. A summary of the subsets used in the analysis and the
information given by each set is presented in Tab. 6.9.
The method is well established and preliminary results are encouraging. Reaching
the desired precision seems to be feasible.
210
Positive Discriminator Negative Discriminator

B-tagged events DSSV HE DSSV HE
Signal Norm., c-tagging eff. background control
OS SS
Dimuon events pT,rel pT,rel
Signal Norm., dimuon eff. background control
Dimuon-b-tagged events pT,rel pT,rel
Signal Norm., c-tagging eff, dimuon eff. background control
Table 6.9: Subsamples used in the proposed analysis and the information given in each
set.
211
WITH CHARMED JETS
212
7
Conclusions
This thesis presents several analysis involving muons detected in the CMS experiment
at the LHC. The final goal was the study of the inclusive W boson production and the
√
associated production of W bosons with charm-jets in pp collisions at s = 7 TeV with
the data collected by CMS during the 2010 LHC run. Being the muons fundamental
elements of many analysis and in particular of the ones presented in this report, the
first part of the thesis is dedicated to performance studies of the muon detector, in
particular of the central detector (DT system) with cosmic data, in the context of the
final commissioning of the detector already in place in its final location.
The central muon system is based on the measurement of the drift time of the
electron cloud produced by the passage of a muon through the detector. One of the
main quantities to be calibrated and monitored regularly to ensure a proper translation
of the physical quantity measured, drift time, to muon track position is the drift velocity.
The drift velocity in the chambers was measured and found to be homogeneous in all
of them once the uncertainty in the timing produced by the asynchronous arrival of the
cosmic muons is corrected for. It was also checked that the measured velocity varies
with the angle of incidence of muons. There was up to a 3% increase in the apparent
drift velocity for muons with an incidence angle of 300 , the maximum angle expected
for muons produced at the interaction vertex. Finally, it was confirmed that the drift
velocity is not affected by the residual magnetic field present in the chambers, except
in the innermost stations of the outermost wheels, where the remnant radial magnetic
field was largest. The maximal variation observed is ∼ 2.6%.
213
7. CONCLUSIONS
One of the design objectives of the muon barrel spectrometer was to guarantee a
100% muon detection and reconstruction efficiency. This is achieved with a highly re-
dundant design, with several layers of detector units, complemented with an optimized
software to perform the track reconstruction from the hits recorded in the chambers.
The muon reconstruction efficiency in the DT system was evaluated with a new method
proposed in this thesis. A careful study taking into account the geometry of the de-
tector, dead zones, overlaps, the irregular illumination of cosmic muons in the different
stations, the different momentum spectra of muons reaching top and bottom parts of
the detector wheels etc. was necessary in order to obtain an unbiased result. The
reconstruction efficiency was shown to be very high, larger than 99% in the bending
plane and bigger than 95% in the logitudinal direction.
An accurate knowledge of the muon performance in terms of momentum assignment
(scale and resolution) is crucial in any physics analysis. Should any effect potentially
affecting the muon measurement had been found, proper corrections would have had
to be incorporated in order to ensure a final result free of any bias. A novel approach
has been proposed in this work to determine the muon momentum scale and resolution.
The description of the behaviour of the CMS detector and of the software algorithms
implemented for muon reconstruction is taken as the baseline description, and new
terms modifying it are included in the MC to reproduce the observed characteristics
of the data. These additional terms are evaluated by careful comparison of the MC
prediction of the reconstructed dimuon mass distribution in the Z mass region with the
experimental distribution.
The method has been applied to the data collected by CMS during the 2010 LHC
run. It showed that the muon momentum scale and resolution are very well described in
the MC and only residual effects were found. A periodical dependence on the azimuthal
angle φ was found, with the same amplitude but opposite sign for positive and negative
muons. A certain modulation of the effect as a function of the muon pseudorapidity
was also observed. Overall the effect is very small and it was not considered to be
large enough that a correction is needed for the physics analysis reported in this thesis.
These effects were properly quantified and parametrized and delivered to the physics
groups of the collaboration working on analysis relying on muon measurement.
The inclusive production of W boson has been studied. The total production cross
sections as well as the production cross section for positive and negative W and their
214
ratio were measured to be:
σ(W) = 10.10 ± 0.03 (stat.) ± 0.16 (syst.) ± 0.40 (lumi.)
σ(W+ ) = 5.93 ± 0.02 (stat.) ± 0.10 (syst.) ± 0.24 (lumi.)
σ(W− ) = 4.17 ± 0.02 (stat.) ± 0.09 (syst.) ± 0.17 (lumi.)
σ(W+ )/σ(W− ) = 1.421 ± 0.008 (stat.) ± 0.11 (syst.)
The result is in agreement with and supersedes the first CMS measurement, performed
with a smaller data sample. The selection process was carefully optimized aiming for
a reduction of the main experimental systematic uncertainty, arising from the back-
ground subtraction, keeping at the same time under control the uncertainty due to
the theoretical uncertainties in the acceptance correction factors. The measurement
also profited from the larger data sample that allowed a better understanding of the
data and a reduction of certain experimental uncertainties such as those associated
to lepton efficiency, signal modeling etc. The precision achieved in the final result is
at the level of 1.6% in the total cross section, 1.7% and 2.1% for the individual W+
and W− cross sections (excluding the luminosity uncertainty) and 2.5% for the ratio.
The experimental contribution to these uncertainties is small, being the theoretical
uncertainties the major contributors to the final budget. These results have validated
next-to-next-to-leading order QCD cross section calculations based on recent parton
distribution functions.
Finally, the large data sample collected permitted the access to W final states with
a rather small production cross section. The process studied was the associated pro-
duction of a W boson with a c-quark jet. This final state is produced dominantly by the
interaction of a strange quark and a charm quark from the sea, thus directly probing the
strange quark content of the proton in a wide range on the x-Bjorken variable. For that
purpose the CMS algorithms for heavy quark tagging (mainly b-tagging) were success-
fully employed for c-tagging for a first time in the experiment. The measurement was
formulated in terms of cross section ratios of the production of a W boson accompanied
by a charm jet over the total associated production W+jets. A second measurement
was the charged ratio production of W+charm, (W+ + charm)/(W− + charm). The
215
7. CONCLUSIONS
measured ratios were:
Rc± = 0.92 ± 0.20 (stat.) ± 0.04 (syst.)

Rc = 0.141 ± 0.015 (stat.) ± 0.030 (syst.)
The Rc± ratio is limited by the statistical precision (∼ 20%), the systematic uncer-
tainty having a reduced impact (∼ 4%). The measured Rc ratio, instead, is affected
by a rather large systematical uncertainty (∼ 20%), arising mainly from the c-tagging
efficiency and top background subtraction.
These measurements constitute a first step towards a comprehensive understanding
of the W+charm production in pp collisions at high energy. Based on the experience ac-
quired with the present study, the analysis strategy to be carried with the data collected
in 2011 is already deployed. Projections from the current measurements indicate the
feasibility to significantly reduce the experimental uncertainty thus contributing to an
improvement in the determination of the strange quark Parton Distribution Function.
216
References
[1] M. B. et al., “The LHC design report v.3: the LHC Infrastructure and General
Services,” CERN-2004-003-V-3, 2004. 5, 11
[2] O. B. et al., “The LHC design report v.1: the LHC Main Ring,”
CERN-2004-003-V-1, 2004. 5, 11, 13
[3] O. B. et al., “The LHC design report v.2 : the LHC Infrastructure and General
Services,” CERN-2004-003-V-2, 2004. 5, 11
[4] M. Herrero, “The standard model,” http::/arxiv.org/abs/hep-ph/9812242v1, 2007.

5
[5] C. Amsler et al., “Review of particle physics,” Phys. Lett., vol. B667, 2008. 6, 29,
57
[6] LEP Electroweak Working Group, ATLAS Collaboration,

http://lepewwg.web.cern.ch/LEPEWWG/. 7
[7] R. Barate et al., “Search for the standard model Higgs boson at LEP,” Phys. Lett.,
vol. B565, pp. 61–75, 2003. 7
[8] Tevatron New Phenomena and Higgs Working Group, “Combined CDF and
DØ Upper Limits on Standard Model Higgs-Boson Production with up to 6.7
fb−1 of Data,” webpage, arXiv:1007.4587 [hep-ex], 2010. 7
[9] “Combined Standard Model Higgs boson searches with up to 2.3 fb−1 of pp colli-
√
sions at s = 7 TeV at the LHC,” Tech. Rep. ATLAS-CONF-2011-157, CERN,
Geneva, Nov 2011. 7
[10] S. P. Martin, “A supersymmetry primer,” arXiv:hep-ph/9709356v6, 2011. 10
217
REFERENCES
[11] K. R. Dienes, “String theory and the path to unification: A review of recent
developments,” arXiv:hep-th/9602045v3, 1997. 10
[12] C. T. Hill and E. H. Simmons, “Strong dynamics and electroweak symmetry break-
ing,” arXiv:hep-ph/0203079v3, 2002. 10
[13] M. J. Hadley, “Classical dark matter,” arXiv:gr-qc/0701100v1, 2007. 10
[14] I. Bigi, “CP violation - an essential mystery in nature’s grand design,” arXiv:hep-
ph/9712475v1, 1997. 10
[15] M. Awramik, M. Czakon, A. Freitas, and G. Weiglein, “Precise prediction for the
W boson mass in the Standard Model,” Phys. Rev. D, vol. 69, p. 053006, Mar
2004. 11
[16] CDF Collaboration, “Observation of top quark production in pp collisions with

the collider detector at Fermilab,” Phys. Rev. Lett., vol. 74, pp. 2626–2631, Apr
1995. 11
[17] ALICE Collaboration, “ALICE: Technical proposal for a large ion collider exper-
iment at the CERN LHC,” CERN-LHCC-95-71, 1995. 14
[18] ATLAS Collaboration, “The ATLAS Experiment at the CERN Large Hadron
Collider,” JINST 3 S08003, 2008. 14
[19] CMS Collaboration, “The CMS experiment at the CERN LHC,” JINST
0803:S08004, 2008. 14, 87
[20] LHCb Collaboration, “LHCb technical design report: Reoptimized detector de-
sign and performance,” CERN-LHCC-2003-030, 2003. 14
[21] CMS Collaboration, “CMS Physics Technical Design Report Vol-

ume 1: Detector Performance and Software,” CERN-LHCC-2006-001,
http://cdsweb.cern.ch/record/922757. 15, 20, 23
[22] G. Bayatian et al., “CMS physics Technical Design Report, Volume II: Physics
performance,” J.Phys.G, vol. G34, pp. 995–1579, 2007. 15
218
REFERENCES
[23] CMS Collaboration, “Precise Mapping of the Magnetic Field in the CMS Barrel
Yoke using Cosmic Rays,” JINST, vol. 5, p. T03021, 2010. 17
[24] CMS Collaboration, “The CMS tracker system project : Technical Design Re-
port,” CMS-TDR, no. CERN-LHCC-98-006; CMS-TDR-005, 1997. 18
[25] CMS Collaboration, “The CMS tracker: addendum to the Technical Design Re-
port,” CMS-TDR, no. CERN-LHCC-2000-016; CMS-TDR-005-add-1, 2000. 18
[26] CMS Collaboration, “Measurement of Momentum Scale and Resolution using

Low-mass Resonances and Cosmic Ray Muons,” CMS-PAS-TRK-10-004, 2010. 21
[27] CMS Collaboration, “Tracking and Vertexing Results from First Collisions,”
CMS-PAS-TRK-10-001, 2010. 22
[28] CMS Collaboration, “The CMS electromagnetic calorimeter project : Technical

Design Report,” CMS-TDR, no. CERN-LHCC-97-033; CMS-TDR-004, 1997. 20
[29] CMS Collaboration, “Electromagnetic calorimeter commissioning and first results

with 7 TeV data,” CMS-PAS-EGM-10-002, 2010. 24, 25
[30] CMS Collaboration, “The CMS hadron calorimeter project : Technical Design
Report,” CMS-TDR, no. CERN-LHCC-97-031; CMS-TDR-002, 1997. 23
[31] CMS Collaboration, “Jet Performance in pp Collisions at 7 TeV,” CMS-PAS-

JME-10-003, 2010. 26
√
[32] CMS Collaboration, “Jet energy resolution in CMS at s = 7 TeV,” CMS note
JME-10-014, 2011. 28
[33] CMS Collaboration, “The CMS muon project : Technical Design Report,” CMS-
TDR, no. CERN-LHCC-97-032; CMS-TDR-003, 1997. 29, 31
[34] P. Igo-Kemenes, “Searches for higgs bosons.” pdg.lbl.gov/2006/reviews/higgs

s055.pdf. 30
[35] CMS Collaboration, “CMS physics technical design report, volume II: Physics
performance,” J. Phys. G: Nucl. Part. Phys, 2007. 30
219
REFERENCES
[36] G. Abbiendi et al., “Muon reconstruction in the CMS detector,” AN-2008/097,

2008. 37, 60, 87, 143
[37] CMS Collaboration, “Performance of the CMS drift tube chambers with cosmic
rays,” JINST, vol. 5, no. 03, p. T03015, 2010. 39, 41, 54
[38] C. Grupen, Particle Detectors. Cambridge University Press, 1996. 88
[39] S. Agostinelli et al., “Geant4: a simulation toolkit,” Nuclear Instruments and

Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors
and Associated Equipment, vol. 506, no. 3, pp. 250 – 303, 2003. 90
[40] CMS Collaboration, “Tracking and vertexing results from first collisions,” CMS
PAS TRK-10-001 (2010). 92
[41] CMS Collaboration, “Tracking and vertexing results from first collisions,” CMS
PAS TRK-10-004 (2010). 92
[42] P. Nason, “A new method for combining NLO QCD with shower Monte Carlo
algorithms,” JHEP. 95, 108, 124, 142, 178
[43] P. S.Frixione and C.Oleari, “Matching NLO QCD computations with parton
shower simulations: the POWHEG method,” JHEP. 95, 142
[44] J.Campbell and K.Ellis, “A Monte Carlo for FeMtobarn processes at Hadron Col-
liders.” http://mcfm.fnal.gov/. 96, 177
[45] A. D. Martin et al., “Parton distributions for the LHC,” arXiv:0901.0002v3 [hep-
ph]. 96, 124
[46] CMS Collaboration, “Results on local muon reconstruction in DT chambers from

analysis of CRAFT data,” CMS CFT 09-012. 95
[47] R. Castello, Alignment of CMS Tracker detector using cosmic ray particles and its
impact on early physics performance. PhD thesis, Turin University, 2010. 109
[48] M. Borgia, The CMS Silicon Tracker: from the performance study in cosmic runs
to the measurement of the momentum resolution from the Z line-shape in early
data. PhD thesis, Turin University, 2010. 121
220
REFERENCES
[49] T. Sjostrand et al., “Pythia 6.2: Physics and manual,” arXiv:hep-ph/0108264v1.

122, 142
[50] D. Bourilkov et al., “LHAPDF: PDF use from the Tevatron to the LHC,”
arXiv:hep-ph/0605240v2. 124
[51] J. Pumplin et al., “New generation of parton distributions with uncertainties from
global QCD analysis,” arXiv:hep-ph/0201195v3. 124
[52] C.Balazs et al., “RESBOS package. QT resummation portal.”

http://hep.pa.msu.edu/resum. 125
[53] P. Langacker, Precision test of the standard electroweak model. World scientific
publishing, 1995. 129
[54] U. Baur et al., “The charm content of W+1 jet events as a probe of the strange
quark distribution function.,” arXiv:hep-ph/9308370v1, 1993. 129
[55] CDF Collaboration, “First run II measurement of the W boson mass at the Fer-
milab Tevatron,” Phys. Rev. D, vol. 77, p. 112001, Jun 2008. 130
[56] CMS Collaboration, “Search for W’ in the leptonic channels in pp collisions at

√
s =7 TeV,” CMS-EXO-11-024, 2011. 131
[57] CMS Collaboration, “Measurements of inclusive W and Z cross sections in pp

√
collisions at s = 7 TeV with the CMS experiment,” JHEP, vol. 2011, pp. 1–40,
2011. 10.1007/JHEP01(2011)080. 130, 140, 145, 171
[58] “Measurement of the lepton charge asymmetry in inclusive W production

√
in pp collisions at s = 7 TeV,” JHEP, vol. 2011, pp. 1–31, 2011.
10.1007/JHEP04(2011)050. 130
[59] C. F. Berger et al., “Next-to-Leading Order QCD Predictions for W+3-Jet Distri-
butions at Hadron Colliders,” Phys. Rev., vol. D80, p. 074036, 2009. 131
[60] CMS Collaboration, “Measurement of the Drell-Yan cross section in pp collisions

√
at s = 7 TeV,” JHEP, vol. 2011, pp. 1–41, 2011. 10.1007/JHEP10(2011)007.
131
221
REFERENCES
[61] CMS Collaboration, “Measurement of the polarization of W bosons with large

transverse momenta in W + jets events at the LHC,” Phys. Rev. Lett., vol. 107,
p. 021802, Jul 2011. 131
[62] CMS Collaboration, “Jet production rates in association with W and Z bosons in
√
pp collisions at s = 7 TeV,” arXiv:1110.3226, 2011. 132
[63] K. Melnikov and F. Petriello, “W boson production cross section at the Large
Hadron Collider with O(αs2 ) corrections,” Phys. Rev. Lett., vol. 96, p. 231803, Jun
2006. 132, 135
[64] C. F. Berger, Z. Bern, L. J. Dixon, F. Febres Cordero, D. Forde, T. Gleisberg,

H. Ita, D. A. Kosower, and D. Maitre, “Precise predictions for W+4-jet production
at the Large Hadron Collider,” Phys. Rev. Lett., vol. 106, p. 092001, Mar 2011.
132
[65] R. K. Ellis, K. Melnikov, and G. Zanderighi, “W+3 jet production at the Teva-
tron,” Phys. Rev., vol. D80, p. 094002, 2009. 132
[66] C. F. Berger, Z. Bern, L. J. Dixon, F. Febres Cordero, D. Forde, T. Gleisberg,

H. Ita, D. A. Kosower, and D. Maitre, “Next-to-leading order QCD predictions
for Z,γ ∗ + 3-jet distributions at the Tevatron,” Phys. Rev. D, vol. 82, p. 074002,
Oct 2010. 132
[67] CMS Collaboration, “Measurement of associated charm production in W final

√
states at s = 7 TeV,” CMS note EWK-11-013, 2011. 132
[68] CMS Collaboration, “Measurement of the Z/γ b-jet cross section in pp collisions
√
at s = 7 TeV,” CMS note EWK-10-012, 2011. 132
[69] “Measurement of the inclusive Z cross section via decays to tau pairs in pp collisions
√
at s = 7 TeV,” JHEP, vol. 2011, pp. 1–32, 2011. 10.1007/JHEP08(2011)117.
132
√
[70] CMS Collaboration, “Observation of W → τ ν production in pp collisions at s
= 7 TeV,” CMS-PAS-EWK-11-002, 2011. 132
222
REFERENCES
[71] CMS Collaboration, “Measurement of forward-backward asymmetry of lepton

pairs and the weak-mixing angle at CMS,” CMS note EWK-10-011, 2011. 133
[72] CMS Collaboration, “Measurement of wγ and zγ production in pp collisions at

sqrt(s) = 7 tev,” Physics Letters B, vol. 701, no. 5, pp. 535 – 555, 2011. 133
[73] CMS Collaboration, “Measurement of w+ w− production and search for the higgs
boson in pp collisions at sqrt(s) = 7 tev,” Physics Letter B, vol. 699, no. 1-2, pp. 25
– 47, 2011. 133
[74] CMS Collaboration, “Measurement of the weak mixing angle with the Drell-Yan
process in proton-proton collisions at the LHC,” Phys. Rev. D, vol. 84, p. 112002,
Dec 2011. 133
[75] A. Straessner, Electroweak physics at LEP and LHC. Springer, 2010. 134
[76] K. Melnikov and F. Petriello, “Electroweak gauge boson production at hadron

colliders through O(αs2 ),” Phys. Rev. D, vol. 74, p. 114017, Dec 2006. 135
[77] C. Diez et al., “Establishing the method to measure the single muon efficiency
from Z → µµ events with the first data,” CMS AN-2010/233. 151
[78] C. Diez, Searches for New Heavy W’ Bosons Decaying into Muon and Neutrino in
pp Collisions at 7 TeV in the CMS Experiment at the LHC. PhD thesis, Univer-
sidad Complutense, 2011. 151
[79] M. Cepeda, Measurement of the inclusive production cross-section of W bosons in

proton-proton collisions at 7 TeV in the CMS experiment at the LHC. PhD thesis,
Universidad Complutense de Madrid, 2011. 153
[80] CMS Collaboration, “Measurement of the inclusive W and Z production cross sec-
√
tions in pp collisions at s = 7 TeV with the CMS experiment,” JHEP, vol. 2011,
pp. 1–76, 2011. 10.1007/JHEP10(2011)132. 169
[81] ATLAS Collaboration, “Measurement of the inclusive W± and Z/γ cross sections
√
in the electron and muon decay channels in pp collisions at s = 7 TeV with the
ATLAS detector,” 2011. 172
223
REFERENCES
[82] J.Rojo, “W+c data with aMCatNLO an their possible impact on strangeness.”
Presentation at the Working Group on Electroweak precision measurements at
the LHC, November 2011. 175, 206
[83] M. W. Krasny, F. Dydak, F. Fayette, W. Placzek, and A. Siodmok, “∆MW <

10 MeV/c2 at the LHC: a forlorn hope?,” Eur. Phys. J., vol. C69, pp. 379–397,
2010. 175
[84] J. Alwall et al., “MadGraph/MadEvent v4: The New Web Generation,” JHEP,
vol. 09, p. 028, 2007. 179
[85] CMS Collaboration, “Measurement of the underlying event activity at the LHC
√ √
with s = 7 TeV and comparison with s = 0.9 TeV,” arXiv:1107.0330v1, 2011.
179
[86] M. Cacciari, G. P. Salam, and G. Soyez, “The anti-kt jet clustering algorithm,”
JHEP, vol. 2008, no. 04, p. 063, 2008. 181
[87] CMS Collaboration, “Particle-flow event reconstruction in CMS and performance

for jets, taus and MET,” CMS PAS PFT-09-001, 2009. 181
[88] CMS Collaboration, “Commissioning of the particle-flow reconstruction in

minimum-bias and jet events from pp collisions at 7 TeV,” CMS PAS PFT-10-
002, 2010. 181
[89] CMS Collaboration, “Commissioning of b-jet identification with pp collisions at

√
s = 7 TeV,” CMS PAS, vol. BTV-10-001, 2010. 186, 200
[90] CMS Collaboration, “Jet energy corrections determination at 7 TeV,” CMS PAS,
vol. JMT-10-010, 2010. 195
[91] CMS Collaboration, “Jet performance in pp collisions at 7 TeV,” CMS PAS,

vol. JMT-10-003, 2010. 195
√
[92] CMS Collaboration, “Performance of b-jet identification with pp collision at s=
7 TeV,” CMS PAS BTV-11-001, 2011. 197
[93] H. Lai et al., “CT10 parton density functions,” arXiv:1007.2241. 198, 204
224
REFERENCES
[94] A. Martin et al., “MSTW2008 parton density functions,” arXiv:0901.0002. 198,

204
[95] R. Ball et al., “NNPDF2.1 parton density functions,” arXiv:1101.1300. 198, 204
[96] L. Gladilin, “Charm hadron production fractions,” arXiv:hep-ex/9912064v1. 198
225

Doctoral Thesis by PHD - Javier Santaolalla

Uploaded by

Copyright:

Available Formats

Doctoral Thesis by PHD - Javier Santaolalla

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Doctoral Thesis by PHD - Javier Santaolalla

Uploaded by

Copyright:

Available Formats

Centro de Investigaciones Energéticas, Medioambientales y

Departamento de Fı́sica Atómica, Molecular y Nuclear

Measurement of Electroweak processes

April 11, 2012

2 The Large Hadron Collider and the CMS experiment 5

3 Muon detection and reconstruction in CMS 29

3.4.2 Muon trajectory and crossing point determination in the test

4 Muon momentum measurement in CMS 87

5 Measurement of the inclusive W boson production cross section in

6 Measurement of associated production of W bosons with charmed

En primer lugar, agradezco enormemente el buen espı́ritu de trabajo y com-

Cientı́ficamente hablando también debo agradecer mucho a Mary Cruz por

Un agradecimiento muy especial a los que han compartido casa conmigo,

Burgos, Canarias, Madrid y Ginebra, a todos. Y por supuesto a mi familia,

Finalmente, un párrafo especial para una mujer demasiado especial e im-

En fin, a todos. A Griffin, Perdue, J.L. Guerra, Cherito, Dieguito, todos

The Large Hadron Collider and

The LHC (Large Hadron Collider) is a 27 km synchrotron that produces proton-proton

2.1 LHC physics program

Charge First Generation Second Generation Third Generation

Table 2.1: Matter content as described in the SM (5).

Electromagnetic Weak Strong Gravitation

Table 2.2: Force carrier particles as described by the SM.

hierarchy problem. Cancellations provided by a new group of particles is the

2.2 The LHC

to other non-physically-interesting processes makes necessary a high collision rate and

Figure 2.4: CMS detector - Representation of CMS detector

• Electromagnetic calorimeter. Exceptional energy resolution, diphoton and dielec-

• Hadronic calorimeter. Reasonable dijet mass resolution and missing transverse

In CMS different detector technologies are applied in order to optimize particle

to compute the momentum of the particles in the transverse plane, to detect

Detector Radius Size Occupancy

Pixel detector. The total amount of pixels in CMS is 66 millions, covering an

• Electromagnetic Calorimeter (ECAL). It is essential in the identification of

Figure 2.8: Representation of CMS tracker detector - Pictorial representation

Figure 2.9: Resolution on transverse momentum. - Resolution on transverse

Cross section Length (Radiation Lenghts) Coverage (η)

Table 2.4: Description of the elements of the ECAL detector.

Figure 2.11: ECAL supermodule energy resolution. - ECAL energy reso-

Figure 2.13: Invariant mass of π 0 candidates. - Mass spectrum of reconstructed

Figure 2.15: HCAL pT resolution - pT resolution as a function of the pT is

Muon detection and

3.1 Importance of muons in particle physics analysis

detectors, an accurate and robust momentum measurement is achieved. The excellent

Other interesting channels in the Higgs boson search are H → W W → lνlν,

• High dimuon mass resonances. Possible SM extensions predict the existence

• Supersymmetry. Supersymmetry (SUSY) is one of the most famous SM ex-

• Electroweak physics. Electroweak physics is important, especially in the early

3.2 CMS muon system

As briefly described in the previous chapter, CMS muon system is required to be

3.2.1 Cathode Strip Chambers.

3.2.2 Resistive Plate Chambers.

3.2.3 Drift Tube Chambers.

Figure 3.6: CMS DT chamber - A DT chamber consist of 2 φ SL, a non-active structure

# RUN B Field # events analyzed Data Taking

Table 3.1: CRAFT08 runs analyzed in this study.

3.3 Determination of the drift velocity

Tmean = (Ti + Tk )/2 + Tj (3.1)

Figure 3.8: Meantime method computation in a SL - This method is based in the