Springer Theses

Recognizing Outstanding Ph.D. Research

You-Hao Chang

Study of Double
Parton Scattering
in Photon + 3 Jets
Final State
In Proton–Proton Collisions at √s = 7 TeV
with the CMS Experiment at the LHC
Springer Theses

Recognizing Outstanding Ph.D. Research

Aims and Scope

The series “Springer Theses” brings together a selection of the very best Ph.D.
theses from around the world and across the physical sciences. Nominated and
endorsed by two recognized specialists, each published volume has been selected
for its scientific excellence and the high impact of its contents for the pertinent field
of research. For greater accessibility to non-specialists, the published versions
include an extended introduction, as well as a foreword by the student’s supervisor
explaining the special relevance of the work for the field. As a whole, the series will
provide a valuable resource both for newcomers to the research fields described,
and for other scientists seeking detailed background information on special
questions. Finally, it provides an accredited documentation of the valuable
contributions made by today’s younger generation of scientists.

Theses are accepted into the series by invited nomination only

and must fulfill all of the following criteria
• They must be written in good English.
• The topic should fall within the confines of Chemistry, Physics, Earth Sciences,
Engineering and related interdisciplinary fields such as Materials, Nanoscience,
Chemical Engineering, Complex Systems and Biophysics.
• The work reported in the thesis must represent a significant scientific advance.
• If the thesis includes previously published material, permission to reproduce this
must be gained from the respective copyright holder.
• They must have been examined and passed during the 12 months prior to
nomination.
• Each thesis should include a foreword by the supervisor outlining the signifi-
cance of its content.
• The theses should have a clearly defined structure including an introduction
accessible to scientists not expert in that particular field.

More information about this series at http://www.springer.com/series/8790

You-Hao Chang

Study of Double Parton

Scattering in Photon + 3
Jets Final State
pffiffi
In Proton–Proton Collisions at s ¼ 7 TeV
with the CMS Experiment at the LHC

Doctoral Thesis accepted by

the National Taiwan University, Taipei, Taiwan

123
Author Supervisor
Dr. You-Hao Chang Prof. Paolo Bartalini
College of Science CERN Meyrin
National Taiwan University Geneva
Taipei Switzerland
Taiwan

ISSN 2190-5053 ISSN 2190-5061 (electronic)

Springer Theses
ISBN 978-981-10-3823-5 ISBN 978-981-10-3824-2 (eBook)
DOI 10.1007/978-981-10-3824-2
Library of Congress Control Number: 2017930604

© Springer Nature Singapore Pte Ltd. 2017

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made. The publisher remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.

Printed on acid-free paper

This Springer imprint is published by Springer Nature

The registered company is Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore
To life—where those amazing science exist in.

To my lovely family—without your

unconditional support, nothing I can achieve.

To myself—the one who was dreaming of

being a scientist from childhood.
Supervisor’s Foreword

This work does focus on the study of photon þ 3 jet final states in proton–proton
pffiffi
collisions at s ¼ 7 TeV, looking for patterns of two (or more) distinct hard
scatterings in the same collision, i.e. the so-called Double Parton Scattering (DPS).
The analysis is performed in the light of the state-of-the-art Monte Carlo generators
providing higher order corrections to the description of the Single Parton Scattering
(SPS) background.
The conclusions on the DPS observation on this channel are prudent as the
current theoretical uncertainties related to the SPS background are found to be
large. At the same time, a rich set of DPS-sensitive measurements is reported for
possible further interpretation. The analysis constitutes an important milestone in
the progress of Multiple Parton Interactions (MPI) studies at the Large Hadron
Collider. It had a big impact on all the DPS papers published by the CMS col-
laboration for reasons of its open-minded approach inviting to deeply review the
methodologies adopted in the early DPS measurements reported at Tevatron, in
particular those focusing on the photon þ 3 jet benchmark channel.

Geneva, Switzerland Prof. Paolo Bartalini

April 2016

vii
Abstract

The probability of having more than one interaction per collision is nonnegligible at
the LHC. These additional interactions might reach a hard scale comparable to the
primary scattering and become experimentally distinguishable at high energies.
Distributions sensitive to double parton scattering are investigated in the photon + 3
jets final state in proton–proton collisions at a center-of-mass energy of 7 TeV. The
data were collected by the CMS experiment at the LHC with an integrated lumi-
nosity of 36 pb−1 in 2010. The cross section r for a final state with a photon and a
jet of transverse momentum pT [ 75 GeV together with 2 jets of pT [ 20 GeV,
where the photon and jets are within the fiducial volume of the CMS detector, is
measured to be 124.9  8.9 (stat.)  22.6 (syst.) pb. The differential cross sections
are measured as a function of the difference in azimuthal angles and the transverse
momentum balance between the photon-jet pair and the di-jet pair. Further it is
investigated whether additional contributions from double parton scattering can
improve the agreement between the measured data and the Monte Carlo predictions.

ix
Acknowledgements

Research is probably the most noble compromise between

freedom and discipline.

—Paolo Bartalini

I would like to show my deepest gratitude to whom I meet with the completion of
this dissertation. I do believe those inspiration accumulated in my life makes who I
am.
First I must thank my supervisors, Prof. Min-Zu Wang and Prof. Paolo Bartalini,
for guiding me into a free learning field of experimental high energy physics. They
are great advisors and mentors who accompanied my course from being guided to
be able to independently complete my research. In addition, I appreciate my
co-supervisors Prof. Shin-Shan Yu and Dr. Yuan Chao. They not only provided me
crucial technical support in academic researches but also wise suggestions and aids
to my life. I also thank the faculties in Lab of NTU High Energy Physics. I have
developed new attitudes concerning cooperation, and staying competitive scien-
tifically or socially because of Prof. Wei-Shu Hou, Prof. Yee Hsiung, Prof. Pao-Ti
Chang, Prof. Kai-Feng Chen, Prof. Stathes Paganis and Dr. Jing-Ge Shiu. I am
especially indebted to Profs. Wei-Shu Hou and Min-Zu Wang because of the
general advices and supports they give, so I learn better to reflect from my mistakes.
Many thanks to Yeon-Jyi Lei, Yeng-Ming Tzeng, Yu-Wei Chang, Kai-Yi Kao,
Kuan-Hsin Chen, Yun-Ju Lu, Jr-Kan Hsieh and Lei Hong for providing me spiritual
and substantial supports on my journey in the field of experimental high energy
physics. With your company beside, I do not feel alone anymore while doing
research abroad. I also appreciate the experiences and advices shared by Mark
Chiang, Jui-Te Wei, Bean Huang, Be-Bo Ho and Po-Yuan 56 which inspired me to
cope with my obstacles in a more mellow way. Aside from scientific works, my
interaction with Jelov Peng, Jia-Jia Chang, Kacaw Kaokoy Cikatopay, Small-Pig
Shiu, Dong-Di 56, Jian-Wei 56, Zu-Ta 56, Yun-Tsung 56, Kai-Ming 56, Po-Hsun
56, Shankar Tien, Yu-Tan Chen, This Lin, Dallas Python Wang, Tzu-An Sheng,
Chia-Hao Tu, Mars Chen, Chou Dong, Jenny Huang, Phoebe Liao and Link Liu
broaden my horizon and enrich my life.

xi
xii Acknowledgements

Last but not least, I am grateful beyond words towards my family and all friends.
With their consideration, I am fearless to walk on this winding long road to do my
Ph.D. With their assistance, I achieve a small goal of my life.
Being a scientist to explore our world has been my dream since I was a child.
This is a dream I would never forget. This is a dream I would ever bear in my mind.
My sincere appreciation to all the people I meet on the road of science.
Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2 Experimental Apparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1 Overview of the LHC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 The CMS Detector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.1 Magnet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.2 Tracking Detectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.3 Electromagnetic Calorimeter (ECAL) . . . . . . . . . . . . . . . . 14
2.2.4 Hadronic Calorimeter (HCAL) . . . . . . . . . . . . . . . . . . . . . 16
2.2.5 Jet Reconstruction at CMS . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.6 Muon System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2.7 Global Event Reconstruction . . . . . . . . . . . . . . . . . . . . . . 21
2.3 Trigger System. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.3.1 Level-1 Trigger (L1). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3.2 High Level Trigger (HLT) . . . . . . . . . . . . . . . . . . . . . . . . 23
2.4 Luminosity Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.4.1 The Primary Luminosity . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.4.2 The Absolute Luminosity . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.5 Computing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3 Data and Monte Carlo Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4 Event Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.1 Trigger Requirement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.2 Pileup Reweighting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.3 Photon Identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.4 Jet Identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5 Discriminating Observables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

xiii
xiv Contents

6 Photon + 3 Jets Events at Detector Level . . . . . . . . . . . . . . . . . . . . . . 49

6.1 DPS Performance at Detector Level . . . . . . . . . . . . . . . . . . . . . . . 49
6.2 Purity of Isolated Prompt Photons . . . . . . . . . . . . . . . . . . . . . . . . 52
7 Correction and Unfolding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
7.1 Acceptance, Background, Purity and Stability . . . . . . . . . . . . . . . 58
7.2 Unfolding with ‘RooUnfold’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
8 Systematic Uncertainties. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
8.1 Photon Energy Scale Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . 65
8.2 Photon Energy Resolution Uncertainty. . . . . . . . . . . . . . . . . . . . . 66
8.3 Jet Energy Scale Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
8.4 Jet Energy Resolution Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . 68
8.5 Model Dependence Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . 69
8.6 Photon Identification Efficiency Uncertainty . . . . . . . . . . . . . . . . 69
8.7 Pileup Reweighting Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . 70
8.8 Second Primary Vertex Selection Uncertainty . . . . . . . . . . . . . . . 70
8.9 Jet-Vertex Association Uncertainty . . . . . . . . . . . . . . . . . . . . . . . 71
8.10 The Sample Size of Monte Carlo Simulation . . . . . . . . . . . . . . . . 71
8.11 HLT Efficiency Correction Uncertainty . . . . . . . . . . . . . . . . . . . . 72
8.12 Total Systematic Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
9 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
10 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
Appendix A: K-Factor Correction for the LO Cross-Section
of PYTHIA, MADGRAPH and SHERPA MCs . . . . . . . . . . . . . . . . 83
Appendix B: Study of Low Acceptance of Photon + 3 Jets Events . . . . . . 87
Appendix C: More DPS Discriminating Observables . . . . . . . . . . . . . . . . 89
Curriculum Vitae . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
Chapter 1
Introduction

Quantum Chromodynamics (QCD) is the underlying theory to describe hard and soft
high-energy hadron collisions. Much of the complexity involved in describing proton-
proton scatters derives from the composite nature of hadrons. In order to compare
theory predictions with experimental data, the set of final-state partons is transformed
into a set of hadrons in the fragmentation step of initial-state (ISR) and final-state
radiation (FSR), hardest parton interaction (HI), multiple parton interaction (MPI)
and beam remnants. The underlying event (UE) then is everything in a single proton-
proton interaction except for the hard scattering component, or, associated with all
particles produced by the hadron remnants “after” the hard scattering. Figure 1.1
shows a typical proton-proton collision.
The basic idea of the multiple interaction models is assuming that QCD fac-
torization theorem is working not only for the hard scattering process but also for
additional scatters. Hence, the hadronic cross section for the additional interaction
(the second, the third…etc.) and the modeling of multiple interactions when simu-
lating proton-proton collisions make use of several adjustable parameters. The set
of parameters which describes a particular final state is referred to as an underlying
event tune. Most underlying event observation only probe the sum of these effects.
But multiple parton interactions models are particularly successful in accounting for
various features of the underlying event at hadron colliders, to exactly quantify their
contribution to the underlying event is a nontrivial task. Indeed, including multiple
parton-parton scatters in the underlying event model has been shown to significantly
improve the predictions of different Monte Carlo (MC) [1–3]. Observation of an
excess of balanced pairs of jets where pairs are uncorrelated with respect to each
other would support the hypothesis of multiple interactions taking place. If these
interactions reach relatively large values of the exchanged transverse momentum,
x, the observation of multiple parton interaction mostly relies on the double parton
scattering (DPS), which is one of the direct evidence for multiple parton interactions.
It is definitely important of understand double parton scattering since such additional
interactions result in higher rate of multi-jet production than the normal predictions,
© Springer Nature Singapore Pte Ltd. 2017 1
Y.-H. Chang, Study of Double Parton Scattering in Photon + 3 Jets Final State,
Springer Theses, DOI 10.1007/978-981-10-3824-2_1
2 1 Introduction

Fig. 1.1 Proton-proton collision at high energies: hard process (in red) with initial-state radiation
(in pink), final-state radiation (in brown), multiple parton interaction (in green) and beam remnants
(in black)

which don’t take the possibilities of additional interactions into account. Thus, it will
produce relevant and background in searching new physic phenomena.
By taking the QCD factorization into account for double parton scattering, the
corresponding cross section σA,BDPS
can be expressed as [4–9]:

m  dx  D(x , x  , b)D(x , x  , b)σ̂ (x , x )σ̂ (x  , x  ).
σDPS = dbdxA dxB dxA B A A B B 1 A B 2 A B
2 p⊥min
(1.1)

The partonic cross sections σ̂1 (xA xB ) and σ̂2 (xA , xB ) where indices 1, 2 reflect
possibly different subprocesses) are convoluted with two-parton density functions
D(xA , xA , b) and D(xB , xB , b) evaluated at the parton momentum fractions xA , xA ,
xB , xB . The m factor has the value m = 2 when two subprocesses are distinguishable
scattering, and m = 1 when they are indistinguishable [10]. A transverse distance
scale b represents the amount of spatial overlap between the hadrons’ wave functions:
The two-parton distribution functions refer to conditional probabilities D(x, x  , b) to
find one parton with momentum fraction x and one parton with momentum fraction
x  separated by a distance b in transverse space in the same proton. The cross section
integral is evaluated for scatters above a minimal transverse momentum scale thresh-
old p ⊥ min, i.e. for scatters producing hard enough jets. In the simplest approach,
the two-parton densities D are assumed to factorize, which is equivalent to assuming
uncorrelated two-parton distributions:

D(x, x  , b) = f eff (x) f eff (x  )F(b), (1.2)

1 Introduction 3

Fig. 1.2 Published measurements of effective cross section, σeff as a function of center of mass
energy of collision

with f eff being an effective parton density function and F(b) a matter overlap density.
If this assumption holds, the cross section for double parton scattering is simply
proportional to the product of the respective cross sections for single scatters:

m m σ1 σ2
σDPS = σ1 σ2 F 2 (b)d2 b = , (1.3)
2 2 σeff

where the dependence on the transverse impact parameter b is absorbed in the effec-
tive cross section eff. The current published values of eff as a function of center of
mass energy of collision are shown in Fig. 1.2.
The production of four high transverse momentum ( pT ) jets is thus the most
prominent process to study the impact of multiple interactions directly: Two inde-
pendent scatters in the same proton-proton collision each produce two jets. But
searches for double parton scattering in four-jet events at hadron colliders face sig-
nificant backgrounds from other sources of jet production, in particular from QCD
Bremsstrahlung. Typical thresholds employed in jet triggers bias the event sample
toward hard scatterings. However, a high- pT jet parton is more likely to radiate addi-
tional partons, thus producing further jets. Thus, the relative fraction of jets from
final-state showers above a given threshold is enlarged in jet trigger streams which
is an unwanted bias. On the other hand, looking for four jets in a minimum-bias
stream will yield very small statistics. Hence, studying the double parton scattering
in one photon and three jets final state is a novel approach. Analyses which try to
identify two hard scatters in multi-jet events typically rely on methods to overcome
combinatorics as there are three possible ways to group four objects into two pairs:
Combinations are commonly selected pairwisely balanced in azimuth and energy.
4 1 Introduction

Fig. 1.3 Dominant Feynman diagrams for the direct photon subprocesses: qg → qγ (left) and
qq → gγ (center) and for the photon + 3 jets production from double parton scattering (right)

In this inclusive analysis, we focus on the study of double parton scattering in one
photon and three jets final state.
At the Large Hadron Collider (LHC), the probability of having more than one
parton-parton interaction per proton-proton collision is sizeable [8]. These addi-
tional interactions become experimentally distinguishable in high energy hadronic
collisions [11]. Up to the present, double parton scattering measurements in hadron-
hadron collisions have been suggested in double Drell-Yan, four-jet, same-sign WW,
W plus 2 jets and single photon plus 3 jets production [4, 7, 12–15].
√ Experimentally,
DPS has been studied in the photon + 3 jets final √ state at s = 1.8 TeV [16] and
1.96 TeV [17–19], in the four-jet final state at s = 63 GeV [20], 630 GeV [21],
1.8 TeV [22], and 7 TeV [23], in √ the 2 b-jets + 2 jets final state at 1.96 TeV [19], and
in the W + 2 jets production at s = 7 TeV [24, 25].
At leading order in perturbative quantum chromodynamics (pQCD), several
processes contribute to the production of single prompt photons with large trans-
verse momentum: photons directly produced in the Compton-like gluon scattering
process qg → qγ (Fig. 1.3 (left)) and the quark-antiquark annihilation process qq →
gγ (Fig. 1.3 (center)) [26]. An additional contribution to prompt photon production
arises from the fragmentation of coloured partons and is referred to as fragmentation
photon. Multiple photons from electromagnetic decays of neutral light mesons (such
as π 0 and η) can mimic prompt photon production, when the opening angle between
the multiple photons from the meson decay is too small to effectively discriminate
them through electromagnetic shower properties. DPS signal events in the photon
+ 3 jets final state include pairwise balanced photon-jet and di-jet combinations
where the photon-jet pair and the di-jet pair are produced in the first and second hard
interaction of the same collision, respectively (Fig. 1.3 (right)). The photon source
of DPS signal events can be direct photons and fragmentation photons. DPS signal
events and single parton scattering (SPS) events are characterized by different kine-
matics, where the SPS events produce one photon and one jet in one interaction,
with the evolution of the jet resulting in additional jets at lower momenta (2 → 4
partonic process). Events containing photon + 3 jets produced from SPS or misiden-
1 Introduction 5

tified photon + 3 jets caused by decays of π 0 and η mesons constitute an irreducible

background.
Photon + 3 jets
√ production is measured in proton-proton collisions at a center-
of-mass energy s = 7 TeV using the data sample collected with the Compact
Muon Solenoid (CMS) detector at the LHC in 2010 with an integrated luminosity of
36 pb−1 . Events with one photon and at least three jets are analyzed by investigating
the azimuthal angular differences for the di-jet pair and the balance in transverse
momentum of the photon-jet and di-jet pairs. The measurement requires the trans-
verse momentum pT > 75 GeV for the leading (highest pT ) photon and the leading
jet and pT > 20 GeV for the other jets. Photons are measured in pseudorapidity
ranges |η| < 1.4442 and 1.566 < |η| < 2.5 and jets are selected up to |η| < 2.4.
The pseudorapidity is defined as η = −ln[tan(θ/2)], where θ is the polar angle
with respect to the counterclockwise-beam direction. Jets are reconstructed using
the anti-kT [27, 28] clustering jet algorithm with distance parameter of R = 0.5.
Various distributions are studied and corrected for detector effects and selection effi-
ciencies by unfolding, a.k.a. removing the known effects of measurements, to the
stable-particle level (lifetime cτ > 10 mm). The data then can be compared to the
predictions of various Monte Carlo (MC) event generators, where the MC simula-
tion contains DPS signal events, possible background SPS events, and misidentified
photon + 3 jets background events.
The structure of this thesis is organized as follows: a brief overview of the basic
theories in particle physics like Standard Model and Quantum Chromodynamics are
included in this chapter with their corresponding mathematical formulas and phe-
nomenological treatment. In Chap. 2, the experimental apparatus, the LHC and the
detailed description of the CMS experiment are introduced. The relevant working
flow of computing in LHC will be also shortly discussed. The details of the analyzed
data collected by the CMS detector and the simulated samples are summarized in
Chap. 3. Event selection criteria are described in full details in Chap. 4. The discrim-
inating kinematic observables, which are sensitive enough to observe the existence
of the DPS phenomenon, are discussed in Chap. 5. After having a basic idea of how
to analyze the DPS with specific observables, a direct study of both data and differ-
ent MC simulated samples is performed in photon + jets channel at detector level
in Chap. 6. Due to the importance of photon in our selected DPS channel, a deep
investigation of photon quality is also included in this chapter. The methodology of
correction and how we extract the true face of data through unfolding are described
in Chap. 7. Detailed systematic uncertainties are studied and presented in Chap. 8.
The measurement of the total integrated cross section and the normalized differential
cross section as a function of different DPS-discriminating observables, fully cor-
rected for detector effects and selection efficiency, are shown in Chap. 9. The final
chapter (Chap. 10) is summary and conclusion.
Chapter 2
Experimental Apparatus

A center-of-mass energy in a collider is efficiently higher than one in a fixed target

with using the same beam energy. At present the Large Hadron Collider √ (LHC) [29]
is designed to collide proton beams at a center-of-mass energy of s = 14 TeV with
a nominal instantaneous luminosity of L = 1034 cm−2 s−1 . The main purpose is to
uncover the nature of electroweak spontaneous symmetry breaking, and prob the
new physics phenomena beyond the Standard Model (SM) at the TeV energy scale.
There are six experiments located on the LHC ring, and the Compact Muon Solenoid
(CMS) experiment is one of them. In this chapter an overview of the LHC and the
detailed description of the CMS experiment will be introduced. Especially, the parts
of the detector relevant to the analyses of this thesis are treated in detail.

2.1 Overview of the LHC

The LHC accelerator is situated at the national boundary between Switzerland and
French in the existing Large Electron-Positron (LEP) tunnel with a length of 26.7 Km
as deep as 175 m. The schematics of the LHC with the beam injection sequence
is presented in Fig. 2.1. Protons are generated from the ionization of H2 plasma
enhanced by an electron beam, and injected into the Linear accelerator (Linac2) to
get an energy of 50 MeV. The Proton Synchrotron Booster (PSB), which is the first
and the smallest circular proton accelerator, speeds up the 50 MeV protons to 1.4 GeV.
Finally, the protons with an energy of 26 GeV coming from the Proton Synchrotron
(PS) are further accelerated by the Super Proton Synchrotron (SPS) to 450 GeV, and
then injected into the LHC, where they will be quickened to 7 TeV in bunches with
a nominal number of 1.15 · 1011 particles per bunch. The cylindrical bunches with a
radius of 16.6 µm and a length of 7.55 cm group the protons within a bunch spacing
of 25 ns, corresponding to a bunch crossing frequency of 40 MHz. In the LHC, two
kinds of superconducting magnets operating at a temperature of 1.9 K with liquid
© Springer Nature Singapore Pte Ltd. 2017 7
Y.-H. Chang, Study of Double Parton Scattering in Photon + 3 Jets Final State,
Springer Theses, DOI 10.1007/978-981-10-3824-2_2
8 2 Experimental Apparatus

Fig. 2.1 The schematics of the LHC with the beam injection sequence

helium are used to control the beam. During the acceleration, 1232 dipole magnets
providing a magnetic field of 8.33 T guide the particles along the design orbit, and
392 quadrupole magnets confine the particles in the vicinity of the design orbit, from
which most particles will unavoidably deviate. The Radio-Frequency (RF) system,
proving a RF voltage of 8–16 MV/beam, accelerates the particles and compensates
an energy loss of 7 KeV per turn due to the synchrotron radiation.
The six experiments operating at the LHC are shown in Fig. 2.2. Two experiments,
including the CMS [30] and ATLAS (A Toroidal LHC Apparatus) [31] detectors, are
optimized towards high- pT physics for general purposes, involving the verification of
the SM, the Higgs boson hunting, new physics beyond the SM, and heavy ion physics.
The motivation of the ALICE (A Large Ion Collider Experiment) [32] detector is to
study heavy ion collisions with an excellent ability of the low- pT measurement. The
purpose of the LHCb (the Large Hadron Collider beauty experiment) [33] detector
2.1 Overview of the LHC 9

Fig. 2.2 The main four experiments operating at the LHC

is to study the B physics in order to solve the CP-violation (CP standing for charge
parity). For another two special-purpose experiments, the TOTEM (TOTal Elastic and
diffractive cross section Measurement) experiment [34] in front of the CMS detector
is dedicated to measure the total cross section, elastic scattering, and diffractive
processes, and the LHCf (Large Hadron Collider forward) experiment [35] located
close to the ATLAS detector studies cosmic rays in laboratory conditions by using
the forward particles generated inside the LHC.
A LHC√ commissioning run with proton–proton collisions at center-of-mass ener-
gies of s = 7 TeV started on 30th March 2010. During the 2010 run, the recorded
efficiency of the CMS is about 92, and 84% of recorded data, (36.1 ± 1.4) pb−1 ,1
are certificated for physics analyses. This dissertation uses the 2010 CMS data in
proton–proton collisions at 7 TeV.

1 ‘p’indicates a number of an order of magnitude, −12, and a barn (symbol ‘b’) stands for a cross
section unit, 10−24 cm2 .
10 2 Experimental Apparatus

2.2 The CMS Detector

The CMS detector is situated on the interaction point 5 (P5) of the LHC at a depth of
100 m. In the LHC environment, the CMS detector should experimentally challenge
the physics benchmark channels with a small cross section against the overwhelm-
ing
√ background from QCD jet production, whose proton–proton cross section at
s = 14 TeV is roughly 100 mb with an event rate of approximately 109 inelas-
tic event/s. For reaching the physics goals, rejecting background with the optimal
efficiency in the interesting rare channels has to be performed with a high power.
Additionally, the design of the CMS is optimized to well identify muons, charged
particles, electrons, and jets with an excellent momentum and corresponding two-
particles mass resolution. Good E Tmiss 2 resolution has also been achieved according
to the related objects with an excellent momentum resolution. Moreover, secondary
vertices and impact parameters have to be precisely measured for an efficient iden-
tification of heavy flavor and τ -lepton decays.
An excess of inelastic (hard-core scattering) collisions happen in the interesting
events. That is, additional proton–proton interactions within the same beam crossing
(pileup). A reduction of the pileup effect can be performed using high-granularity
detectors with a good time resolution, resulting in low occupancy. In addition, the
sub-detectors and front-end electronics have to be required radiation-hard because
of a high radiation arising from the large flux of particles near the interaction region.
The CMS coordinate system adopts a right-handed style such that the origin
is centered at the nominal collision point, the x-axis directs south to the center
of the LHC ring, the y-axis points vertically upward, and the z-axis is along the
beam direction to the west. The azimuthal angle φ is measured from the x-axis in
the x y plane, and the polar angle θ is measured from the z-axis. Pseudorapidity,
defined as η = −ln tan(θ/2), is introduced to replace the polar angle θ because of
a property of Lorentz invariant for boosting along the z-axis. Besides, η is a good
approximation of the rapidity ( 21 ln E+ pz c
E− pz c
) in a massless-particles case. The transverse
momentum (denoted pT ) relative to the beam direction is computed from the x- and
y-components, while the transverse energy E T is obtained from Esinθ .
The overview of the CMS apparatus is presented in Fig. 2.3. Briefly speaking, the
CMS detector is composed of a silicon tracking system, an electromagnetic calorime-
ter (ECAL), a hadronic calorimeter (HCAL), and a muon system. A superconducting
solenoid magnet at the heart of the CMS detector provides a magnetic field of 3.8 T.
The CMS detector is 21.6 m long with a diameter of 14.6 m, and has overall weight
of 12500 tons. The detailed descriptions are presented in the following sub-sections.

2 The imbalance of energy is measured in the transverse plane, and also denoted by MET.
2.2 The CMS Detector 11

Fig. 2.3 The overview of the CMS detector

Table 2.1 Parameters of the Parameter Value

CMS superconducting
solenoid Magnetic field 3.8 T
Inner bore 5.9 m
Length 12.9 m
Number of turns 2168
Current 19.5 kA
Stored energy 2.7 GJ
Hoop stress 64 atm

2.2.1 Magnet

The CMS superconducting solenoid provides a large bending power to perform

a good muon momentum resolution of Δp/ p ≈ 10%3 at p = 1 TeV/c, while the
bending starts at the primary vertex. During the current data taking, the magnetic
field reaches 3.8 T, and the magnetic flux is returned through the iron yoke in which
the muon detector is installed. The designed parameters are shown in Table 2.1.
The commissioning was executed at the experimental surface hall during the second
quarter of 2006.

3 The momentum resolution is proportional to p/(B R 2 ), where B is the magnetic field in Tesla, and
R indicates the radius of the charged-particle curvature in meter.
12 2 Experimental Apparatus

Fig. 2.4 The CMS tracker layout. The outer radius is around 110 cm, and the total length is approx-
imately 540 cm

2.2.2 Tracking Detectors

In terms of various radii, the tracker shown in Fig. 2.4 within a range of |η| < 2.5
can be divided into three regions coping with the different charged-particle flux. In
the innermost region (r ≈ 10 cm), since the particle flux closest to the interaction
vertex is the highest as 107 /s, pixel detectors with the fine-granularity size of ≈
100 × 150 µm2 /pixel are installed. In the intermediate region (20 < r < 55 cm),
the particle flux becomes lower to allow use of silicon microstrip detectors with a
minimum cell size of 10 cm × 80 µm. In the outermost region (r > 55 cm), the
particle flux has sharply decreased, and then use of larger-pitch silicon microstrips
with a maximum cell size of 25 cm × 180 µm is enough. With the design of the
inner tracker system, a measurement of the charged-particle trajectory and vertices
reconstructions can be performed. The Fig. 2.5 displays the material budget of the
CMS tracker in units of radiation length (X 0 ) [36].4 Taking muons with pT = 1 GeV/c
as an example, the pT resolution (δpT / pT ) shown in Fig. 2.6 has a good performance
of ∼1% [36]. The (δd0 ) is around 100 µm, and the longitudinal part (δz0 ) is between
100 and 1000 µm.

Pixel Tracker

The pixel detector comprise three barrel layers and two endcap disks on each side
shown in Fig. 2.7. The three barrel layers are located at mean radii of 4.4, 7.3, and
10.2 cm with a length of 53 cm. The two endcap disks, covering from 6 to 15 cm in
radius, are placed on each side at |z| = 34.5 and 46.5 cm in a turbine-like geometry
with blades rotated by 20◦ .5

4 The radiation length (X

0 ) stands for a characteristic of a material, related to the energy loss of high
energy electromagnetic-interacting particles in the material.
5 It’s to benefit from the Lorentz effect.
2.2 The CMS Detector 13

Fig. 2.5 Material budget of the CMS tracker in units of radiation length X 0 as a function of |η| for
different detector types

Fig. 2.6 Resolution of track pT (left), d0 (center), and z0 (right) with muon pT of 1, 10, 100 GeV/c

Strip Tracker

The strip tracker in the barrel region is composed of two parts including a TIB
(Tracker Inner Barrel) and a TOB (Tracker Outer Barrel). The TIB with a half-length
of |z| < 65 cm is made of four layers, whose first two layers consist of ‘stereo’
modules in order to provide a measurement in both r-φ and r-z coordinates. The
detailed parameters for different detector types are presented in Table 2.2. The TOB
with a half-length of |z| < 110 cm is made of six layers, whose first two layers also
provide a ‘stereo’ measurement.
14 2 Experimental Apparatus

Fig. 2.7 Layout of pixel detectors in the CMS tracker

Table 2.2 Detector types in the silicon tracker

Part No. detectors Thickness (µm) Mean pitch (µm) r−φ(z) resolution
(µm)
TIB 2724 320 81/118 23–34(230)
TOB 5208 500 81/183 35–52(530)
TID 816 320 97/128/143
TEC 2512 320 96/126/128/143
TEC(2) 3888 500 143/158/183

In the endcap region, the strip tracker can be separated into the TEC (Tracker
End Cap) and TID (Tracker Inner Disks). Each TEC, extending to the region 120 <
|z| < 280 cm, comprises nine disks whose first two and fifth rings are made of ‘stereo’
modules. Each TID, filling the gap between the TIB and the TEC, consists of three
small disks whose first two rings have ‘stereo’ modules.

2.2.3 Electromagnetic Calorimeter (ECAL)

The ECAL is designed to measure energy depositions coming from electromagnetic-

interacting particles. In order to achieve this goal, the ECAL is made of lead
tungstate (PbWO4 ) crystals, which can provide a short radiation length (0.89 cm),
small Moliere radius (2.2 cm),6 a fast response time (emissions of 80% of the light

6 The Moliere radius is a characteristic constant of a material, indicating the radius of a cylinder
containing on average 90% of the shower’s energy deposition.
2.2 The CMS Detector 15

Fig. 2.8 Layout of the CMS Electromagnetic Calorimeter system (ECAL)

within 25 ns) and high radiation resistance (up to 10 Mrad).7 This sub-detector is
hermetic and homogeneous with a large coverage in pseudorapidity (|η| < 3), and
divided into barrel and endcap parts as illustrated in Fig. 2.8. The ECAL barrel (EB)
in the region of |η| < 1.479 consists of 61200 crystals with a crystal front face of
22 × 22 mm2 , and the ECAL endcap (EE) in the rest region comprises 7324 crystals,
whose front face is 28.6 × 28.6 mm2 . Thus, a design of a fine granularity and a com-
pact calorimeter is ensured. A barrel crystal is 23 cm long corresponding to 25.8 X 0 ,
while the length of a endcap crystal is 22 cm equal to 24.7 X 0 . The scintillation light
emitting from the crystals is received by avalanche photodiodes (APD)8 with an
active area of 5 × 5 mm2 in the barrel, and by vacuum phototriodes (VPT)9 with an
active area of 280 mm2 in the EE. The readout has a strong temperature dependence,
and this effect induces a challenge to the ECAL energy calibration.
A special-purpose sub-detector, prEShower (ES) system, placed in front of the
EE crystal is designed to have a good spatial resolution to measure the position of
incoming particles, and aid particle identification in the EE. Furthermore, the ES
can distinguish a single photon from two closely-separated photons so that rejecting
neutral pions (π 0 ) can be performed. In each endcap, there are two orthogonal layers
of 1.9 mm-pitch silicon sensors, each preceded by thin layers of lead (3 X 0 ) that
initiate electromagnetic showers.
The ECAL energy resolution can be parametrized by the following expression:

 σ 2  2  2
S N
= √ + + C2 , (2.1)
E E E

7 The rad is a unit of absorbed radiation dose.

8 It’s
a highly sensitive semiconductor electronics which can convert light to electricity.
9 VPT is the first batches of series production used in the CMS. The VPT is more fast and sensitive

than the APD since the radiation levels are highest in the EE.
16 2 Experimental Apparatus

Fig. 2.9 ECAL energy resolution as a function of the energy, which is measured in an electron test
beam. The central value of the stochastic (S), noise (N), and constant (C) terms are shown in the
legend

where S is the stochastic term, N stands for the noise term, and C represents the
constant term. In an electron test beam, the parameters of ECAL energy resolution
1
were measured to be S = 0.028 GeV 2 , N = 0.12 GeV, and C = 0.003 shown in
Fig. 2.9 [36].

2.2.4 Hadronic Calorimeter (HCAL)

The energy measurement in the calorimeter system is completed using the ECAL and
the hadronic calorimeter (HCAL), which is designed to measure energies coming
from hadrons, jets, or neutrinos. The HCAL is a sampling calorimeter which com-
prises alternating layers of massive absorbing brass plates providing a short inter-
action length (λl ),10 and plastic scintillator tiles, which embed wavelength-shifting
fibers.11 The photodetection readout is based on multi-channel hybrid photodiodes
(HPDs). The first scintillators in front of the first absorber plate is used to sample

10 It indicates the mean free path, meaning the average distance between collisions.
11 It shifts the blueviolet light emitted by the scintillator to green light which is sent to HPDs.
2.2 The CMS Detector 17

Fig. 2.10 Layout of the CMS Hadronic Calorimeter system (HCAL)

Table 2.3 Tower segmentation in φ and η for the hadronic barrel, endcap, and forward calorimeter
HB/HO HE |η| ≤ 2.5 HE |η| > 2.5 HF |η| ≤ 4.7 HF |η| > 4.7
δφ × δη 0.087 × 0.087 0.087 × 0.087 0.175 × 0.175 0.175 × 0.175 0.175 × 0.35

showers producing in the material, and the last scintillators installed in back of the last
absorber plate ensures if the late producing showers leak out. The HCAL is divided
into three components, including the barrel and outer hadronic calorimeter (HB and
HO) in the barrel region of |η| < 1.3, the endcap part (HE) at 1.3 < |η| < 3, and the
forward sub-detector (HF) extending up to |η| < 5.2 as presented in Fig. 2.10.
In the barrel region, hadron showers can not be completely absorbed by the EB plus
HB. Therefore, the HB is supplemented by the HO placed between the solenoid and
the muon detectors, while the solenoid is regarded as additional absorbing material
providing sufficient containment with a thickness of 11.8 λl . The granularity of the
calorimeters in δη × δφ is summarized in Table 2.3.
In the forward region, the HF placed at 11.2 m starting from the interaction point is
designed to measure the energetic forward jets12 using steel absorber plates, which
is used to detect the Cherenkov light.13 Thus, signals coming from electrons and
photons can be separated from signals generated by hadrons by considering a longi-
tudinal segmentation.

12
√For example, there is an energy of 760 GeV deposited on average in proton–proton collisions at
(s) = 14 TeV.
13 The Cherenkov light is emitted by a particle when its speed is fast more than the speed of light in

the material.
18 2 Experimental Apparatus

2.2.5 Jet Reconstruction at CMS

In this subsection, we will summarize the basic concepts employed to reconstruct

jets in the CMS calorimeters, where calorimeter towers are subject to jet algorithms.
Other analyses use calorimeter energies together with track momenta or individually
reconstructed particles as input to the jet algorithm.

2.2.5.1 Tower Definition and Thresholds

Calorimeter towers combine the measurements of ECAL and HCAL in appropriate

(η, φ) bins. Due to the much finer granularity of the ECAL, tower sizes follow
the HCAL segmentation of Δη × Δφ = 0.087 × 0.087 in the barrel region (and
correspondingly larger in the endcap and forward regions). Towers thus follow a
pattern in (η, φ) space, which is projective to the nominal interaction point. Tower
energies are obtained from summing all contributing cells with energies above pre-
configured thresholds, typically of the order of E T = 0.5 GeV. In total, 4176 towers
are clustered by a number of jet algorithms, where towers are treated as massless
particles with energies equal to the tower energy and emerging from the nominal
interaction point to the center of the tower.

2.2.5.2 Jet Algorithms

Jets are reconstructed offline from the energy deposits in the calorimeter towers,
clustered by the anti-kt algorithm [27, 28] with a size parameter of 0.5. In this
process, the contribution from each calorimeter tower is assigned a momentum,
the absolute value and the direction of which are given by the energy measured in
the tower, and the coordinates of the tower. The raw jet energy is obtained from
the sum of the tower energies, and the raw jet momentum by the vectorial sum of
the tower momenta, which results in a nonzero jet mass. The raw jet energies are
then corrected to establish a relative uniform response of the calorimeter in η and a
calibrated absolute response in transverse momentum pT .

2.2.5.3 Jet Energy Correction at CMS

CMS has developed a factorized multi-level approach to relate measured calorimeter

jet energies to final-state-particle jet energies. Note that any correction beyond par-
ticle level is ambiguous: Different hadronization models and/or different underlying
event descriptions result in a range of possible predictions.
Calorimeter cells have been calibrated using test beam data. The approaches
described in the following calibrate calorimeter cells in situ, i.e. without requiring
dedicated runs or measurements. Typically, jet energy corrections are determined in
2.2 The CMS Detector 19

2 → 2 processes, where both final-state objects must conserve transverse momentum

( pT -balance). pT -balance-techniques are employed in di-jet events, γ +jet-events and
Z +jet-events [37]. A schematic equation to determine the corrected jet energy reads:
Corrected Jet Energy = (Measured Jet Energy - Offset) × (Relative η Correction)
× (Absolute pT Correction) [× EMF Correction] [× Flavor Correction], where the
squared brackets indicate optional corrections.
Jet momentum is determined as the vectorial sum of all particle momenta in the
jet, and is found from simulation to be within 5 to 10% of the true momentum
over the whole pT spectrum and detector acceptance. An offset correction is applied
to jet energies to take into account the contribution from additional proton–proton
interactions within the same bunch crossing [38]. Jet energy corrections are derived
from simulation, and are confirmed with in situ measurements of the energy balance
in di-jet and photon+jet events. Additional selection criteria are applied to each event
to remove spurious jet-like features originating from isolated noise patterns in certain
HCAL regions. The jet energy resolution amounts typically to 15% at 10 GeV, 8% at
100 GeV, and 4% at 1 TeV, to be compared to about 40, 12, and 5% obtained when
the calorimeters alone are used for jet clustering.

2.2.6 Muon System

The muon system, situated at the outermost part of the CMS detector, consists of
three kinds of gaseous detector chambers in order to identify muon and measure
momentum. The coverage of the muon system reaches up to |η| = 2.4. In the barrel
region, drift tube (DT) chambers are adopted since the muon rate and the neutron
induced background are both small under a lower magnetic field. On the other hand,
the muon and the background flux is much higher in the endcap region, and then
cathode strip chambers (CSCs) are used because of a faster response, a higher gran-
ularity, and a better resistance against radiation. Besides, resistive plate chambers
(RPCs) construct a redundant trigger system. In total, the muon system possesses
250 DT chambers, 540 CSCs, and 610 RPCs, and the overview is shown in Fig. 2.11.
For the performance of the muon system, the momentum resolution (Δp/ p) only
based on the muon system shown in Fig. 2.12 can be improved by considering the
inner tracker because the effects from multiple scattering and energy loss are not
included [36].

DT Chamber

Each DT chamber comprises 12 layers of DT cells, which is a 4 cm wide gas tube

with a positively charged-stretched wire inside. The maximum drift length is 2.0 cm
with a single-point resolution of ∼200 µm.
20 2 Experimental Apparatus

Fig. 2.11 Layout of the CMS muon system

Fig. 2.12 The muon momentum resolution (Δp/ p) as function of the muon momentum for the
muon system only, the inner tracker only, or both (‘full system’) with a region of |η| < 0.2 (left)
and 1.8 < |η| < 2.0 (right)

CSC

The CSCs are multiwire proportional chambers with a trapezoidal shape. This type
provides a two-dimensional position measurement, where the r and φ coordinates
are determined by the copper strips and the anode wires, respectively.
2.2 The CMS Detector 21

RPC

The RPCs are composed of two high resistive plastic plates with a bias, and separated
by a gas volume. This type providing a fast response with a time resolution of 1 ns is
used in the muon trigger system. The RPCs are sandwiched between DT chambers
or CSCs.

2.2.7 Global Event Reconstruction

The global event reconstruction (also called particle-flow event reconstruction [39,
40]) consists in reconstructing and identifying each single particle with an optimized
combination of all subdetector information. In this process, the identification of
the particle type (photon, electron, muon, charged hadron, neutral hadron) plays an
important rôle in the determination of the particle direction and energy. Photons (e.g.
coming from π 0 decays or from electron bremsstrahlung) are identified as ECAL
energy clusters not linked to the extrapolation of any charged particle trajectory to
the ECAL. Electrons (e.g. coming from photon conversions in the tracker material or
from b-hadron semileptonic decays) are identified as a primary charged particle track
and potentially many ECAL energy clusters corresponding to this track extrapolation
to the ECAL and to possible bremsstrahlung photons emitted along the way through
the tracker material. Muons (e.g. from b-hadron semileptonic decays) are identified
as a track in the central tracker consistent with either a track or several hits in the
muon system, associated with an energy deficit in the calorimeters. Charged hadrons
are identified as charged particle tracks neither identified as electrons, nor as muons.
Finally, neutral hadrons are identified as HCAL energy clusters not linked to any
charged hadron trajectory, or as ECAL and HCAL energy excesses with respect to
the expected charged hadron energy deposit.
The energy of photons is directly obtained from the ECAL measurement, cor-
rected for zero-suppression effects. The energy of electrons is determined from a
combination of the track momentum at the main interaction vertex, the correspond-
ing ECAL cluster energy, and the energy sum of all bremsstrahlung photons attached
to the track. The energy of muons is obtained from the corresponding track momen-
tum. The energy of charged hadrons is determined from a combination of the track
momentum and the corresponding ECAL and HCAL energy, corrected for zero-
suppression effects and for the response function of the calorimeters to hadronic
showers. Finally, the energy of neutral hadrons is obtained from the corresponding
corrected ECAL and HCAL energy.
In the barrel section of the ECAL, energy resolutions as good as 1% are achieved
for unconverted or late-converting photons in the tens of GeV energy range. The
remaining barrel photons have a resolution of about 1.3% up to a pseudorapidity of
|η| = 1, rising to about 2.5% at |η| = 1.4. In the endcaps, the resolution of uncon-
verted or late-converting photons is about 2.5%, while the remaining endcap photons
have a resolution between 3 and 4% [41].
22 2 Experimental Apparatus

2.3 Trigger System

The CMS trigger system is designed to deal with the unprecedented high interaction
rate ∼109 Hz, caused by the bunch crossing rate of 40 MHz at nominal luminosity.
The data recording rate should be reduced to the order of a few 100 Hz, which allows
for permanent storage of an event. The trigger system is divided into two steps,
including the Level-1 trigger (L1) providing a fast decision with custom electronics,
and the High Level Trigger (HLT) system relying on commercial processors based
on software algorithms. The flow chart is displayed in Fig. 2.13 [42].

Fig. 2.13 CMS Trigger system

2.3 Trigger System 23

2.3.1 Level-1 Trigger (L1)

The L1 trigger latency, allocated for the transit and for archiving a decision from a
bunch crossing, is 3.2 µs. It turns out that the L1 trigger calculation must be done
in less than 1 µs. The L1 trigger is decided using information from the calorimeters
plus the muon system, and  the presence of trigger primitive objects such as photons,
missing
electrons, muons, jets, E T , and E T . Once an event is triggered by L1, the
full detector information (∼1 MB) is transferred to the processor farms (also called
‘event filter farms’) via the data acquisition (DAQ) system at a rate of up to 100 kHz.

2.3.2 High Level Trigger (HLT)

Events, passing through the L1 trigger, are proceeding with the HLT. A data recording
rate is limited up to few 100 Hz, while only data accepted by the HLT is recorded for
offline physics analysis. Traditionally, this is performed in two stages, involving a
hardware/software-based Level-2 trigger quickly providing a large rejection factor,
and a Level-3 processor farm based on more sophisticated algorithms.

2.4 Luminosity Measurement

A significant component of the LHC physics program includes precision measure-

ments associated with an accurate luminosity determination. The primary luminosity
is measured using the online HF system, and absolutely calibrated with Van der Meer
(VdM) scans [43] offline. The luminosity in a colliding beam accelerator can read
as,
N1 N2 νor bit n b
L0 = , (2.2)
Ae f f

where L0 means the peak instantaneous luminosity, Ni is the bunch intensity in

beam i,14 νor bit represents the LHC orbit frequency of 11, 246 Hz,15 n b stands for
the number of colliding bunches, and Ae f f expresses the effective overlap area of
the colliding beams. N1 or 2 , and νor bit are well-known, and n b can be measured with
the beam current transformer.16 The only unknown parameter, Ae f f , is obtained by
scanning the beams through one another in the transverse directions under the VdM
approach.

14 Inproton–proton collisions, it can be regarded as the number of protons in the beam.

15 ν can be obtained from lor bit × νor bit = c, where lor bit is the LHC ring (∼26.7 km), and c
or bit
indicates the speed of light.
16 The beam current transformer measures the current in each 25-ns LHC bunch.
24 2 Experimental Apparatus

2.4.1 The Primary Luminosity

Two methods for extracting a relative instantaneous luminosity with the online HF
system have been studied. The first method based on zero counting infers the mean
number of interactions per bunch crossing according to the average fraction of empty
towers. In the second method, the linear relationship between the average transverse
energy per tower and the luminosity is exploited. However, one unexpected result
first observed in 2011 running is that the HF scale is non-linear for single-bunch
luminosities due to the pile-up effect. Thus, the zero-counting method is mainly
applied for the primary luminosity measurement.

2.4.2 The Absolute Luminosity

The separation scan method for determining the absolute luminosity was first pio-
neered by S. Van Der Meer at the Intersecting Storage Rings (ISR) [44]. The beam
size can be measured using the relative interaction rate as a function of the transverse
beam separations. Once the beam profile in x and y is assumed to be given by the
function F(x, y) = f x (x) f y (y), the instantaneous luminosity can be re-written as,
N1 N2 νor bit n b F(0, 0)
L0 =   , (2.3)
f x (Δx) dΔx f y (Δy) dΔy

where Δx and Δy are the beam separations in the x (horizontal) and y (verti-
cal) planes, respectively. Besides, double-Gaussian distributions are adopted for the
functions f x and f y . It results in the instantaneous luminosity with beam profiles
expressed as,
N1 N2 νor bit n b
L0 ≡ ef f ef f
, (2.4)
2π σx σy

ef f
where σ j means the effective beam size for each scan plane j.

2.5 Computing

All data has to be stored for the lifetime of the experiment. In particular, data analy-
sis of all running periods has to be supported at all times. Consequently, transfer,
storage and analysis of large data sets is foreseen by the CMS Computing Model.
Real-time (raw) detector data passes pattern recognition, filtering and data reduction
steps. Physics analysis activities are carried out regardless of the physical loca-
tion of the physicist at several computing centers distributed throughout the world.
2.5 Computing 25

Furthermore, theory predictions have to be simulated and distributed on a large scale

as well to enable physicists to compare the experimental data with theory predic-
tions. Run conditions, calibration information, and additional non-event data are
made accessible via databases.
The CMS off-line computing model comprises a fully distributed system through-
out worldwide computing centers. Data storage, processing, and analysis will mainly
happen outside CERN at designated sites. The Worldwide LHC Computing Grid
(WLCG) provides the necessary services to the LHC experiments to maintain such
an infrastructure. The CMS computing model’s key components shall be described
in the following.
Event processing, filtering and analysis are applications common to off-line and
on-line computing. The aim of the CMS application framework is to provide well-
defined interfaces which decouple physics analyses from event I/O, user interfaces
and other constraints. This is accomplished by interfacing software modules to one
standard application. All software modules may filter, analyze, or modify the central
entity of the CMS data model, the Event. Communication between modules also
happens exclusively through the Event. A user who wishes to process data (real or
simulated) will have to define a sequence of modules to be executed. The application
framework hides complexity such as accessing global services through database
interfaces, etc.
Several event formats with well-defined data content exist which are typical for
different processing steps of a data sample:

RAW: This type of event content consists of the entire detector information, includ-
ing trigger decisions and meta data, for instance on the run conditions. According to
the trigger signature, RAW data is classified into distinct primary datasets.
RECO: Reconstructed events are obtained after RAW data has been passed through
pattern recognition and compression algorithm steps. Thus, in this event content,
high-level physics objects are contained, including all basic detector input from
which they were reconstructed.
AOD: The Analysis Object Data event content consists of high-level physics objects
together with information needed to refit the kinematics. AOD data is obtained from
RECO data by means of filtering.
In addition, event reconstruction relies on information about non-event data, e.g.
construction data, equipment management data, configuration data, and conditions
data. Conditions data includes information on alignment and calibration constants
as well as information on the detector status.
The CMS computing needs exceed all presently available resources of a single
site. Computing resources provided by collaborating institutes worldwide are thus
integrated into a common hierarchical architecture of Tiered centers: One Tier-0
center at CERN, six Tier-1 centers at large computing facilities and approximately
25 Tier-2 centers at medium to large computing facilities. The data ow between tiers
is sketched in Fig. 2.14 and described in the following [36].
26 2 Experimental Apparatus

Fig. 2.14 Data flow between CMS computing Centres

The CERN Tier-0 center records experimental data at a mass storage system,
promptly reconstructs RAW data into first-pass RECO datasets, and exports a copy of
the RAW data to Tier-1 centers for re-reconstruction. The Tier-1 centers are operated
around the clock and provide large computing farms, dedicated mass storage, and
fast network links to each other and to the Tier-0. RAW data are stored permanently
to offer reconstruction whenever necessary, for instance if improved alignment and
calibration constants are available. In addition, skimming and data-intensive analysis
applications run on the Tier-1 with the output being transferred to associated Tier-2
centers. These centers support the bulk of user analysis and thus have to grant access
to large computing farms to all CMS users with a valid grid certificate. Off-line
calibration and alignment studies as well as detector analyses are typically performed
at the Tier-2 centers as well. Thirdly, Tier-2s produce large amounts of simulated
data which are transferred to the associated Tier-1 center for storage. In order to
provide fast feedback on key analyses, the CERN Analysis Facility (CAF) provides
large CPU resources together with fast access to the full data.
The Worldwide LHC Computing Grid provides an infrastructure for job submis-
sions and remote data access. It does so, while ensuring robust security and account-
ing. Nonetheless, some CMS-specific software is needed on sites. The Dataset Book-
keeping System catalogues existing event data. Existing event data is mapped to its
physical location or locations (if more than one copy of a dataset exists) with the
help of the Data Location Service. Each site employs a Local File Catalogue which
translates a logical file name into the actual path in the respective file system. CMS-
specific tools for large scale data transfers schedule, monitor and verify the data
transfer. Very large-scale data processing is performed with specialized tools, com-
prised of automated job managers. Since remote analysis introduces additional layers
of complexity to a physics analysis, dedicated tools have been developed to facilitate
job submission, monitoring, and retrieval [45].
Chapter 3
Data and Monte Carlo Simulation

This analysis is based on the full 2010 data sample, collected with the CMS
detector and reprocessed with the CMS reconstruction software. The full data sam-
ple processed for this analysis amounts to 36.1 ± 1.4 pb-1 . A detailed list of the
reprocessed data sets can be found in Table 3.1. Only certified good runs are used
in the official list with respect to specific un-prescaled single photon triggers, which
are summarized in Chap. 4. The treatment of pile-up effect in data is also described
in Chap. 4.
Simulated event samples for single photon production associated with 3 jets
are produced with different MC event generators: pythia 8 (version 8.165) [46],
pythia 6 (version 6.426) [47], and MadGraph 5 (version 5.1.1.0) [48–50] in com-
bination with pythia interface. Multiple parton interactions (MPI) are simulated in
both pythia 6 and pythia 8. The modeling of MPI simulation makes use of several
adjustable parameters in the underlying event tune. MPI-related parameters of base
tunes for LHC experiments are studied in Reference [51]. All simulated event sam-
ples are processed and reconstructed in the same manner as done for the collision
data. The full detector response is simulated with geant4 [52, 53]. MC samples
from different generators are used to compare predictions at both stable-particle and
detector level, which can be also used for unfolding process of data.
pythia 6 and pythia 8 are leading order (LO) MC event generators with 2 →
2 matrix element (ME) calculations. The pythia 6 generator with the tune Z2 [54]
uses LO CTEQ6L1 [55] parton distribution functions (PDF) and applies a new MPI
model [56]. The pythia 8 generator with the tune 4C [57] also uses LO CTEQ6L1
PDFs and implements a similar MPI model with respect to pythia 6. It tunes the
free parameters to the underlying event data obtained at the LHC and considers color
reconnection and rescattering between the partons [58]. The pythia 8 generator can
also be used to study the DPS signal distribution by generating two hard scatterings in
the same proton-proton collision. The MadGraph 5 generator with LO CTEQ6L1
PDFs contains LO ME calculations up to 2 → 5 processes. Events generated by

© Springer Nature Singapore Pte Ltd. 2017 27

Y.-H. Chang, Study of Double Parton Scattering in Photon + 3 Jets Final State,
Springer Theses, DOI 10.1007/978-981-10-3824-2_3
28 3 Data and Monte Carlo Simulation

Table 3.1 Summary of analyzed data sets. The full data sets correspond to 36.1 pb-1
Run ranges Dataset
136035–144114 /EG/Run2010A-Apr21ReReco-v1/AOD
146428–149294 /Photon/Run2010B-Apr21ReReco-v1/AOD

Table 3.2 Summary of pythia-8 MC samples used in the analysis at detector level. All MC samples
listed here are produced using pythia-8 generator with the 4 C tune. The cross section values are
all LO predictions
pythia-8, γ +jet 4 C tune

MC set σ , pb # events L dt, pb-1 pˆT range, GeV/c
PhotonJet_Pt30to50_7TeV 1.80×104 2.5×105 13.9 30–50
PhotonJet_Pt50to80_7TeV 2.90×103 1.5×105 51.7 50–80
PhotonJet_Pt80to120_7TeV 470 1.0×105 213 80–120
PhotonJet_Pt120to170_7TeV 89.0 1.0×105 1.12×103 120–170
PhotonJet_Pt170to300_7TeV 22.6 5.0×104 2.21×103 170–300
PhotonJet_Pt300to470_7TeV 1.49 5.0×104 3.35×104 300–470
pythia-8, QCD 4C tune

MC set σ , pb # events L dt, pb-1 pˆT range, GeV/c
QCD_Pt30to50_7TeV 6.10×107 1.1×106 1.81×10−2 30–50
QCD_Pt50to80_7TeV 7.22×106 1.1×106 1.53×10−1 50–80
QCD_Pt80to120_7TeV 8.83×105 1.1×106 1.23 80–120
QCD_Pt120to170_7TeV 1.29×105 8.5×105 6.6 120–170
QCD_Pt170to300_7TeV 2.70×104 8.5×105 31.7 170–300
QCD_Pt300to470_7TeV 1.29×103 5.3×105 412 300–470

means of the MadGraph generator are interfaced with pythia 6, tune Z2 and tune
Z2* [56, 59], for hadronization, parton showering, and MPI simulation.
The data are also compared to the multijet-improved MC event generator, sherpa
2.1.0 [60–64] with next-to-leading order (NLO) CT10 [65] PDFs and tune CT10 [60]
is considered. sherpa contains particle production at tree level with merging between
ME and parton shower (PS) (2 → 2 + N, where N = 1 in this analysis). The pre-
diction of MadGraph, interfaced with pythia 8 tune 4C, is also included for the
comparisons at stable-particle level. We also preform a deep investigation of LO
and NLO cross section of different MCs in Appendix A. The lists of fully simulated
samples are presented in Tables 3.2, 3.3, 3.4, 3.5, 3.6 and 3.7.
For computation reason, the productions of all MC samples are organized by
splitting the different processes in several bins. For pythia and sherpa productions, a
generation in slices according to the transverse momentum exchanged in the partonic
3 Data and Monte Carlo Simulation 29

Table 3.3 Summary of pythia-6 MC samples used in the analysis at detector level. All MC samples
listed here are produced using pythia-6 generator with the Z2 tune. The cross section values are
all LO predictions
pythia-6, γ +jet Z2 tune

MC set σ , pb # events L dt, pb-1 pˆT range, GeV/c
PhotonJet_Pt30to50_7TeV 1.67×104 2.19×106 131 30–50
PhotonJet_Pt50to80_7TeV 2.72×103 2.04×106 748 50–80
PhotonJet_Pt80to120_7TeV 447 2.05×106 4.58×103 80–120
PhotonJet_Pt120to170_7TeV 84.2 2.09×106 2.48×104 120–170
PhotonJet_Pt170to300_7TeV 22.6 2.07×106 9.14×104 170–300
PhotonJet_Pt300to470_7TeV 1.49 2.08×106 1.39×106 300–470
pythia-6, QCD Z2 tune

MC set σ , pb # events L dt, pb-1 pˆT range, GeV/c
QCD_Pt30to50_7TeV 5.31×107 6.58×106 1.24×10−1 30–50
QCD_Pt50to80_7TeV 6.36×106 6.60×106 1.04 50–80
QCD_Pt80to120_7TeV 7.84×105 6.59×106 8.4 80–120
QCD_Pt120to170_7TeV 1.15×105 6.13×106 53.2 120–170
QCD_Pt170to300_7TeV 2.43×104 6.22×106 256 170–300
QCD_Pt300to470_7TeV 1.17×103 6.43×106 5.51×103 300–470

Table 3.4 Summary of MadGraph MC samples used in the analysis at detector level. All MC
samples listed here are generated by MadGraph and passed to pythia-6 Z2 tune hadronizer for
hadronization and parton showering. The cross section values are all LO predictions
MadGraph, γ +jet, with pythia-6 Z2 tune Hadronizer

MC set σ , pb # events L dt, pb-1 HT range, GeV/c
PhotonJet_Ht40to100_7TeV 2.57×104 1.27×107 496 40–100
PhotonJet_Ht100to200_7TeV 5.21×103 1.54×106 295 100–200
PhotonJet_Ht200toinf_7TeV 798.3 9.38×106 1.17×104 200–∞
MadGraph, QCD, with pythia-6 Z2 tune Hadronizer

MC set σ , pb # events L dt, pb-1 HT range, GeV/c
QCD_Ht100to250_7TeV 4.19×106 2.26×107 5.38 100–250
QCD_Ht250to500_7TeV 1.99×105 2.07×107 104 250–500
QCD_Ht500to1000_7TeV 5.86×103 1.44×107 2.47×103 500–1000
QCD_Ht1000toinf_7TeV 122.6 6.29×106 5.13×104 1000–∞
30 3 Data and Monte Carlo Simulation

Table 3.5 Summary of MadGraph MC samples used in the analysis at stable-particle level. All
MC samples listed here were generated by MadGraph and passed to pythia-6 Z2* tune hadronizer
for hadronization, parton showering and multiple parton interaction
MadGraph, γ +jet, with pythia-6 Z2* tune Hadronizer with MPI

MC set σ , pb # events L dt, pb-1 HT range, GeV/c
PhotonJet_Ht40to100_7TeV 2.55×104 1.17×107 460.4 40–100
PhotonJet_Ht100to200_7TeV 5.30×10 3 1.54×106 290.0 100–200
PhotonJet_Ht200toinf_7TeV 779.4 9.79×106 1.26×104 200–∞
MadGraph, QCD, with pythia-6 Z2* tune Hadronizer with MPI

MC set σ , pb # events L dt, pb-1 HT range, GeV/c
QCD_Ht100to250_7TeV 4.17×106 1.54×107 3.68 100–250
QCD_Ht250to500_7TeV 1.99×105 2.14×107 108.0 250–500
QCD_Ht500to1000_7TeV 5.82×103 1.49×107 2.56×103 500–1000
QCD_Ht1000toinf_7TeV 122.5 7.84×106 6.10×104 1000–∞

Table 3.6 Summary of MadGraph MC samples used in the analysis at stable-particle level. All
MC samples listed here were generated by MadGraph and passed to pythia-8 4 C tune hadronizer
for hadronization, parton showering and multiple parton interaction
MadGraph, γ +jet, with pythia-8 4 C tune Hadronizer with MPI

MC set σ , pb # events L dt, pb-1 HT range, GeV/c
PhotonJet_Ht40to100_7TeV 2.33×104 1.07×107 460.3 40–100
PhotonJet_Ht100to200_7TeV 4.67×10 3 1.35×10 6 290.0 100–200
PhotonJet_Ht200toinf_7TeV 624.9 7.85×106 1.26×104 200–∞
MadGraph, QCD, with pythia-8 4 C tune Hadronizer with MPI

MC set σ , pb # events L dt, pb-1 HT range, GeV/c
QCD_Ht100to250_7TeV 3.27×106 1.21×107 3.68 100–250
QCD_Ht250to500_7TeV 1.51×105 1.63×107 108.0 250–500
QCD_Ht500to1000_7TeV 4.60×103 1.18×107 2.56×103 500–1000
QCD_Ht1000toinf_7TeV 98.3 6.00×106 6.10×104 1000–∞

Table 3.7 Summary of sherpa tune CT10 MC samples used in the analysis at stable-particle level
sherpa, γ +jet

MC set σ , pb # events L dt, pb-1 pˆT range, GeV/c
PhotonJet_Pt45_7TeV 5.42×103 5.00×106 922.5 45–∞
sherpa, QCD

MC set σ , pb # events L dt, pb-1 pˆT range, GeV/c
QCD_Pt55_7TeV 6.31×106 1.20×109 190.5 55–∞
3 Data and Monte Carlo Simulation 31

interaction ( pˆT ) is processed. And, for MadGraph production, generation slices are
based on the scalar summation of the transverse momentum of all the jets (HT ).
In this analysis, the contributions of different bins are combined together. Accord-
ing to the corresponding cross section and number of generated events, each bin is
weighted to the effective integrated luminosity of 2010 data sets, 36.1 pb-1 . The data
is then compared with different combined MCs of photon+jet and QCD samples. It
must be noted that for the QCD processes, the size of the MC samples corresponds
to an integrated luminosity much lower than the one analyzed in data, especially,
for pythia-8 samples. Although after applying all the selection designed for this
analysis, which would be introduced in the following section, there is no any con-
tribution from pˆT range between 30 to 50 GeV/c for both pythia-8 and pythia-6.
The sample size of MadGraph production is relatively sufficient and is mainly used
for detector level comparisons in this analysis and also the unfolding procedure for
correcting data. The related results pythia-8 and pythia-6 is taken as comparison
between different theoretical models. All the data sets and MC samples are analyzed
using the cmssw framework with recommended global setting.
Chapter 4
Event Selection

The analyzed data sample corresponds to approximately √ 36 pb−1 of integrated lumi-

nosity, recorded by the CMS detector in 2010 at s = 7 TeV with low-pileup con-
ditions. The mean value of overlapping proton-proton interactions ranges between
1.5 and 3 for the considered data. All the MC samples include the pileup simulation
and are reweighted to match with the pileup distribution measured in data.
At least four tracks
 2are required for the reconstructed vertex fit. The vertex with the
largest value of pT i−th track is selected as the primary vertex, where pT i−th track is
the transverse momentum of the i-th track associated to the vertex. The reconstructed
primary vertices are required to have a longitudinal position, z, within ±24 cm along
the proton beam axis (z-direction) and a radial position within 2 cm of the nominal
interaction point (beamspot). In addition, in order to filter off beam-scraping events,
at least 10 tracks with pT > 0.25 GeV were required to be present in each event.
The triggers used to collect the analyzed data sample require the presence of at
least one reconstructed electromagnetic cluster, with transverse energy (E T ) trigger
thresholds from 20 to 70 GeV. Events with one photon and at least three jets are
selected. Reconstructed photon candidates are required to have pT > 75 GeV to
assure a fully efficient trigger, and |η| < 1.4442 and 1.566 < |η| < 2.5 to avoid
instrumental inefficiencies of the calorimeter.
Events passing the trigger and vertices selections are also required to have at
least one reconstructed photon candidate fulfilling isolation criteria and at least three
jets passing the basic qualities selection. The inner tracking system and calorimeters
are the primary detector used to reconstruct photons and jets. Details of the related
selection criteria for the photons and the jets are given in this chapter. In order to keep
enough events for analyzing, loose photon and jet identification criteria are beneficial
for this analysis. The photon purity and also the influences of different background
events are studied in the analysis.

© Springer Nature Singapore Pte Ltd. 2017 33

Y.-H. Chang, Study of Double Parton Scattering in Photon + 3 Jets Final State,
Springer Theses, DOI 10.1007/978-981-10-3824-2_4
34 4 Event Selection

4.1 Trigger Requirement

Due to the importance of isolated photon requirement of double parton scattering

study in photon + 3 jets final state, photon candidate events used in this analysis are
selected by two-level trigger system implemented in the CMS detector. Photon are
reconstructed in the CMS detector through the energy deposited in the Electromag-
netic Calorimeter (ECAL). Dedicated algorithms are designed in the CMS trigger
system to select events with large energy depositions in the ECAL. In the Level-1
trigger system (L1), electromagnetic energy deposits are reconstructed through the
sum of the transverse energy (E T ) deposited in two neighboring groups of 5 × 5
ECAL crystals (trigger towers). The E T sum of two neighbouring trigger towers is
required to be above a pre-defined threshold of 8 GeV.
The events are passed on to the second trigger level, the high level trigger (HLT), if
they satisfy the selection at L1. In the HLT part, the same clustering algorithm used in
the offline photon reconstruction would be used to cluster the crystals [66]. According
to the known phenomenon observed in collision events, anomalous interactions arisen
from direct ionization of the avalanche photodiode (APD) would lead to an apparently
high energy deposit localized in an single crystal. In order to suppress the influence
comes from anomalous energy deposits in the ECAL APD, a simple topological
selection criterion is implemented at HLT level: the highest energy deposit in a
single crystal, E 1 , is compared to the summation of energy deposited in the four
adjacent crystals in the η and φ, E 4 for each channel. Such anomalous signals with
1− E 4 /E 1 > 0.95 are rejected [67] and the corresponding HLT paths are then labeled
as “Cleaned”.
The events that have at least one reconstructed cluster in the ECAL with a E T
above a programmable threshold (E THLT ) are stored for the further analysis. In this
analysis, E THLT of 20, 30, 50 and 70 are used, depending on the run period [68]. Over
the course of the 2010 proton-proton collision with respect to different run period, the
analysis data sets are required with various trigger conditions. The available triggers
and their prescaling factors changed over time due to the increased instantaneous
luminosity. To simplify the analysis, only events with un-prescaled trigger paths
are selected for the different run ranges. The effective integrated luminosities for
different HLT paths and corresponding run ranges are summarized in Table 4.1. The
behavior of the employed HLT as a function of the leading photon pT is studied in

Table 4.1 Summary of the single photon triggers, corresponding run ranges and effective integrated
luminosities

HLT path name Run range L dt, pb−1
HLT_Photon20_Cleaned_L1R 138564–143962 2.46
HLT_Photon30_Cleaned_L1R 144010–147116 5.81
HLT_Photon50_Cleaned_L1R_v1 147196–148058 9.47
HLT_Photon70_Cleaned_L1R_v1 148822–149294 18.4
4.1 Trigger Requirement 35

order to correct for possible inefficiencies in the different regions of the phase space.
According to the selection of HLT in this analysis, The HLT efficiency of data has
+0.5%
been measured, 99.1%−1.2% , in the considered kinematic region.

4.2 Pileup Reweighting

The phenomenon when there is more than one interaction appeared inside the same
bunch crossing would be defined as pileup condition. Generally, the pileup effect
sometime would seriously affect or bias the results of analysis. This effect must be
well-understood. In order to study the pileup contribution for each data sample, all
the MC samples have been generated with distributions for the number of pileup
interaction.
The reweighting procedure is based on an iterative process: the distribution of
reconstructed vertex extracted from full data and MCs in real cross section scale. The
MCs are reweighted according to the ratio as a function of pileup interaction s for
each event. Then the weighted distribution of vertex reconstructed from MCs is used
to study the difference with respect to the distribution of vertex reconstructed from
data by calculating their ratio. MCs are reweighted repeatedly until the difference is
negligible. After at least 4 times of iteration, there is a good agreement between the
data and weighted MCs. The related results of pileup reweighting for different MC
samples are shown in Fig. 4.1.

106 3
7TeV DATA(2010)
Pythia8 MC(GJet+QCD), PU-reweight

105 Pythia8 MC(GJet+QCD), no PU-reweight

Pythia6 MC(GJet+QCD), PU-reweight 2.5 Pythia8 MC(GJet+QCD)

Pythia6 MC(GJet+QCD), no PU-reweight

number of events

Pythia6 MC(GJet+QCD)
Madgraph MC(GJet+QCD), PU-reweight
104 Madgraph MC(GJet+QCD), no PU-reweight
2 Madgraph MC(GJet+QCD)
DATA / MC

3
10
1.5
102
1
10
0.5
1
0
0 5 10 15 20 25 30 0 2 4 6 8 10 12 14
number of vertex number of vertex

Fig. 4.1 Reconstructed vertex distribution of uncorrected data and MC samples. a Comparison
between uncorrected data, MCs without pileup reweighting and MCs with pileup reweighting. b
Ratio plot between uncorrected data and reweighted MCs
36 4 Event Selection

Fig. 4.2 The square pT sum 104

CMS Preliminary 2010, s = 7 TeV, L = 36.14 pb-1, pp → γ +3jets

of tracks associated to 2nd γ -jet: p > 75 GeV

T DATA
2jets: p > 20 GeV
primary vertex. Only events T
γ : |η|<1.4442, 1.566<|η|< 2.5 Pythia8 4C

with at least two primary jets: |η|<2.4

number of events
103 Pythia6 Z2

vertices are shown Madgraph+Pythia6 Z2

102

10

1
0 5 10 15 20 25 30 35 40

Σ P2T,i-th track(2nd vertex)

Ideally, most of analyses focus on studying the physic phenomena inside a single
interaction. The LHC circulates protons inside its beam-pipes not in a continuous
stream but in several very closely packed bunches. As a matter of known fact, the
pileup effect is already unavoidable in the later run of 2010 and even becomes much
more serious. To some extent, analyzing particles from different proton-proton colli-
sion cause large uncertainties in final results when the pileup part is high pT interac-
tion. But in order not to over-kill the problem of background rejection from pileup,
we try to reject part of beam-crossings where there are more than one proton-proton
interactions as they may give rise to background from pileup. An special selection
on the number of reconstructed
 2 vertex is introduced to reduce influences from pileup
effect. Events with pT i−th track (2nd vertex) < 10 GeV2 are accepted, where pT
nd
i−th track is the transverse momentum of the i-th track associated to the 2 primary
vertex. Figure 4.2 shows the related distribution for events with at least two primary
vertices. The value of selection criterion is a relatively safe setting by taking into
the pT threshold of jets account. This selection criterion rejects additional jets from
pileup interactions because the contribution of tracks associated to the 2nd primary
vertex is low in the jet reconstruction. The uncertainty associated to this selection cri-
terion has been taken into account as a source of systematic uncertainty, as discussed
later.
Figure 4.3 shows the comparison of number of reconstructed vertices between
uncorrected data and different combined MC samples. A good agreement between
data and MCs can be found in the normalization scale. But different MCs have
significant differences in the absolute scale. Such disagreement is quite uniform in
all the bins of vertices distribution. The prediction of MadGraph interfaced with
pythia 6 tune Z2 describes data better than pythia 6 tune Z2 and pythia 8 tune
4 C within the statistics uncertainty. pythia 6 tune Z2 and pythia 8 tune 4 C are
underestimated and with relatively large statistic uncertainty.
4.3 Photon Identification 37

CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j
105 10
γ -j: pT γ , pT jet1 > 75GeV DATA γ -j: pT γ , pT jet1 > 75GeV DATA
2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV
104 PYTHIA8 4C PYTHIA8 4C
γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
jets: | η|<2.4 PYTHIA6 Z2 1 jets: | η|<2.4 PYTHIA6 Z2
number of events

3
10 MADGRAPH+PY6 Z2 MADGRAPH+PY6 Z2

dN/N
102 10-1

10
10-2

10-3
2.5 3.5
MC / DATA

MC / DATA
3
2 number of vertex 2.5 number of vertex
1.5 2
1 1.5
1
0.5
0.5
0 0
0 2 4 6 8 10 0 2 4 6 8 10
number of vertex number of vertex

Fig. 4.3 The detector level comparison of number of reconstructed vertices between uncorrected
data and different combined MC samples in a real cross section scale and b normalization scale

4.3 Photon Identification

Photon candidates are reconstructed from energy deposits in the ECAL, and are
referred to as super-clusters (SC). The calorimeters and tracker system of CMS are
contained within a 3.8 T magnetic field. The radiation length of the material between
the interaction point and the ECAL varies as a function of pseudorapidity (η), and
is significant enough to cause about half of photons from the hard interaction to
convert before reaching the ECAL. To allow recovery of energy emitted by the
bremsstrahlung from e+ e− pairs via photon conversions in the barrel regions of
CMS, super-clusters are up to 35 crystals wide in φ and 5 crystals wide in η. In the
endcap region of CMS, super-clusters are constructed out of at least one array of
5×5 crystals. Further details can be found in [66].
Reconstructed photon candidates are required to have pT > 75 GeV to assure a
fully efficient trigger, and |η| < 1.4442 and 1.566 < |η| < 2.5 to avoid instrumental
inefficiencies of the calorimeter. Photons are reconstructed and selected as described
in Reference [69, 70].

• ISOTRK : The track isolation is the sum of the pT of tracks located in a cone
ΔR = 0.4 centered around the line connecting the center of the super-cluster to
the primary vertex. To avoid misidentification of converted photons, tracks within
ΔR = 0.04 and Δη = 0.015 are not included in the sum.
• ISOECAL : The ECAL isolation is the sum of the E T of energy deposits in the indi-
vidual crystals located in a cone of ΔR = 0.4, centered around the super-cluster.
In order to remove the contribution due to the photon candidate, an exclusion
region consisting of an inner cone of radius ΔR = 0.06 and a strip of dimension
Δη × Δφ = 0.04 × 0.4 is defined. The sum is performed only for crystals where
|E| > 80 MeV.
38 4 Event Selection

• ISOHCAL : The Hadron Calorimeter (HCAL) isolation is defined as the sum of the
energy deposited in the HCAL towers in a cone ΔR = 0.4 around the photon
direction, excluding a cone of radius ΔR = 0.15.
• H/E: A photon reaching the ECAL without undergoing conversion to an e+ e− pair
deposits most of its energy in a 3 × 3 crystal matrix. Only a very small fraction of
the energy from the resulting shower leaks into the HCAL. The ratio of Hadronic
to Electromagnetic energy is defined by taking the ratio of energy deposited in the
HCAL and ECAL in a cone of radius ΔR = 0.15 around the photon direction.
• σ ηη : A particular shower shape variable which measures the width of the photon
super-cluster in the η direction is defined as:

Σ(ηi − η̄)ωi Σηi ωi

σηη
2
= , η̄ = (4.1)
Σωi Σωi

where the sum runs over all the 5 × 5 crystal matrix and ωi = max(0, 4.7 +
log(E i /E 5×5 )), where E i is the energy of i-th crystal and E 5×5 is the total energy
of the crystal matrix; ηi = η̂i × δη, where η̂i is the η index of i-th crystal and
δη = 0.0174.

To have high purity of selected photons, and to be above the turn on curve for all
trigger paths used for 2010 data, listed in Sect. 4.1. Photon candidates are required
to have pT > 75 GeV. Furthermore, to remove contamination from beam halo, an
additional shower shape variable, σφφ (which have similar definition of σηη but in
the φ plane), is required to have σφφ > 0.009 and measured hit time of the most
energetic crystal, tseed , is required to have |tseed | < 1.5. These two selection criteria
are 100% for MC signal photons, but remove 99% of photon arising from beam halo.
The position of super-cluster was extrapolated to the interaction point and required

Table 4.2 Summary of photon identification criteria

Selection criteria for photon
pT > 75 GeV
|ηSC | < 1.4442, the barrel region
1.566 < |ηSC | < 2.5, the endcap region
γ
ISOTRK < 2.0 GeV + 0.001 ∗ E T
γ
ISOECAL < 4.2 GeV + 0.003 ∗ E T
γ
ISOHCAL < 2.2 GeV + 0.001 ∗ E T
H/E < 0.05
σηη < 0.01 in the barrel region
σηη < 0.03 in the endcap region
σφφ > 0.009
|tseed | < 1.5
Pixel seed veto
4.3 Photon Identification 39

CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j
γ -j: pT γ , pT jet1 > 75GeV DATA γ -j: pT γ , pT jet1 > 75GeV DATA
104 2j: pT jet2, pT jet3 > 20GeV 10 2j: pT jet2, pT jet3 > 20GeV
PYTHIA8 4C PYTHIA8 4C
γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
jets: | η|<2.4 PYTHIA6 Z2 jets: | η|<2.4 PYTHIA6 Z2
Events / 25 GeV
3
10 1
MADGRAPH+PY6 Z2 MADGRAPH+PY6 Z2

dN/N
2 -1
10 10

10 10-2

10-3
1

3.5 4
MC / DATA

MC / DATA
3 3.5
2.5 3
2.5
2
2
1.5
1.5
1 1
0.5 0.5
0 0
75 100 125 150 175 200 225 250 275 300 325 75 100 125 150 175 200 225 250 275 300 325
p (γ ) (GeV) p (γ ) (GeV)
T T
CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j
1

dN/N
Events / 0.4 (rad)

γ -j: pT γ , pT jet1 > 75GeV DATA γ -j: pT γ , pT jet1 > 75GeV DATA
2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV
PYTHIA8 4C PYTHIA8 4C
γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
103 jets: | η|<2.4 PYTHIA6 Z2 jets: | η|<2.4 PYTHIA6 Z2
MADGRAPH+PY6 Z2 MADGRAPH+PY6 Z2

10-1
102

2.5 3
MC / DATA

MC / DATA

2 2.5
2
1.5
1.5
1
1
0.5 0.5
0 0
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
η (γ ) η (γ )
CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j
1
dN/N
Events / 0.315 (rad)

γ -j: pT γ , pT jet1 > 75GeV DATA γ -j: pT γ , pT jet1 > 75GeV DATA
2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV
PYTHIA8 4C PYTHIA8 4C
103
γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
jets: | η|<2.4 PYTHIA6 Z2 jets: | η|<2.4 PYTHIA6 Z2
MADGRAPH+PY6 Z2 MADGRAPH+PY6 Z2

10-1
102

3 3
MC / DATA

MC / DATA

2.5 2.5
φ (γ )
2 2
1.5 1.5
1 1
0.5 0.5
0 0
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
φ (γ ) φ (γ )

Fig. 4.4 The detector level comparison of leading photon’s transverse momentum pT (top row),
pseudorapidity η (center row) and azimuthal angle φ (bottom row) between uncorrected data and
different combined MC samples in real cross section scale (left) and normalization scale (right)

to have no match with hits in the pixel detector compatible with tracks from the hard
interaction. Such kind of requirement could reduce the possible source of background
from electron misidentification.
The summary of photon identification criteria used in this analysis is listed in
Table 4.2.
The photon energy scale, 0.43%, is determined with electrons from reconstructed
Z -boson decays with an uncertainty estimated to be 0.6% (1.5%) in the barrel (end-
40 4 Event Selection

cap) region [71–73]. Electron samples from Z -boson decays are used to estimate
the efficiency of the isolation criteria, εisolation . The εisolation in the data is found to
be higher (lower) than in simulation in the barrel (endcap) region, therefore a scale
factor of 1.008 ± 0.005 (0.984 ± 0.005) is applied to the simulation. The photon
pixel veto efficiency, εpixel veto , is estimated by analyzing final-state-radiation photons
in Z → μ+ μ− γ events. The εpixel veto in the data is found to be higher (lower) than
in simulation in the barrel (endcap) region, therefore a scale factor of 1.009 ± 0.029
(0.959 ± 0.062) is included in the simulation [70].
Comparison of basic properties of the selected leading photon with highest pT
are shown from Fig. 4.4.

4.4 Jet Identification

Jets used in this analysis are reconstructed by using the particle flow method [74].
Associated energy correction of jets are applied to correct the pT of jets and to make
jet response flat in the η plane [75]. Jets are reconstructed with the anti-k T clustering
algorithm with a cone size of ΔR = 0.5 [27, 28]. All the jets are requested to pass
loose jet identification criteria [76]. The leading jet is requested to have pT > 75 GeV.
The other jets are requested to have pT > 20 GeV. This kind of setting is helpful
to discriminate the different sub-processes, photon-jet process and di-jet process.
Besides, all the jets are counted only if they are within the acceptance of the CMS
tracker(|η| < 2.4). Jets are rejected if they overlap with the selected leading photon
within a cone of ΔR = 0.5. The leading photon and leading jet are tagged as the
“photon-jet pair”, while the other two jets are the “di-jet pair”.
In order to reduce the probability to select jets from pile-up, jet-vertex association
algorithm is introduced. The basic idea of jet-vertex association algorithm is to select
jets according to the correlation between the target jet and different vertices. By
checking the component tracks, we can know their dependence on each vertex. For

Table 4.3 Summary of jet identification criteria

Selection criteria for jets
pT > 75 GeV for leading jet
pT > 20 GeV for the other jets
|η| < 2.4
ΔR(γ , jet)>0.5
Charged Hadron Fraction > 0.0
Neutral Hadron Fraction < 1.0
Charged Multiplicity > 0.0
Charged Electromagnetic Fraction < 1.0
Neutral Electromagnetic Fraction < 1.0
4.4 Jet Identification 41

CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j
104 10

dN/N
Events / 25 GeV
γ -j: pT γ , pT jet1 > 75GeV DATA γ -j: pT γ , pT jet1 > 75GeV DATA
2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV
PYTHIA8 4C PYTHIA8 4C
γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
jets: | η|<2.4 PYTHIA6 Z2 1 jets: | η|<2.4 PYTHIA6 Z2
103
MADGRAPH+PY6 Z2 MADGRAPH+PY6 Z2

10-1

102

10-2

10

4 4
MC / DATA

MC / DATA
3.5 3.5
3 3
2.5 2.5
2 2
1.5 1.5
1 1
0.5 0.5
0 0
75 100 125 150 175 200 225 250 275 300 325 75 100 125 150 175 200 225 250 275 300 325
p (jet1) (GeV) p (jet1) (GeV)
T T

CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j
1

dN/N
Events / 0.4 (rad)

γ -j: pT γ , pT jet1 > 75GeV DATA γ -j: pT γ , pT jet1 > 75GeV DATA
2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV
PYTHIA8 4C PYTHIA8 4C
103 γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
jets: | η|<2.4 PYTHIA6 Z2 jets: | η|<2.4 PYTHIA6 Z2
MADGRAPH+PY6 Z2 MADGRAPH+PY6 Z2
-1
10
102

10
10-2

3 3
MC / DATA

MC / DATA

2.5 2.5
2 2
1.5 1.5
1 1
0.5 0.5
0 0
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
η (jet1) η (jet1)

CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j
103 1
dN/N
Events / 0.315 (rad)

γ -j: pT γ , pT jet1 > 75GeV DATA γ -j: pT γ , pT jet1 > 75GeV DATA
2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV
PYTHIA8 4C PYTHIA8 4C
γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
jets: | η|<2.4 PYTHIA6 Z2 jets: | η|<2.4 PYTHIA6 Z2
MADGRAPH+PY6 Z2 MADGRAPH+PY6 Z2

102
10-1

3 3
MC / DATA

MC / DATA

2.5 2.5
2 2
1.5 1.5
1 1
0.5 0.5
0 0
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
φ (jet1) φ (jet1)

Fig. 4.5 The detector level comparison of leading jet’s transverse momentum pT (top row), pseudo-
rapidity η (center row) and azimuthal angle φ (bottom row) between uncorrected data and different
combined MC samples in real cross section scale (left) and normalization scale (right)
42 4 Event Selection

CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j
104 10

dN/N
Events / 20 GeV
γ -j: pT γ , pT jet1 > 75GeV DATA γ -j: pT γ , pT jet1 > 75GeV DATA
2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV
PYTHIA8 4C PYTHIA8 4C
γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
103 jets: | η|<2.4 PYTHIA6 Z2 1 jets: | η|<2.4 PYTHIA6 Z2
MADGRAPH+PY6 Z2 MADGRAPH+PY6 Z2

102 10-1

10 10-2

3 50 100 150 200 250 4 50 100 150 200 250

MC / DATA

MC / DATA
3.5
2.5 p (jet2) (GeV) 3
p (jet2) (GeV)
2 T T
2.5
1.5 2
1 1.5
1
0.5 0.5
0 0
20 40 60 80 100 120 140 160 180 200 220 240 260 20 40 60 80 100 120 140 160 180 200 220 240 260
p (jet2) (GeV) p (jet2) (GeV)
T T

CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j
1

dN/N
Events / 0.4 (rad)

γ -j: pT γ , pT jet1 > 75GeV DATA γ -j: pT γ , pT jet1 > 75GeV DATA
2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV
PYTHIA8 4C PYTHIA8 4C
γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
3
10
jets: | η|<2.4 PYTHIA6 Z2 jets: | η|<2.4 PYTHIA6 Z2
MADGRAPH+PY6 Z2 MADGRAPH+PY6 Z2

10-1
102

10

10-2
3-3 -2 -1 0 1 2 3 3-3 -2 -1 0 1 2 3
MC / DATA

MC / DATA

2.5
η (jet2) 2.5
η (jet2)
2 2
1.5 1.5
1 1
0.5 0.5
0 0
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
η (jet2) η (jet2)
CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j
1
dN/N
Events / 0.315 (rad)

γ -j: pT γ , pT jet1 > 75GeV DATA γ -j: pT γ , pT jet1 > 75GeV DATA
3
2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV
10 PYTHIA8 4C PYTHIA8 4C
γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
jets: | η|<2.4 PYTHIA6 Z2 jets: | η|<2.4 PYTHIA6 Z2
MADGRAPH+PY6 Z2 MADGRAPH+PY6 Z2

10-1

102

3 -3 -2 -1 0 1 2 3 3 -3 -2 -1 0 1 2 3
MC / DATA

MC / DATA

2.5
φ (jet2) 2.5
φ (jet2)
2 2
1.5 1.5
1 1
0.5 0.5
0 0
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
φ (jet2) φ (jet2)

Fig. 4.6 The detector level comparison of second leading jet’s transverse momentum pT (top row),
pseudorapidity η (center row) and azimuthal angle φ (bottom row) between uncorrected data and
different combined MC samples in real cross section scale (left) and normalization scale (right)
4.4 Jet Identification 43

CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j
103

dN/N
Events / 33.3 GeV
105 γ -j: pT γ , pT jet1 > 75GeV DATA γ -j: pT γ , pT jet1 > 75GeV DATA
2j: pT jet2, pT jet3 > 20GeV 102 2j: pT jet2, pT jet3 > 20GeV
PYTHIA8 4C PYTHIA8 4C
104 γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
jets: | η|<2.4 PYTHIA6 Z2 10 jets: | η|<2.4 PYTHIA6 Z2
3
10
MADGRAPH+PY6 Z2 1 MADGRAPH+PY6 Z2

102
10-1

10 10-2

1 10-3

10-4
10-1

40
10-54
20 40 60 80 100 120 140 160 180 200 220 0 20 40 60 80 100 120 140 160 180 200 220
MC / DATA

MC / DATA
3.5 3.5
3
p (jet3) (GeV) 3
p (jet3) (GeV)
T T
2.5 2.5
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 50 100 150 200 0 50 100 150 200
p (jet3) (GeV) p (jet3) (GeV)
T T

CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j
103 1

dN/N
Events / 0.4 (rad)

γ -j: pT γ , pT jet1 > 75GeV DATA γ -j: pT γ , pT jet1 > 75GeV DATA
2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV
PYTHIA8 4C PYTHIA8 4C
γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
jets: | η|<2.4 PYTHIA6 Z2 jets: | η|<2.4 PYTHIA6 Z2
MADGRAPH+PY6 Z2 MADGRAPH+PY6 Z2

102 10-1

10-2
3-3 -2 -1 0 1 2 3 3-3 -2 -1 0 1 2 3
MC / DATA

MC / DATA

2.5
η (jet3) 2.5
η (jet3)
2 2
1.5 1.5
1 1
0.5 0.5
0 0
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
η (jet3) η (jet3)
CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j
103 1
dN/N
Events / 0.315 (rad)

γ -j: pT γ , pT jet1 > 75GeV DATA γ -j: pT γ , pT jet1 > 75GeV DATA
2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV
PYTHIA8 4C PYTHIA8 4C
γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
jets: | η|<2.4 PYTHIA6 Z2 jets: | η|<2.4 PYTHIA6 Z2
MADGRAPH+PY6 Z2 MADGRAPH+PY6 Z2

102
10-1

3 -3 -2 -1 0 1 2 3 3 -3 -2 -1 0 1 2 3
MC / DATA

MC / DATA

2.5
φ (jet3) 2.5
φ (jet3)
2 2
1.5 1.5
1 1
0.5 0.5
0 0
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
φ (jet3) φ (jet3)

Fig. 4.7 The detector level comparison of third leading jet’s transverse momentum pT (top row),
pseudorapidity η (center row) and azimuthal angle φ (bottom row) between uncorrected data and
different combined MC samples in real cross section scale (left) and normalization scale (right)
44 4 Event Selection

CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j
105 102
γ -j: pT γ , pT jet1 > 75GeV DATA γ -j: pT γ , pT jet1 > 75GeV DATA
2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV
PYTHIA8 4C 10 PYTHIA8 4C
104 γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
jets: | η|<2.4 PYTHIA6 Z2 jets: | η|<2.4 PYTHIA6 Z2
MADGRAPH+PY6 Z2 1 MADGRAPH+PY6 Z2
103
Events

dN/N
10-1
102

10-2
10

10-3
1
2 2.5
MC / DATA

MC / DATA
1.8
1.6 2
1.4
1.2 1.5
1
0.8 1
0.6
0.4 0.5
0.2
0 0
2 4 6 8 10 12 2 4 6 8 10 12
number of jets (PT > 20 GeV) number of jets (PT > 20 GeV)

Fig. 4.8 The detector level comparison of inclusive number of jets with pT > 20 GeV between
uncorrected data and different combined MC samples in real cross section scale (left) and normal-
ization scale (right)

example, there are two reconstructed vertices (ver tex A and ver tex B) and two
reconstructed jets ( jet 1 and jet 2) in one specific event. If most of component
tracks of jet 1 overlap to those tracks which are used to reconstructed ver tex A,
then we will say jet 1 is highly associated with ver tex A. That is, jet 1 is from
interaction A which is respect to ver tex A. The summary of jet identification
criteria used in this analysis is listed in Table 4.3.
Comparison of basic properties of the selected leading jets and also the exclusive
jet multiplicity ( pT > 20 GeV) are shown from Figs. 4.5, 4.6, 4.7 and 4.8.
Chapter 5
Discriminating Observables

A final state arising from background SPS events tends to have a strongly correlated
configuration in the azimuthal angle and pT -balance distribution between the two
sub-systems, while a DPS signal event has a preferred back-to-back topology and
good pT -balance for the separated systems that are not correlated. To differentiate
between SPS and DPS processes, a set of three discriminating observables are inves-
tigated:

• the difference in azimuthal angle between the jets belonging to the di-jet pair,

Δφ23 = φjet2 − φjet3 ; (5.1)

DPS and SPS will have different configurations for the photon and 3 jets in the
final state in the phase space of angular variable. Δφ23 computes the difference of
azimuthal angle in the di-jet sub-system. SPS background events lead to a broad
distribution for these quantities while DPS signal events contribute most at values
close to π . The later would perform a back-to-back topology of the di-jet pair.
• the transverse momentum balance of the jets belonging to the di-jet pair,
|pT jet2 + pT jet3 |
Δrel pT,23 = ; (5.2)
|pT jet2 | + |pT jet3 |

Δrel pT,23 is sensitive to the momentum balance associated with the di-jet sub-
system. A back-to-back configuration of the di-jet sub-system contributes at low
value Δrel pT,23 , while correlated pairs of jets result in a broader distribution over
the whole phase space.

© Springer Nature Singapore Pte Ltd. 2017 45

Y.-H. Chang, Study of Double Parton Scattering in Photon + 3 Jets Final State,
Springer Theses, DOI 10.1007/978-981-10-3824-2_5
46 5 Discriminating Observables

Fig. 5.1 A possible orientation of the transverse momenta of the photon and the three jets in photon
+ 3 jets events (left). Illustration of the definition of the ΔS observable, applied to a photon + 3 jets
SPS event (center) and a photon + 3 jets DPS event (right)

• the azimuthal angle between the pT vectors of the photon-jet pair and the di-jet
pair,

ΔS ≡ Δφ(pT (γ , jet1), pT (jet2, jet3)), (5.3)

where γ , jet1, jet2, and jet3 stand for the leading photon, the leading jet, the second
leading jet, and the third leading jet, respectively. For the di-jet pair of DPS event,
the two jets balance each other and are also independent from the first interaction.
Hence Δφ23 will be close to π and Δrel pT,23 will be small. Figure 5.1 (left) illustrates
the definition of ΔS with a possible orientation of the transverse momentum vectors
of the selected photon and jets. Figure 5.1 (center) shows a possible topology of
a background SPS event with two radiation jets emitted close to the leading jet’s
direction, which is recoiling against the photon’s direction in φ are preferred. The
correlation due to momentum conservation biases ΔS towards 180◦ (π ). For a typical
DPS signal event shown in Fig. 5.1 (right), we have better pairwise balance among the
photon-jet pair and the di-jet pair. The low correlation between the two sub-systems
leads to a flatter ΔS distribution.
Except of the above observables, there are more defined in Appendix C which can
be used to analyze the contributions of the two sub-systems. But we only include the
relevant results of the most powerful observables in the main body of this thesis.
The stable-particle level comparisons between the DPS signal and all possible
background components for different MC predictions are shown in Fig. 5.2 using
area-normalized distributions of these three observables within the phase space as
defined in Table 5.1. The hadronic energy surrounding the photon is required to be
at most 5 GeV within ΔR < 0.4 [69].
All DPS signal distributions are produced by pythia 8 tune 4C, and are taken
as a reference for the comparisons. In such DPS signal production, one hard parton
interaction generates a prompt photon and a jet and the second one generates 2
jets in the same proton–proton collision. pythia 8 tune 4C, MadGraph interfaced
with pythia 8 tune 4C, and sherpa tune CT10 are separately used to produce
different background distributions: direct photon + 3 jets SPS events, fragmentation
5 Discriminating Observables 47

CMS Simulation Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Simulation Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Simulation Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j

1 γ -j: pT γ , pT jet1 > 75GeV DPS signal γ -j: p , pT jet1 > 75GeV
Tγ
DPS signal γ -j: p , pT jet1 > 75GeV
Tγ
DPS signal
σ dΔφ [1/rad]
2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV
direct photon+3jets, SPS direct photon+3jets, SPS direct photon+3jets, SPS
γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5

σ dΔS [1/rad]
jets: |η|<2.4 frag. photon+3jets, SPS 1 jets: |η|<2.4 frag. photon+3jets, SPS 10 jets: |η|<2.4 frag. photon+3jets, SPS
23
1 dσ

Signal: PYTHIA8 4C Signal: PYTHIA8 4C Signal: PYTHIA8 4C

fake photon+3jets fake photon+3jets 1 Background: PYTHIA8 4C fake photon+3jets
Background: PYTHIA8 4C Background: PYTHIA8 4C

T,23

1 dσ
σ dΔrelp
1 dσ
10-1 10-1

10-1

10-2

10-2
0 0.5 1 1.5 2 2.5 3 0 0.2 0.4 0.6 0.8 1 0 0.5 1 1.5 2 2.5 3
Δφ (rad) ΔrelpT,23 ΔS (rad)
23
CMS Simulation Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Simulation Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Simulation Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j

1 γ -j: pT γ , pT jet1 > 75GeV DPS signal γ -j: p , pT jet1 > 75GeV
Tγ
DPS signal γ -j: p , pT jet1 > 75GeV
Tγ
DPS signal
σ dΔφ [1/rad]

2j: pT jet2, pT jet3 > 20GeV direct photon+3jets, SPS 2j: pT jet2, pT jet3 > 20GeV direct photon+3jets, SPS 2j: pT jet2, pT jet3 > 20GeV direct photon+3jets, SPS
γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
frag. photon+3jets, SPS frag. photon+3jets, SPS frag. photon+3jets, SPS

σ dΔS [1/rad]
jets: |η|<2.4 1 jets: |η|<2.4 10 jets: |η|<2.4
23

fake photon+3jets fake photon+3jets fake photon+3jets

1 dσ

Signal: PYTHIA8 4C Signal: PYTHIA8 4C Signal: PYTHIA8 4C

Background: MADGRAPH+PY8 4C Background: MADGRAPH+PY8 4C
1 Background: MADGRAPH+PY8 4C
T,23

1 dσ
σ dΔrelp
1 dσ

10-1 10-1

10-1

10-2

10-2
0 0.5 1 1.5 2 2.5 3 0 0.2 0.4 0.6 0.8 1 0 0.5 1 1.5 2 2.5 3

Δφ (rad) ΔrelpT,23 ΔS (rad)

23
CMS Simulation Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Simulation Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Simulation Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j

1 γ -j: pT γ , pT jet1 > 75GeV DPS signal γ -j: p , pT jet1 > 75GeV
Tγ
DPS signal 10 γ -j: pT γ , pT jet1 > 75GeV DPS signal
σ dΔφ [1/rad]

2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV
direct photon+3jets, SPS direct photon+3jets, SPS direct photon+3jets, SPS
γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
σ dΔS [1/rad]
jets: |η|<2.4 frag. photon+3jets, SPS 1 jets: |η|<2.4 frag. photon+3jets, SPS jets: |η|<2.4 frag. photon+3jets, SPS
23
1 dσ

Signal: PYTHIA8 4C Signal: PYTHIA8 4C Signal: PYTHIA8 4C

fake photon+3jets fake photon+3jets 1 Background: SHERPA CT10 fake photon+3jets
Background: SHERPA CT10 Background: SHERPA CT10
T,23

1 dσ
σ dΔrelp
1 dσ

10-1 10-1

10-1

10-2

10-2
0 0.5 1 1.5 2 2.5 3 0 0.2 0.4 0.6 0.8 1 0 0.5 1 1.5 2 2.5 3

Δφ (rad) ΔrelpT,23 ΔS (rad)

23

Fig. 5.2 Comparison of MC-based signal and background distributions for different DPS-
discriminating observables: Δφ23 (left), Δrel pT,23 (center), and ΔS (right). The signal distributions
are all produced by pythia 8 tune 4C. The comparison of background distributions are shown for
three MC event generators: MadGraph interfaced with pythia 8 tune 4C (top row), pythia 8 tune
4C (center row), and sherpa tune CT10 (bottom row). Three kinds of background contributions
are included in these comparisons: direct photon + 3 jets SPS events, fragmentation photon (frag.
photon) + 3 jets SPS events and misidentified photon (fake photon) + 3 jets events

Table 5.1 Phase space Leading photon pT > 75 GeV, |η| < 1.4442,
definition for photon + 3 jets 1.566 < |η| < 2.5
production at stable-particle
level Leading jet pT > 75 GeV, |η| < 2.4
Second leading jet pT > 20 GeV, |η| < 2.4
Third leading jet pT > 20 GeV, |η| < 2.4
48 5 Discriminating Observables

photon + 3 jets SPS events and misidentified photon + 3 jets events. The background
distributions of direct photon + 3 jets SPS events are obtained by requiring the final
state photon to come from qg Compton scattering or qq annihilation and by removing
the MPI contribution. The background distributions of fragmentation photon + 3 jets
SPS events are obtained with the same method except for requiring the final state
photon to be from the fragmentation of a parton. The background distributions of
misidentified photon + 3 jets events are extracted by requiring the final state photon
to stem from a π 0 or η meson.
The area-normalized distribution of Δφ23 is shown in Fig. 5.2 (left column). The
jets are strongly (weakly) correlated at high (low) Δφ23 values. A local maximum
is visible around Δφ23 0.5–1.0 because the anti-kT clustering jet algorithm merges
jets with a separation less than the distance parameter of R = 0.5. Figure 5.2 (center
column) shows the transverse momentum balance of the jets belonging to the di-jet
pair, Δrel pT,23 . Better transverse momentum balance of jets tends to smaller values
of Δrel pT,23 . The peak around Δrel pT,23 ∼ π indicates the imbalance in pT of di-jet
pair, which is because the jets are inexactly matched or originate from radiation of
the initial- or final-state of the leading jet. The azimuthal angle between the planes
of the photon-jet pair and the di-jet pair, ΔS, is shown in Fig. 5.2 (right column). At
low ΔS values, the photon-jet pair and the di-jet pair are not correlated.
The DPS signal distribution shows different features with respect to each back-
ground distribution and is distinguishable in the DPS-sensitive region: at high Δφ23
values, low Δrel pT,23 values, and low to medium ΔS values. A sizeable model depen-
dence can be observed from comparisons of background distributions predicted by the
investigated MC event generators indicating e.g. differences in the merging between
matrix elements and parton showers.
For pythia 8 tune 4C (Fig. 5.2 top row), the three kinds of background distribu-
tions present similar behavior to each other in Δφ23 and Δrel pT,23 but give distinct
contributions in the DPS-sensitive region of ΔS distribution. MadGraph, interfaced
with pythia 8 tune 4C (Fig. 5.2 center row), predicts the distributions of fragmenta-
tion photon + 3 jets SPS events with less back-to-back topology of Δφ23 and worse
pT -balance of Δrel pT,23 for the di-jet pair than pythia 8 tune 4 C and sherpa tune
CT10. In the region of ΔS < 2.5, fragmentation photon + 3 jets SPS events give less
contributions with respect to the other background components. sherpa tune CT10
(Fig. 5.2 bottom row) gives quite different predictions of misidentified photon + 3
jets background events with respect to the other two MC generators. Its misidentified
photon + 3 jets background events behave similar to DPS signal events in ΔS distri-
bution and also show a more back-to-back topology of Δφ23 and better pT -balance
of Δrel pT,23 for the di-jet pair than the other two kinds of background distributions.
Chapter 6
Photon + 3 Jets Events at Detector Level

In this chapter, the discriminating observables defined in Chap. 5 are used to measure
the photon + 3 jets event in data and compared with the predictions of various MC
simulations, pythia 6 tune Z2, pythia 8 tune 4C and MadGraph interfaced with
pythia 6 tune Z2, at the detector level. Furthermore, the photon purity, defined
as the fraction of isolated prompt photons in data, is estimated from an extended
maximum likelihood two-component fit with related photon signal and background
fitting templates.

6.1 DPS Performance at Detector Level

After passing complete selections, predictions of different MC simulations are re-

weighted with respect to the pile-up condition of 2010 data. Contributions of different
processes are weighted to the effective integrated luminosity of 2010 full data, 36
pb-1 , and combined together for each MC simulation. Related corrections, such as
energy scale and energy resolution of photon and jets and photon identification
efficiency, are applied to make MC simulation being more comparable with complete
2010 data sets. The corresponding systematic uncertainties are studied and shown in
the later chapter. A detailed list with the total number of events and with the number
of events progressively after each applied selection can be found in Tables 6.1 and
6.2 for each data and various MC simulations.
The 2nd primary vertex selection rejects about 65% of events which have more than
two vertices. Such selection rate is uniform for data and different MC simulations.

© Springer Nature Singapore Pte Ltd. 2017 49

Y.-H. Chang, Study of Double Parton Scattering in Photon + 3 Jets Final State,
Springer Theses, DOI 10.1007/978-981-10-3824-2_6
50 6 Photon + 3 Jets Events at Detector Level

Table 6.1 Summary of number of events passing the applied selection criteria for different 2010
data sets
Data sample Good vertex Trigger Photon Jets selection 2nd vertex
requirement requirement selection selection
EG 4.59 × 107 1.35 × 107 2956 279 209
Photon 2.04 × 107 6.07 × 106 31408 3122 1581

Table 6.2 Summary of number of events passing the applied selection criteria for different MC
simulations. Contributions of different processes are weighted to the effective integrated luminosity
of 2010 data, 36 pb-1 , and combined together
Data sample Good vertex Trigger Photon Jets selection 2nd vertex
requirement requirement selection selection
pythia-8 7.21 × 105 1.16 × 105 19987 1053 648
photon-jet
pythia-8 2.20 × 109 7.26 × 105 9514 1799 657
QCD
pythia-6 6.68 × 105 1.08 × 105 17378 999 508
photon-jet
pythia-6 1.95 × 109 6.60 × 105 9357 1550 771
QCD
MadGraph 1.06 × 106 1.33 × 105 22092 2600 1358
photon-jet
MadGraph 1.46 × 108 2.55 × 105 2732 312 183
QCD

Figure 6.1 shows the detector level comparisons of the data with MC simulations
for Δφ23 , Δrel pT,23 , and ΔS. In real scale, MadGraph interfaced with pythia 6
tune Z2 describes data quite well within statistic uncertainties. pythia 6 tune Z2
and pythia 8 tune 4C simulations predict a slightly smaller total number of events
than data by about 20–30%. But both two kinds of pythia simulations are with
larger statistic uncertainties, especially for pythia 8 tune 4C. There is an overall
agreement in shape of all DPS-discriminating distributions for the data and different
MC simulations except for some DPS-sensitive regions. MadGraph, interfaced with
pythia 6 tune Z2, is consistent with the data. At low and medium ΔS values, pythia 8
tune 4C and pythia 6 tune Z2 underestimate the data. At low Δrel pT,23 values,
pythia 6 tune Z2 shows much worse pT -balance of the di-jet pair than the other two
MC simulations.
6.1 DPS Performance at Detector Level 51

CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j
103 1
γ -j: pT γ , pT jet1 > 75GeV DATA γ -j: pT γ , pT jet1 > 75GeV DATA
2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV
PYTHIA8 4C PYTHIA8 4C

Events / 0.315 rad

γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
jets: | η|<2.4 PYTHIA6 Z2 jets: | η|<2.4 PYTHIA6 Z2
MADGRAPH+PY6 Z2 MADGRAPH+PY6 Z2

dN/N
10-1

102

2 2
1.8 1.8
MC / DATA

1.6 1.6

MC / DATA
1.4 1.4
1.2 1.2
1 1
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
Δφ23 (rad) Δφ23 (rad)
CMS Preliminary, s = 7 TeV, L = 36 pb , pp → γ +3j -1
CMS Preliminary, s = 7 TeV, L = 36 pb , pp → γ +3j -1
104
γ -j: pT γ , pT jet1 > 75GeV DATA γ -j: pT γ , pT jet1 > 75GeV DATA
2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV
PYTHIA8 4C PYTHIA8 4C
γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
Events / 0.101

jets: | η|<2.4 PYTHIA6 Z2 1 jets: | η|<2.4 PYTHIA6 Z2

103
MADGRAPH+PY6 Z2 MADGRAPH+PY6 Z2
dN/N

10-1
102

10-2
10
2.5 2.5
MC / DATA

2 2
MC / DATA

1.5 1.5
1 1
0.5 0.5
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Δrelp Δrelp
T,23
T,23

CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j
γ -j: pT γ , pT jet1 > 75GeV DATA γ -j: pT γ , pT jet1 > 75GeV DATA
2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV
PYTHIA8 4C 10 PYTHIA8 4C
104 γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
Events / 0.315 (rad)

jets: | η|<2.4 PYTHIA6 Z2 jets: | η|<2.4 PYTHIA6 Z2

MADGRAPH+PY6 Z2 MADGRAPH+PY6 Z2
1
dN/N

103

10-1
102

10-2
10

2.5 2
1.8
MC / DATA

2 1.6
MC / DATA

1.4
1.5 1.2
1
1 0.8
0.6
0.5 0.4
0.2
0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
ΔS (rad) ΔS (rad)

Fig. 6.1 The detector level comparisons of DPS discriminating observables, Δφ23 (top row),
Δrel pT,23 (center row), and ΔS (bottom row) between the data and various MC simulations in
real scale (left column) and normalization scale (right column), The lower panel shows the ratio
of the MC simulation to the data. The error bars on the ratio histogram represent the statistical
uncertainty of the data and simulated samples added in quadrature
52 6 Photon + 3 Jets Events at Detector Level

6.2 Purity of Isolated Prompt Photons

The photon purity, defined as the fraction of isolated prompt photons in data, of
reconstructed events satisfying the full selection is estimated with a binned extended
maximum likelihood fit to the σηη distribution with the expected signal and back-
ground components of photons. The shapes of signal and background distributions
are obtained as described in [69]. According to the definition of signal photon, the
photon source of DPS signal events can be direct photons and fragmentation photons.
The background photon is due to the electromagnetic decays of light mesons such
as π 0 and η. Events are selected with almost the same selection criteria except for
the σηη selection criteria in Table 4.2. These are performed by requiring two regions
of barrel, |η| < 1.4442, and endcap, 1.566 < |η| < 2.5, for photons and two bins of
E T,γ : 75 < E T,γ < 95 GeV and E T,γ > 95 GeV, in each region separately.
The signal distribution is extracted from the photon-jet samples of the MC simu-
lation of MadGraph interfaced with pythia 6 tune Z2 by taking the σηη distribution
of direct photons identified using MC truth information. These signal distribution
of direct photons are cross-checked with the signal distribution of fragmentation
photons, which is derived from the di-jet samples of the MC simulation of

CMS Preliminary 2010 CMS Preliminary 2010

1.4 s=7TeV, L=36 pb-1 direct photon 1.4 s=7TeV, L=36 pb-1 direct photon
pp→γ +3jet pp→γ +3jet
γ : 75<ET<95 (GeV), Barrel γ : E >95 (GeV), Barrel
1.2 γ : |η|<1.4442, Barrel 1.2 T
γ : |η|<1.4442, Barrel
fragmentation photon fragmentation photon

1 1
dN/N

dN/N

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.005 0.01 0.015 0.02 0.025 0.03 0 0.005 0.01 0.015 0.02 0.025 0.03
σiηiη σiηiη
1 1
CMS Preliminary 2010 CMS Preliminary 2010
s=7TeV, L=36 pb-1 direct photon s=7TeV, L=36 pb-1 direct photon
0.9 0.9
pp→γ +3jet pp→γ +3jet
γ : 75<E <95 (GeV), Endcap γ : ET>95 (GeV), Endcap
0.8 T
γ : 1.566<|η|<2.5, Endcap 0.8 γ : 1.566<|η|<2.5, Endcap
fragmentation photon fragmentation photon

0.7 0.7

0.6 0.6
dN/N

dN/N

0.5 0.5

0.4 0.4

0.3 0.3

0.2 0.2

0.1 0.1

0 0
0 0.01 0.02 0.03 0.04 0.05 0 0.01 0.02 0.03 0.04 0.05
σiηiη σiηiη

Fig. 6.2 Distributions of σηη for direct photons from the photon-jet MadGraph MC sample and
fragmentation photons from the di-jet MadGraph MC sample in different E T,γ bins and different
η regions. The two types of photons have the same σηη distributions
6.2 Purity of Isolated Prompt Photons 53

0.8 0.8
CMS Preliminary 2010 Data side-band CMS Preliminary 2010 Data side-band
s=7TeV, L=36 pb-1 s=7TeV, L=36 pb-1
0.7 pp→γ +3jet 0.7 pp→γ +3jet
γ : 75<E <95 (GeV), Barrel MC side-band γ : E >95 (GeV), Barrel MC side-band
T T
γ : |η|<1.4442, Barrel γ : |η|<1.4442, Barrel
0.6 MC background 0.6 MC background

0.5 0.5
dN/N

dN/N
0.4 0.4

0.3 0.3

0.2 0.2

0.1 0.1

0 0
0 0.005 0.01 0.015 0.02 0.025 0.03 0 0.005 0.01 0.015 0.02 0.025 0.03
σiηiη σiηiη
0.6 0.6
CMS Preliminary 2010 Data side-band CMS Preliminary 2010 Data side-band
s=7TeV, L=36 pb-1 s=7TeV, L=36 pb-1
pp→γ +3jet pp→γ +3jet
0.5 γ : 75<ET<95 (GeV), Endcap
MC side-band 0.5 γ : ET>95 (GeV), Endcap
MC side-band
γ : 1.566<|η|<2.5, Endcap γ : 1.566<|η|<2.5, Endcap
MC background MC background

0.4 0.4
dN/N

dN/N

0.3 0.3

0.2 0.2

0.1 0.1

0 0
0 0.01 0.02 0.03 0.04 0.05 0 0.01 0.02 0.03 0.04 0.05
σiηiη σiηiη

Fig. 6.3 Distributions of σηη for true background in MC (orange) and candidates from the ISOTRK
side-band in data (black) and MC (green) for different E T,γ bins and different η regions

MadGraph interfaced with pythia 6 tune Z2. All of the σηη distributions are shown
in Fig. 6.2.
The background σηη distribution is obtained from the data by requiring the iso-
lation criterion of the scalar sum of the charged-particle pT within ΔR = 0.4 for
photons larger than 2 GeV but less than 5 GeV, while keeping all other selection cri-
teria (Table 4.2) unchanged. This can be thought of as a side-band to the signal cut,
ISOTRK < 2 GeV/c. Under this procedure, the σiηiη background component shapes
are obtained for:
• The side-band in data, labelled Data side-band.
• The side-band in MC, labelled MC side-band.
• The signal region in MC, requiring the final state photon to have ISOTRK < 2
GeV/c and come from a hadron or have no match, labelled MC backgr ound.
From Fig. 6.3, we can find the agreement between all of three distributions is good
but the comparison is limited by the statistics in data in endcap region.
For each region, data is fitted with f (σηη ) = NS S(σηη ) + NB B(σηη ), where NS and
NB are the estimated numbers of signal and background events in different E T,γ and
different η region. The fit is performed using a binned extended maximum likelihood
by minimizing:
54 6 Photon + 3 Jets Events at Detector Level

CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j+X CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j+X
600 800
Data Data
γ : 75 < E < 95 GeV γ: E > 95 GeV
T,γ 700 T,γ
500 |η| < 1.4442, Barrel Fit result |η| < 1.4442, Barrel Fit result
Background 600 Background
400
500
γ candidates

γ candidates
300 400

300
200
200
100
100

0 0
0 0.005 0.01 0.015 0.02 0.025 0.03 0 0.005 0.01 0.015 0.02 0.025 0.03
σηη σηη
CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j+X CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j+X
200
140 Data Data
γ : 75 < E < 95 GeV 180 γ: E > 95 GeV
T,γ T,γ
1.566 < |η| < 2.5, Endcap Fit result 1.566 < |η| < 2.5, Endcap Fit result
120 160
Background Background
140
100
120
γ candidates

γ candidates

80
100

60 80

60
40
40
20
20

00 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
00 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05

σηη σηη

Fig. 6.4 Measured σηη distributions. The extended maximum likelihood fit result is overlaid in
each plot and the component for background is shown separately

Table 6.3 Estimated signal yield as a function of the photon η. For illustration, the purity in the
region σηη < 0.01 of barrel and the region σηη < 0.03 of endcap is also reported
η region, E T bin Fitted yield Purity
Barrel, 75–95 388.16 ± 11.82 85.9% ± 3.0% (stat.)
Barrel, 95–∞ 530.06 ± 16.02 85.5% ± 3.0%3.0 (stat.)
Endcap, 75–95 243.59 ± 10.72 78.8% ± 3.0%4.4 (stat.)
Barrel, 95–∞ 368.15 ± 11.56 90.3% ± 3.0%3.1 (stat.)


n
L = −lnL = −(NS + NB ) + Ni ln(NS Si + NB Bi ) (6.1)
i=1

where n is the number of bins in σηη , Ni is the observed number of events for the i-th
σηη bin and Si and Bi are the values of the corresponding components in that bin. The
results of fitting for different E T,γ and different η regions are shown in Fig. 6.4. The
raw yield of isolated photons (direct photons and fragmentation photons) is reported
in Table 6.3.
6.2 Purity of Isolated Prompt Photons 55

χ2 / ndf 64.25 / 52 χ2 / ndf 55.18 / 49

Constant 317.1 ± 5.7 Constant 321 ± 5.7
350 Mean 0.05048 ± 0.01434 350 Mean 0.06655 ± 0.01416
Sigma 0.9939 ± 0.0107 Sigma 0.9838 ± 0.0105
300 300

250 250

200 200

150 150

100 100

50 50

0 0
-10 -8 -6 -4 -2 0 2 4 6 8 10 -10 -8 -6 -4 -2 0 2 4 6 8 10

χ2 / ndf 55.87 / 47 χ2 / ndf 51.85 / 49

Constant 326.3 ± 5.8 Constant 315.8 ± 5.6
Mean 0.02801 ± 0.01388 Mean 0.04449 ± 0.01438
350 Sigma 0.9673 ± 0.0105 350 Sigma 1.001 ± 0.011

300 300

250 250

200 200

150 150

100 100

50 50

0 0
-10 -8 -6 -4 -2 0 2 4 6 8 10 -10 -8 -6 -4 -2 0 2 4 6 8 10

Fig. 6.5 Distributions of pull from the fitter test. The input values of shape parameters have been
Gaussian randomized. From the top left to the bottom are for (1) 75 < E T,γ < 95 GeV and |η| <
1.4442, (2) 75 < E T,γ < 95 GeV and 1.566 < |η| < 2.5, (3) E T,γ > 95 GeV and |η| < 1.4442 and
(4) E T,γ > 95 GeV and 1.566 < |η| < 2.5

Table 6.4 The fit bias in η region, E T bin Fitter bias (%)
different E T,γ and different η
regions Barrel, 75–95 5.0
Barrel, 95–∞ 6.7
Endcap, 75–95 2.8
Barrel, 95–∞ 4.4

Test of unbinned likelihood fitter is performed by generating pseudo-experiments

and fitting with the same fitter settings that are used to fit real data. In the pseudo-
experiments, input values of number of signal and background events, Ns and Nb ,
and other shape parameters are Gaussian randomized where the Gaussian means and
widths are the values determined in real-data fit.
Input
Pull distributions of different E T,γ and different η regions, (NsFit − Ns /NsFit ),
Input
is fitted to a Gaussian function, where the Ns is the number of input and the
(Ns ) is the number of fitting result. Figure 6.5 gives the pull distributions for all
Fit

E T,γ and different η regions and the Gaussian fit results. If the pull mean differs from
Input
zero it means that the fitter model is not ideal and a fit bias (NsFit − Ns /NsFit ) is
introduced into the results of photon purity. Table 6.4 summarizes the fit bias of for
photons in different E T,γ bins and barrel and endcap regions, respectively. The fit
bias is included in the final systematic uncertainty of photon purity study.
The background component shape, the σηη < distribution for data in a ISOTRK
side-band of 2 < ISOTRK < 5 GeV/c, may systematically affect the results. Instead of
using side-band distribution of data, Data side-band, we perform the same fitting
method for data by using the other two kinds of background templates, MC side-
band and MC backgr ound. Then the difference will be introduced as the systematic
uncertainty of background shape. Besides, the background templates for fitting are
56 6 Photon + 3 Jets Events at Detector Level

Table 6.5 The final systematic uncertainty of the background component shape for different E T,γ
and different η regions
η region, E T bin Background component shape (%) Background fluctuation (%)
Barrel, 75–95 3.0 1.4
Barrel, 95–∞ 7.5 1.4
Endcap, 75–95 1.5 2.4
Barrel, 95–∞ 4.3 1.1

Table 6.6 The photon purity of reconstructed events satisfying the full selection in the different
E T,γ and different η regions
E T,γ bin, GeV |η| < 1.4442 1.566 < |η| < 2.5
75–95 85.9% ±3.0% (stat.) ±4.0% (syst.) 78.8% ±4.4% (stat.) ±4.0% (syst.)
95–∞ 85.5% ±3.0% (stat.) ±10.2% (syst.) 90.3% ±3.1% (stat.) ±6.2% (syst.)

fluctuated in order to take into account the effects of insufficient sample size on
producing the background fitting templates. The final systematic uncertainties of the
background component shapes are summarized in Table 6.5.
The final result of photon purity is listed in Table 6.6 for each E T,γ bin in the barrel
and endcap regions. The dominant contribution of systematic uncertainty is caused
by the background distribution [69]. For σηη < 0.01 in the barrel region, the photon
purity in the data after full selection is about 86%. However, for σηη < 0.03 in the
endcap region, the photon purity increases from 79% at E T,γ = 75 GeV to 90% at
E T,γ = 95 GeV.
Chapter 7
Correction and Unfolding

Our goal is to provide distributions of data which can be compared to predictions

of various Monte Carlo generators without reference to any detector. The relation-
ship between the desired true distributions, which is represented as f true (xtrue ), and
the measured distribution at detector-reconstruction level, which is represented as
f reco (xreco ), can be expressed as:

f (x ) = R(xtrue , xreco )f true (xtrue )dxtrue
reco reco
(7.1)

where R(xtrue , xreco ) is the response matrix. The response matrix encapsulates the
detector response for the measurement. Response matrices are obtained from Mad-
Graph interfaced with pythia 6 tune Z2 full simulation sample here. The steps
to unfolding involves building a response matrix that maps the true distribution at
generator-truth level to the measured distribution at detector-reconstruction level.
For a given number of events in a bin of the true distribution, the response matrix
element R(xtrue , xreco ) gives the fraction of events that end up in the measured bin.
The response matrices are constructed for each measured observables, as follows.

• Events are accepted if they can satisfy the generator-truth level selection and
detector-reconstruction level selection with the same kinematic acceptance.
• If the event satisfies the generator-truth level acceptance criteria, but has no coun-
terpart inside the detector acceptance, such event is tagged as ”missing” and is
recorded as lost due to detector inefficiency.
• If the event accepted at detector-reconstruction level, but has no counterpart at
generator-truth level, such event is tagged as “fake” and is also recorded in the
response matrices.

© Springer Nature Singapore Pte Ltd. 2017 57

Y.-H. Chang, Study of Double Parton Scattering in Photon + 3 Jets Final State,
Springer Theses, DOI 10.1007/978-981-10-3824-2_7
58 7 Correction and Unfolding

7.1 Acceptance, Background, Purity and Stability

In order to further investigate the reliability of response matrices of the measured

observables, the migration effects inside and outside the defined phase space have to
be deeply studied. Due to the detection inefficiencies or resolution, event selected at
detector-reconstruction level may not have its corresponding event at generator-truth
level, or event selected at generator-truth level may not be measured at detector-
reconstruction level. Such phenomena are defined as the migration effect outside the
phase space. Furthermore, the quantities of measured observables of events selected
at both two level may correspond to different values or even belong to different bins.
This is labeled as the migration effect inside the phase space. Some quantities used
to study the migration effects are introduced in the following paragraphs. It should
be noted that the events selected at both generator-truth and detector-reconstruction
level are labeled as “matched events”.
For the migration effects outside the phase space, we study the corresponding
quantities, acceptance and background, respectively. The acceptance is the fraction
of events selected at generator-truth level which have their counterparts at detector-
reconstruction level. That is, the fraction of generated event at generator-truth level
which can be really measured at detector-reconstruction level. The background is the
fraction of measured events at detector-reconstruction level which actually have no
their corresponding events at generator-truth level. In other words, they are simply
fake events due to the detection effects, which can pass the corresponding selection
criteria at detector-reconstruction level. The detailed formulas of these two quantities
are shown below:
Nmatched (Egen ∈ bin i)
Ai = (7.2)
Nall (Egen ∈ bin i)

Nmatched (Edet ∈ bin i)

Bi = 1 − (7.3)
Nall (Edet ∈ bin i)

For the migration effects inside the phase space, purity and stability are typical
quantities used in related investigation. The purity is defined as the fraction of the
matched events that remain in the same bin as the detector level selection, while the
stability is the fraction of the matched events that stay in the same bin as the generator
level selection. The detailed formulas of these two quantities are shown below:
Nmatched (Edet ∈ bin i ∧ Egen ∈ bin i)
Pi = (7.4)
Nmatched (Edet ∈ bin i)

Nmatched (Edet ∈ bin i ∧ Egen ∈ bin i)

Si = (7.5)
Nmatched (Egen ∈ bin i)

Acceptance, background, purity and stability are studied for every measured dis-
criminating observable. The related results are shown in Fig. 7.1. Basically, a good
7.1 Acceptance, Background, Purity and Stability 59

CMS Preliminary 2010

1.4 CMS Preliminary 2010
s=7TeV, L=36 pb-1 Acceptance s=7TeV, L=36 pb-1 Acceptance
1.2 pp→γ +3jet
Purity
pp→γ +3jet
Purity
Stability 1.2 Stability
Background Background

Events / 0.101
Events / 0.315 rad
1
1
0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.5 1 1.5 2 2.5 3 0 0.2 0.4 0.6 0.8 1
Δφ(jet2,jet3) (rad) ΔrelPT(Jet2,jet3)

1.4
CMS Preliminary 2010
s=7TeV, L=36 pb-1 Acceptance
Purity
1.2 pp→γ +3jet
Stability
Background
Events / 0.315 (rad)

0.8

0.6

0.4

0.2

0
0 0.5 1 1.5 2 2.5 3
ΔS (photonjet,dijet) (rad)

Fig. 7.1 Acceptance, purity, stability and background for Δφ23 (upper left), Δrel pT,23 (upper right)
and ΔS (lower)

measurement should have high purity, high stability, high acceptance but low back-
ground over the whole phase space. These properties directly reflect the quality of
measurement. All the related results are performed by analyzing MadGraph inter-
faced with pythia 6 tune Z1 full simulation sample.
The acceptance for each measured observable is overall around 0.3–0.4 without
significant fluctuations. The phenomenon of low acceptance had been investigated. It
is due to the photon identification and the selection of 2nd primary vertex. Basically,
we present a comparison between original study and a special study by removing the
photon identification selection, which is described in Sect. 4.3, except for selection
criteria of the pT threshold and η acceptance region. In order to have higher photon
purity, photon isolation criteria are introduced to reduce the background rate, caused
by any possible fake photons or photons come from the neutral light meson decay,
such as π 0 and η. But the cost is that some of signal photons fail the selection cuts.
Moreover, in order to reduce the influence from pileup, a selection on the 2nd vertex
described before will make the analysis only focus on events with only one vertex or
events whose pileup is relatively soft. The acceptance decreases due to the composite
influence of these effects.
60 7 Correction and Unfolding

On the other hand, the background curves for different observables are fluctu-
ating a bit more around 0.1–0.3. As described before, background events are due
to the unmatched events, especially those events which are accepted at the detector-
reconstruction level but are not presented at the generator-truth level because of detec-
tion inefficiencies, migration effects, or pileup events, where one or more selected
particles at detector-reconstruction level come from pileup interactions; such events
can not have a counterpart at generator-truth level because there is no information of
particles from the pileup.
The purity and stability are overall above 0.6 for all the measured observ-
ables except for the low degree region of ΔS. For the angular observables, i.e.
Δφ23 , the curves are around 0.7–0.9, while for the pT balance observables, i.e.
Δrel pT,23 , the purity and stability are relatively lower and around 0.5–0.9 due to the
effect of pT resolution of selected photon and jets. And discriminating observables
involved more jets will also have relatively lower purity and stability. In the region
ΔS < 2.5, the purity and stability curves are fluctuating a bit more around 0.2–0.4.
ΔS is a composite discriminating observable contain both angular and momentum
properties. The lower purity and stability in the low ΔS region indicate the migration
effect would be relatively serious.

7.2 Unfolding with ‘RooUnfold’

Measurements must be unfolded for detector effects. A direct comparison is then per-
mitted of distributions from different experiments without knowledge of the detector
response for each experiment. Data after unfolding can facilitated the automated
tuning of Monte Carlo generators using in multiple measurements, which can be a
great feedback to theorists to improve their models.
In this analysis, the iterative (D’Agostini) unfolding method [77], which is imple-
mented in the RooUnfold package [78], is mainly used to remove the known effects
of measurement resolutions, systematic biases, and detector inefficiency.
This unfolding method is based on Bayes’ theorem in the following form:
P(Ek | Cj )P(Cj )
P(Cj | Ek ) = (7.6)

nC
P(Ek | Cj )P(Cj )
j=1

where P(Cj ) is the probability of the j-th cause (generator-truth level object),
P(Ek | Cj ) is the conditional probability of j-th cause to produce the k-th effect
(detector-reconstruction level object), and P(Cj | Ek ) is the probability that the k-th
effect is due to the j-th cause.
7.2 Unfolding with ‘RooUnfold’ 61

By replacing the corresponding term with into our formalism:

tj Rjk
P(Cj | Ek ) =  (7.7)
tj Rjk
k

where tj denotes the content of the j-the truth bin.

The algorithm starts by using an initial distribution as an estimate of the final dis-
tribution. As we mentioned before, we use the relevant distribution of MadGraph
interfaced with pythia 6 tune Z2 as the initial distribution for the first iteration,
to produce from which the algorithm produces an improved estimate of the final
distribution. Each subsequent iteration processes the output distribution of the pre-
vious iteration, to produce yet another improved estimate. Since each iteration uses
Bayes Theorem in an intermediate step, the algorithm is often referred to as Bayesian
Iterative Unfolding.
We describe the mechanical aspects of the algorithm as it pertains to the current
analysis, and how it uses the response matrix R to perform the unfolding. The process
of iteration can be stopped until the output has stabilized. This algorithm uses a
regularization parameter to prevent statistical fluctuations in the data from appearing
as structure in the unfolded distribution and also properly takes into account bin-to-
bin migration.

• Step 1. Let p0 ≡ (p1 , p2 , ..., pM ) be an initial set of probabilities (derived from the
input distribution,
 or even a constant value) for an event to be found in each bin,
and ntotal = ni be the total number of entries.
i
• Step 2. Define
μ̂0 ≡ ntotal p0 (7.8)
⎛ ⎞
1 
N
⎜ Rij pi ⎟
μ̂i = ⎜ ⎟ nj (7.9)
εi ⎝ ⎠
j=1 Rkj pk
k

and we update a new value for μ by using the response matrices Rjk .
• Step 3. Extract a new set of probability p

μ̂k
pk = (7.10)
ntotal

• Step 4. Iterate step 2 and 3 on Monte Carlo simulation sample, until the change
in Λ2 between iterations indicates the corrected distribution has stabilized. A
number of 4–6 iterations has turned to be sufficient to get a stable result for the
matrix inversion.
62 7 Correction and Unfolding

In order to check the consistence of unfolding, we compare the results by imple-

menting two different methods for unfolding: the iterative unfolding method and
singular value decomposition (SVD) method [79], we separate the MadGraph sam-
ples into two sets, called Set 1 and Set 2. The SVD unfolding method takes a totally
different way to extract the response matrix. To make short of the matter, A SVD of
a real m × n response matrix R is its factorization of the form:

R = U S V T, (7.11)

where U is an m × m orthogonal matrix, U U T = U T U = I(Iisidentitymatrix), V

is an n × n orthogonal matrix, V V T = V T V = I. S is an m × n diagonal matrix
with non-negative diagonal elements:

Sij = 0 for i = j, Sii ≡ si ≥ 0, (7.12)

where the numbers si are singular values of the response matrix R.

10 4
CMS Preliminary 2010 CMS Preliminary 2010
Bayesian Unfolding Bayesian Unfolding
s = 7 TeV, L = 36 pb-1, pp → γ +3jets s = 7 TeV, L = 36 pb-1, pp → γ +3jets
γ jet: p > 75 GeV, dijet: pT > 20 GeV SVD Unfolding γ jet: p > 75 GeV, dijet: pT > 20 GeV SVD Unfolding
3
T
γ : |η| < 1.4442, 1.566 < |η| < 2.5 Gen. MC truth 10 T
γ : |η| < 1.4442, 1.566 < |η| < 2.5 Gen. MC truth
10 2 Reco. Measurement Reco. Measurement
jets: |η| < 2.4 jets: |η| < 2.4

2
10
σ [pb]
σ [pb]

10 10

1 10-1
0 0.5 1 1.5 2 2.5 3 0 0.2 0.4 0.6 0.8 1
Δφ(jet2,jet3) (rad) ΔrelPT(jet2,jet3)
3.5
3 4
Reco. Level ratio

3.5
Reco. Level ratio

2.5 3
2 2.5
1.5 2
1.5
1
1
0.5 0.5
0 0
0 0.5 1 1.5 2 2.5 3 0 0.2 0.4 0.6 0.8 1
5
10
CMS Preliminary 2010
-1 Bayesian Unfolding
s = 7 TeV, L = 36 pb , pp → γ +3jets
4
10 γ jet: p > 75 GeV, dijet: p > 20 GeV
T T
SVD Unfolding

γ : |η| < 1.4442, 1.566 < |η| < 2.5 Gen. MC truth
jets: |η| < 2.4 Reco. Measurement
3
10
σ [pb]

2
10

10

-1
10
0 0.5 1 1.5 2 2.5 3
ΔS (photonjet,dijet) (rad)

3.5
Reco. Level ratio

3
2.5
2
1.5
1
0.5
0
0 0.5 1 1.5 2 2.5 3

Fig. 7.2 Cross section measurements obtained from the Set 2 MadGraph sample at detector-
reconstruction and generator-truth level and unfolded distributions obtained with the SVD correction
and the Bayesian method: Δφ23 (upper left), Δrel pT,23 (upper right) and ΔS (lower). The response
matrices are extracted from the Set 1 MadGraph sample
7.2 Unfolding with ‘RooUnfold’ 63
4
10
CMS Preliminary 2010 CMS Preliminary 2010
-1 -1
s = 7 TeV, L = 36 pb , pp → γ +3jets Corrected data, Bayesian Unfolding s = 7 TeV, L = 36 pb , pp → γ +3jets Corrected data, Bayesian Unfolding
γ jet: p > 75 GeV, dijet: pT > 20 GeV Corrected data, SVD Unfolding 3 γ jet: p > 75 GeV, dijet: pT > 20 GeV Corrected data, SVD Unfolding
T
γ : |η| < 1.4442, 1.566 < |η| < 2.5
10 T
γ : |η| < 1.4442, 1.566 < |η| < 2.5
2
10 Uncorrected data Uncorrected data
jets: |η| < 2.4 jets: |η| < 2.4

2
10

σ [pb]
σ [pb]
10
10

1 10
-1

0 0.5 1 1.5 2 2.5 3 0 0.2 0.4 0.6 0.8 1

Δφ(jet2,jet3) (rad) ΔrelP (jet2,jet3) T

Corrected data/Uncorrected data

Corrected data/Uncorrected data

4.5 4.5
4 4
3.5 3.5
3 3
2.5 2.5
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 0.5 1 1.5 2 2.5 3 0 0.2 0.4 0.6 0.8 1

105
CMS Preliminary 2010
-1
s = 7 TeV, L = 36 pb , pp → γ +3jets Corrected data, Bayesian Unfolding
104 γ jet: p > 75 GeV, dijet: pT > 20 GeV Corrected data, SVD Unfolding
T
γ : |η| < 1.4442, 1.566 < |η| < 2.5
Uncorrected data
jets: |η| < 2.4
103
σ [pb]

2
10

10

10-1
0 0.5 1 1.5 2 2.5 3
ΔS (photonjet,dijet) (rad)
Corrected data/Uncorrected data

0
0 0.5 1 1.5 2 2.5 3

Fig. 7.3 Cross section measurements obtained with the complete 2010 data samples at detector-
reconstruction level and unfolded distributions obtained with the SVD correction and the Bayesian
method: Δφ23 (upper left), Δrel pT,23 (upper right) and ΔS (lower). The response matrices are
extracted from MadGraph sample

At first, we extract the response matrices by using Set 1 MadGraph sample.

Set 2 MadGraph sample is then unfolded through the two different methods, the
SVD correction and the Bayesian method, implemented in RooUnfold package. The
results are shown in Fig. 7.2. This study can be also devoted to the validation of
MadGraph samples for consistence test. It can be found that systematic difference
in sampling is neglected. The MC samples of MadGraph is self-consistent.
Furthermore, the corrected distributions of data are checked by directly compar-
ing the results processed with different unfolding method. The related results are
shown in Fig. 7.3, where the response matrices are extracted by using the complete
MadGraph sample. The two methods are always comparable within uncertainties
and the SVD method is also able to give reliable results. The corrected distribution of
discriminating observables of data are compared to various Monte Carlo predictions
in the Chap. 9.
Chapter 8
Systematic Uncertainties

Several studies are performed to assess in which way quantities entering the absolute
cross section measurements and normalized shape distributions can be systematically
affecting the results. Analyses using photons and jets have to consider in particular
the impact of their corresponding energy scale and energy resolution. The uncer-
tainty on the correction of photon identification efficiency needs to be taken into
account. Studies on the 2nd primary vertex selection, jet-vertex association algo-
rithm, pileup reweighting, luminosity, trigger efficiency and model dependence of
unfolding are also performed. The sample size of MC simulation are treated as an
additional systematic uncertainty since the sample size affect the output response
matrix of unfolding.
All of the systematic uncertainties are expressed as variations with respect to
the nominal values, Ni /Ni − 1, where Ni is the perturbed bin content and Ni is
the nominal bin content of i th bin in the measured distribution. That is, the average
discrepancies between the perturbed and the nominal distribution are taken bin-by-
bin and referred as the uncertainties. Whenever an effect is found to be smaller than
0.1% it is not reported.
For both of real cross sections and normalized shape distributions, the minimum
and the maximum value of the uncertainties along the whole phase space are listed
together with their average. This average is taken as the reference of systematic
uncertainty.

8.1 Photon Energy Scale Uncertainty

The uncertainty of photon energy scale was estimated from Z → e+ e− events in

both data and signal control samples. The analysis was repeated shift the photon
energy scale by the full 0.6% uncertainty in the barrel region and 1.5% in the endcap
region [71–73]. The resulting effects in the event yield and normalized shape are
© Springer Nature Singapore Pte Ltd. 2017 65
Y.-H. Chang, Study of Double Parton Scattering in Photon + 3 Jets Final State,
Springer Theses, DOI 10.1007/978-981-10-3824-2_8
66 8 Systematic Uncertainties

Table 8.1 Effect of the photon energy scale uncertainty on the event yield for measurements of
different discriminating observables in absolute cross section scale. The maximum, the minimum
and the average value are listed
Measured observable Δφ23 (%) Δrel pT,23 (%) ΔS (%)
Max. 6.8 5.0 3.9
Min. 1.5 1.2 1.8
Avg. 3.9 2.8 2.7

Table 8.2 Effect of the photon energy scale uncertainty on the normalized shape for measurements
of different discriminating observables in normalization scale. The maximum, the minimum and
the average value are listed
Measured observable Δφ23 (%) Δrel pT,23 (%) ΔS (%)
Max. 4.6 2.8 4.2
Min. 0.7 0.4 0.3
Avg. 2.2 1.2 0.8

shown in Tables 8.1 and 8.2. A contribution is around 2–5% for the absolute cross
section and around 1–2% for the normalized shape distributions.

8.2 Photon Energy Resolution Uncertainty

The difference of photon energy resolution between data and MC is investigated

and assigned as a systematic uncertainty for this analysis. It is found that resolution
in data is worse by comparing Z → e+ e− data and MC [80]. In order to study
this uncertainty, the pT of photon is smeared out around its true value by matching
every photon at the detector level to the closest photon at the generator level with
the hadronic energy surrounding the photon to be less than 5 GeV within ΔR < 0.4.
The matching
 is performed through an angular cone algorithm with an aperture
ΔR ≡ (Δη)2 + (Δφ)2 = 0.15. The MC reconstructed photon energy is then
gen gen
smeared in the following way: ( pTreco smear ed) = pT ±1.3×( pTreco − pT ) where the
factor 1.3 is chosen as the typical ratio between data and MC energy resolutions [81]
gen
and pT , pTreco are transverse momenta of the two matched photons at the generator
level and detector level. The resulting effects in the event yield and normalized shape
are shown in Tables 8.3 and 8.4. A contribution is around 0.1–1% for both the absolute
cross section and the normalized shape distributions but except for the results of ΔS.
Relatively larger systematic uncertainties of photon energy resolution could be up to
2% for ΔS.
8.3 Jet Energy Scale Uncertainty 67

Table 8.3 Effect of the photon energy resolution uncertainty on the event yield for measurements
of different discriminating observables in absolute cross section scale. The maximum, the minimum
and the average value are listed
Measured observable Δφ23 (%) Δrel pT,23 (%) ΔS (%)
Max. 1.1 1.7 6.2
Min. 0.3 0.2 0.4
Avg. 0.6 0.8 2.4

Table 8.4 Effect of the photon energy resolution uncertainty on the normalized shape for measure-
ments of different discriminating observables in normalization scale. The maximum, the minimum
and the average value are listed
Measured observable Δφ23 (%) Δrel pT,23 ΔS (%)
Max. 0.6 2.2% 6.6
Min. 0.1 – 0.2
Avg. 0.3 0.7% 2.2

8.3 Jet Energy Scale Uncertainty

Besides the high quality requirement is needed for photon, this analysis is also based
on the selection of high jet multiplicity. The detailed understanding of the jet energy
scale is of crucial importance for analyses involved jets. The corresponding uncer-
tainty of jet energy scale is believed to play an important role that affects the total
systematic uncertainty. The transverse momentum of each jet is varied up and down
according to the uncertainty associated to the reconstructed jet pT [82]. The resulting
effects in the event yield and normalized shape are shown in Tables 8.5 and 8.6. A
contribution is around 5–17% for the absolute cross section and around 1–9% for
the normalized shape distributions. Relatively, the effect of jet energy scale makes
smaller uncertainties in those discriminating observables connected to the configura-
tion of the azimuthal angle for the absolute cross section. But the situation is opposite
for the normalized shape distribution.

Table 8.5 Effect of the jet energy scale uncertainty on the event yield for measurements of different
discriminating observables in absolute cross section scale. The maximum, the minimum and the
average value are listed
Measured observable Δφ23 (%) Δrel pT,23 (%) ΔS (%)
Max. 11.1 11.1 29.1
Min. 2.9 7.2 6.8
Avg. 7.5 9.3 17.2
68 8 Systematic Uncertainties

Table 8.6 Effect of the jet energy scale uncertainty on the normalized shape for measurements of
different discriminating observables in normalization scale. The maximum, the minimum and the
average value are listed
Measured observable Δφ23 (%) Δrel pT,23 (%) ΔS (%)
Max. 6.0 2.2 18.4
Min. 0.6 0.5 2.3
Avg. 2.7 1.5 8.9

8.4 Jet Energy Resolution Uncertainty

The jet energy response does not exactly correspond to the true value of the measured
physical quantity. Measurements show that the jet energy resolution in data is worse
than in the simulation. The method of studying the effect of jet energy resolution is
similar to what had
 been done for photon energy resolution. Except for the matching
aperture ΔR ≡ (Δη)2 + (Δφ)2 , the value would be 0.3. The MC reconstructed jet
gen gen
energy was smeared in the following way: ( pTreco smear ed) = pT ± A×( pTreco − pT )
where the factor A is the official parameters of the detector resolution for the 2010
gen
data as described in [82] and pT , pTreco were transverse momenta of the two matched
jets at the generator level and detector level. The resulting effects in the event yield
and normalized shape are shown in Tables 8.7 and 8.8. A contribution is around
5–7% for the absolute cross section and around 4–6% for the normalized shape
distributions.

Table 8.7 Effect of the jet energy resolution uncertainty on the event yield for measurements of
different discriminating observables in absolute cross section scale. The maximum, the minimum
and the average value are listed
Measured observable Δφ23 (%) Δrel pT,23 (%) ΔS (%)
Max. 9.7 13.2 10.9
Min. 0.8 0.8 2.4
Avg. 5.2 7.0 7.0

Table 8.8 Effect of the jet energy resolution uncertainty on the normalized shape for measurements
of different discriminating observables in normalization scale. The maximum, the minimum and
the average value are listed
Measured observable Δφ23 (%) Δrel pT,23 (%) ΔS (%)
Max. 8.2 8.1 6.2
Min. 1.1 1.2 1.2
Avg. 4.0 5.5 5.0
8.5 Model Dependence Uncertainty 69

Table 8.9 Effect of the model dependence uncertainty on the event yield for measurements of
different discriminating observables in absolute cross section scale. The maximum, the minimum
and the average value are listed
Measured observable Δφ23 (%) Δrel pT,23 (%) ΔS (%)
Max. 10.6 10.5 18.0
Min. 3.7 1.9 4.0
Avg. 8.0 6.9 11.6

Table 8.10 Effect of the model dependence uncertainty on the normalized shape for measurements
of different discriminating observables in normalization scale. The maximum, the minimum and
the average value are listed
Measured observable Δφ23 (%) Δrel pT,23 (%) ΔS (%)
Max. 3.9 3.9 11.9
Min. 1.0 1.1 3.1
Avg. 2.5 2.3 5.7

8.5 Model Dependence Uncertainty

The half discrepancy between MadGraph and pythia-6 samples at the detector level
is used to study the effect arising from different physic models. The data are unfolded
with response matrices obtained from the two different MC generators pythia 6 and
MadGraph interfaced with pythia 6 tune Z2. The resulting effects in the event yield
and normalized shape are shown in Tables 8.9 and 8.10. A contribution is around
7–12% for the absolute cross section and around 2–6% for the shape distributions.

8.6 Photon Identification Efficiency Uncertainty

Systematic uncertainty for photon identification efficiency is assigned for the differ-
ence between the efficiency results in MC and those measured from data by using
the “Tag and Probe” technique. Both corrections of the isolation and the pixel seed
veto efficiency are taken into account for evaluating their effects on the final results,
at described in Sect. 4.3. The efficiency of the isolation criteria, εisolation , in the data is
found to be higher (lower) than in simulation in the barrel (endcap) region, therefore
a scale factor of 1.008 ± 0.005 (0.984 ± 0.005) is applied to the simulation. The
photon pixel veto efficiency, εpixel veto , in the data is found to be higher (lower) than
in simulation in the barrel (endcap) region, therefore a scale factor of 1.009 ± 0.029
(0.959 ± 0.062) is included in the simulation [70]. Such uncertainties arising from
the correction of photon identification efficiencies are quite uniform. Around 4% was
estimated in the real cross section distributions and 0.1% in the normalized shape
distributions.
70 8 Systematic Uncertainties

Fig. 8.1 Distribution of the 700

χ2 / ndf
Const
16.39 / 9
2062 ± 46.1
CMS Preliminary 2010 Mean 2.461 ± 0.036
number of pile-up in data s=7TeV, L=36 pb-1
600
with a fitted Poisson function
500

number of events
400

300

200

100

0
0 2 4 6 8 10 12 14 16 18 20
number of pile-up

8.7 Pileup Reweighting Uncertainty

The number of interactions in the data is estimated from the measured luminosity
in each bunch crossing times an average total inelastic cross section. The uncer-
tainty due to pileup reweighting can be estimated by shifting the overall mean of the
“reweighting target” (i.e., the data distributions) number of interactions up or down
by the corresponding amount, then redoing the reweighting. Figure 8.1 shows the
estimated distribution of number of pileup with a fitted Poisson function. By taking
a variation ±3.6% [83] for shifting the mean of the fitted Poisson function is suffi-
cient to cover the uncertainties due to pileup modeling. The resulting effects in the
event yield and normalized shape are shown in Tables 8.11 and 8.12. A contribution
is around 2–3% for the absolute cross section and less than 1% for the normalized
shape distributions.

8.8 Second Primary Vertex Selection Uncertainty


An mentioned in the Sect. 4.2, events with pT 2i−th track (2nd vertex) < 10 GeV 2 are
accepted, where pT i−th track is the transverse momentum of the i-th track associated to
the 2nd vertex. In order to evaluating the uncertainty related to this selection criterion,

Table 8.11 Effect of the pileup reweighting uncertainty on the event yield for measurements of
different discriminating observables in absolute cross section scale. The maximum, the minimum
and the average value are listed
Measured observable Δφ23 (%) Δrel pT,23 (%) ΔS (%)
Max. 2.7 4.2 4.8
Min. 2.4 2.3 1.7
Avg. 2.6 2.9 2.9
8.8 Second Primary Vertex Selection Uncertainty 71

Table 8.12 Effect of the pileup reweighting uncertainty on the normalized shape for measurements
of different discriminating observables in normalization scale. The maximum, the minimum and
the average value are listed
Measured observable Δφ23 Δrel pT,23 (%) ΔS (%)
Max. 0.9% 1.9 4.1
Min. – 0.1 0.1
Avg. 0.3% 0.7 1.0

the pT of track associated to 2nd primary vertex is varied up and down according
to its corresponding uncertainty [84]. And the resulting effects in the event yield
and shape are quite uniform. Around 1.4% was estimated in the real cross section
distributions and around 0.5% in normalized shape distributions.

8.9 Jet-Vertex Association Uncertainty

The detailed definition of jet-vertex association algorithm had been described in

Sect. 4.4. It can reduce the chance to have selected jets from pileup. The systematic
uncertainty related to this jet selection is simply studied by comparing the results with
the jet-vertex association algorithm and the results without the jet-vertex association
algorithm. Related uncertainty is quite uniform for each measured distributions. Less
than 0.1% uncertainties are estimated in both the real cross section distributions and
normalized shape distributions.

8.10 The Sample Size of Monte Carlo Simulation

The sample size of MC simulation is treated as an additional systematic uncertainty

since it affects the correction factors extracted from response matrix for unfold-
ing. The statistical uncertainty in each bin of the distributions on the number of
reconstructed events satisfying the full selection is 4–8% and is used to assign the

Table 8.13 Statistics uncertainties of Monte Carlo (MadGraph) simulation. The maximum, the
minimum and the average value are listed
Measured observable Δφ23 (%) Δrel pT,23 (%) ΔS (%)
Max. 7.3 10.5 11.2
Min. 2.4 2.0 1.4
Avg. 4.0 6.4 8.0
72 8 Systematic Uncertainties

systematic uncertainty due to the size of MC samples in both absolute cross section
scale and normalization scale. Table 8.13 shows the related uncertainties for different
discriminating observables.

8.11 HLT Efficiency Correction Uncertainty

From Table 8.14 and the pT threshold used in this analysis, we can know only Pho-
ton70 trigger has inefficiency problem in the +η endcap region. Furthermore, the
inefficiency only exists for events whose photon’s E T is between 75 and 80 GeV.
From Fig. 8.2, only 16 events with HLT_Photon70 requirement are in the +η
endcap region. Hence, the correction of HLT efficiency in such region should be
+0.5%
negligible. The HLT efficiency of data has been measured, 99.1%−1.2% , in the con-
sidered kinematic region. We simply quote the systematic uncertainty of HLT by
taking every uncertainty of HLT_Photon70 and the fractions of events in different
region into account. The systematic uncertainty of HLT is 1.2% and assigned to both
results in absolute and normalization scales.

Table 8.14 Efficiency of single photon HLT paths as a function of the photon transverse energy in
barrel and endcaps
HLT path name Photon E T ( GeV) −η Endcap Barrel +η Endcap
HLT_Photon70 75 - Inf. 100 +0.
−11.6 100 +0.
−1.4 90.0 +5.5
−7.9

CMS Preliminary 2010, s = 7 TeV, L = 36.14 pb-1, pp → γ +3jets CMS Preliminary 2010, s = 7 TeV, L = 36.14 pb-1, pp → γ +3jets
100 100
γ -jet: pT > 75 GeV γ -jet: pT > 75 GeV
90 2jets: p > 20 GeV 90 2jets: p > 20 GeV
T T
γ : 1.566<η< 2.5, +η endcap γ : 1.566<η< 2.5, +η endcap
80
jets: |η|<2.4 DATA 80
jets: |η|<2.4
DATA, HLT_Photon70
Events / 5 GeV

Events / 5 GeV

70 70

60 60

50 50

40 40

30 30

20 20

10 10

0 0
100 150 200 250 300 100 150 200 250 300
PT (γ ) (GeV) PT (γ ) (GeV)

Fig. 8.2 Photon pT distribution of data in the in the +η endcap region. a all passed events. b events
with HLT_Photon70 requirement
8.12 Total Systematic Uncertainty 73

8.12 Total Systematic Uncertainty

For the separate contributions, jet energy scale, jet energy resolution, photon energy
scale, model dependence and the sample size of MC simulation dominate the sys-
tematic uncertainties in both absolute scale of real cross section and normalization
scale of shape. Basically, the systematic uncertainties in the normalization scale are
smaller than those in the absolute cross section scale.

0.4 0.3
CMS Preliminary 2010 Total Uncertainty CMS Preliminary 2010 Total Uncertainty
Jet Energy Scale Jet Energy Scale
s=7TeV, L=36 pb-1 Jet Energy Resolution s=7TeV, L=36 pb-1 Jet Energy Resolution
0.35 pp→γ +3jet Photon ID Efficiency pp→γ +3jet Photon ID Efficiency
Photon Energy Scale 0.25
Relative uncertainty

Relative uncertainty
Photon Energy Scale
Photon Energy Resolution
Photon Energy Resolution
Model dependence
0.3 Model dependence
MC Statistics
HLT efficiency MC Statistics
2nd vertex selection 0.2 HLT efficiency
0.25 Pile-up Re-weighting 2nd vertex selection
Luminosity Pile-up Re-weighting

0.2 0.15

0.15
0.1

0.1

0.05
0.05

0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3

Δφ(jet2,jet3) (rad) Δφ(jet2,jet3) (rad)

0.4 0.3
CMS Preliminary 2010 Total Uncertainty CMS Preliminary 2010 Total Uncertainty
Jet Energy Scale Jet Energy Scale
s=7TeV, L=36 pb-1 Jet Energy Resolution s=7TeV, L=36 pb-1 Jet Energy Resolution
0.35 pp→γ +3jet Photon ID Efficiency pp→γ +3jet Photon ID Efficiency
Photon Energy Scale 0.25
Relative uncertainty

Relative uncertainty

Photon Energy Scale

Photon Energy Resolution
Photon Energy Resolution
0.3 Model dependence
Model dependence
MC Statistics
HLT efficiency MC Statistics
2nd vertex selection 0.2 HLT efficiency
0.25 Pile-up Re-weighting 2nd vertex selection
Luminosity Pile-up Re-weighting

0.2 0.15

0.15
0.1

0.1

0.05
0.05

0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3

ΔrelPT(jet2,jet3) ΔrelPT(jet2,jet3)
0.5 0.5
Total Uncertainty Total Uncertainty
CMS Preliminary 2010 CMS Preliminary 2010
Jet Energy Scale Jet Energy Scale
0.45 s=7TeV, L=36 pb-1 Jet Energy Resolution 0.45 s=7TeV, L=36 pb-1 Jet Energy Resolution
pp→γ +3jet Photon ID Efficiency pp→γ +3jet Photon ID Efficiency
Photon Energy Scale
Photon Energy Scale
0.4 0.4
Relative uncertainty

Relative uncertainty

Photon Energy Resolution

Photon Energy Resolution
Model dependence
Model dependence
MC Statistics
0.35 HLT efficiency 0.35 MC Statistics
2nd vertex selection HLT efficiency
Pile-up Re-weighting 2nd vertex selection
0.3 Luminosity 0.3 Pile-up Re-weighting

0.25 0.25

0.2 0.2

0.15 0.15

0.1 0.1

0.05 0.05

0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3

ΔS (photonjet,dijet) (rad) ΔS (photonjet,dijet) (rad)

Fig. 8.3 Single contributions and total systematic uncertainty of Δφ23 (top row), Δrel pT,23 (center
row) and ΔS (bottom row) in absolute scale of real cross section (left column) and normalization
scale of shape (right column)
74 8 Systematic Uncertainties

Table 8.15 Total systematic uncertainties affecting the absolute cross section distributions for each
measured discriminating observable: the maximum, the minimum and the average value are listed
Measured observable Δφ23 (%) Δrel pT,23 (%) ΔS (%)
Max. 18.1 22.6 35.3
Min. 14.0 12.7 13.2
Avg. 15.8 17.6 25.4

Table 8.16 Total systematic uncertainties affecting the normalized distributions for each measured
discriminating observable: the maximum, the minimum and the average value are listed
Measured observable Δφ23 (%) Δrel pT,23 (%) ΔS (%)
Max. 11.9 13.1 24.7
Min. 4.6 4.4 5.0
Avg. 7.9 9.8 15.3

For 2010 data, the official uncertainty on the luminosity is around 4% and this
value is taken into account for this analysis and included in the final combination
of different systematic uncertainties. The combination has been evaluated by sum-
ming in quadrature the single contributions, assuming absence of correlation among
the different sources. The final results of total systematic uncertainties and sepa-
rate contribution are shown in Fig. 8.3 for different discriminating observables. In
Tables 8.15 and 8.16, the total systematic uncertainties are listed for each measured
discriminating observables in absolute cross section scale and normalization scale.
In absolute scale of real cross section, the value of total systematic uncertainties are
about 15–25%. Jet energy scale, jet energy resolution, photon energy scale, photon
efficiency correction, model dependence and Monte-Carlo statistics have relatively
large systematic uncertainties while the others are small or even almost negligible.
And the largest uncertainties are caused by jet energy scale. In normalization scale of
shape, the total systematic uncertainties are smaller than those in the absolute scale
by 5–10%. The values are around 8–15%. The uncertainties of Monte-Carlo statistics
remain the same and become relatively important in this scale. And the constant 4%
of luminosity uncertainty is not taken into account since its effect will disappear after
normalization. ΔS observable has relatively larger total systematic uncertainties than
the other observables by 8–10% in absolute scale of real cross section and 5–6% in
normalization scale of shape. Especially, large systematic uncertainties exist in the
DPS sensitive region of ΔS.
Chapter 9
Results

After the data unfolding, the measurement of the total integrated cross section and the
normalized differential cross section as a function of different DPS-discriminating
observables, fully corrected for detector effects and selection efficiency, are shown
in Table 9.1, Figs. 9.1 and 9.2. The cross sections for a final state with one photon and
at least three jets span a phase space with a photon and a jet of PT > 75 GeV together
with 2 jets of PT > 20 GeV, where photon is measured within η < 1.4442 or 1.566
<η < 2.5 and jets are within η < 2.5. The hadronic energy surrounding the photon
is required to be at most 5GeVwithin ΔR < 0.4 at the stable-particle level. This
isolation sum is calculated as the sum of the E T ofall charged and neutral particles,
after removing the photon, within a cone of ΔR≡ (η − ηγ )2 + (φ − φ γ )2 , ηγ and
φ γ being the coordinates of the photon. The 5GeVthreshold at the generator level
is chosen to ensure greater than 95% efficiency for direct photons and minimize
dependence of the efficiency on the variation of underlying event models. For each
i-th bin, the absolute and normalized differential cross sections are defined as

dσi Ni
= ; (9.1)
dβ L · Ci · Δβi

1 dσi 1 Ni
= · ; (9.2)
σ dβ Ntotal L · Ci · Δβi

where σ is the total integrated cross section, β is the DPS-discriminating observable,

Ntotal is the number of total events, L is the integrated luminosity, Ci is a correction
factor from the unfolding, and Δβi is the bin width.
The measured value of the total integrated cross section σ for a final state with
one photon and at least 3 jets is 124.9 ± 8.9 (stat.) ± 22.6 (syst.)pb. We compare
this with various theoretical predictions in Table 9.1. pythia 6 tune Z2 predicts a
lower value for the cross section than the measured value, while pythia 8 tune 4 C
is in agreement with it. There are two results of MadGraph by using different
© Springer Nature Singapore Pte Ltd. 2017 75
Y.-H. Chang, Study of Double Parton Scattering in Photon + 3 Jets Final State,
Springer Theses, DOI 10.1007/978-981-10-3824-2_9
76 9 Results

Table 9.1 Total integrated cross sections for measurement of the data and different MC predictions
in photon + 3 jets final state
Sample Tune Total Cross Section σ , pb
Data – 124.9 ± 8.9 (stat.)± 22.6 (syst.)
pythia 6 Z2 79.7 ± 3.2 (stat.)
pythia 8 4C 119.1 ± 0.7 (stat.)
MadGraph + pythia 6 Z2* 104.5 ± 1.8 (stat.)
Z2* (NO MPI) 101.6 ± 1.8 (stat.)
MadGraph + pythia 8 4C 95.2 ± 2.1 (stat.)
4C (NO MPI) 84.9 ± 2.0 (stat.)
sherpa CT10 190.2 ± 1.2 (stat.)
CT10 (NO MPI) 159.7 ± 1.3 (stat.)

pythia interfaces for hadronization, parton showering and also MPI simulation. The
MadGraph, interfaced with pythia 6 tune Z2*, gives a lightly lower value but still
in agreement with the measured cross section. The prediction of the MadGraph,
interfaced with pythia 8 tune 4C, is compatible with the measurement even if it
shows a lower value than the prediction of the MadGraph, interfaced with pythia 6
tune Z2*. sherpa tune CT10 overestimates the cross section value. Switching off
the MPI simulation for multijet-improved MC affect the prediction of cross section.
It reduces the prediction of cross section by 3% for the MadGraph, interfaced with
pythia 6 tune Z2*, and by 10% for the MadGraph, interfaced with pythia 8 tune
4C. The value of the cross section predicted by sherpa tune CT10 without MPI is
decreased by 16%, yet is still higher than the measured result.
The fractions of the different types of events, direct photon + 3 jets, fragmentation
photon + 3 jets and misidentified photon + 3 jets, contributing to each MC prediction
are shown in Table 9.2. Each fraction is calculated as the ratio of the integrated cross
section of the corresponding photon + 3 jets events to the total integrated cross section.
Different types of events are categorized according to the definitions described in
Chap. 5.
The main difference between pythia 6 and pythia 8 is the contribution of frag-
mentation photon + 3 jets events. MadGraph predicts a higher contribution of direct
photon + 3 jets events than pythia and sherpa by 40–60% and gives a lower con-
tribution of fragmentation photon + 3 jets events by 40–60%. sherpa shows similar
compositions of different photon + 3 jets events to pythia 6. The contributions of
misidentified photon + 3 jets events are similar (3–9%) among different MC simula-
tions. The theoretical fragmentation scale used in the calculation of MC simulation
will affect the consideration of a photon as a fragmentation photon or a higher-order
photon. Implementing different fragmentation scales then changes the SPS contri-
bution of fragmentation photon + 3 jets events.
9 Results 77

Table 9.2 Fractions of different types of events based on the MC estimation. Three kinds of
contributions are included in this table: direct photon + 3 jets events, fragmentation photon (frag.
photon) + 3 jets events, and misidentified photon (fake photon) + 3 jets events
Sample Tune Direct photon + 3 Frag. photon (%) Fake photon + 3
jets (%) + 3 jets jets (%)
pythia 6 Z2 42.6 48.2 9.2
pythia 8 4C 28.6 66.1 5.3
MadGraph + Z2* 92.1 3.5 4.4
pythia 6
Z2* (NO MPI) 85.0 9.1 5.9
MadGraph + 4C 83.8 13.3 2.9
pythia 8
4C (NO MPI) 83.6 14.0 2.4
sherpa CT10 52.0 41.9 6.0
CT10 (NO MPI) 49.2 45.5 5.3

The absolute real cross section as a function of Δφ23 is shown in Fig. 9.1 (left
column) and is well described by pythia 8 and MadGraph, with some slight fluctu-
ations. The DPS contribution is visible at Δφ ∼ π , as illustrated by the predictions of
MadGraph and sherpa without MPI. pythia 6 has lower prediction than pythia 8,
MadGraph. And sherpa shows significantly overestimated prediction w.r.t. all the
other MC predictions. In Fig. 9.1 (center column), the absolute real cross section as
a function of Δrel pT,23 is shown to be in agreement with pythia 8 and MadGraph
for Δrel pT,23 < 0.5 but visible differences arise in the high value region. pythia 6
has lower prediction than pythia 8, MadGraph. And sherpa shows significantly
overestimated prediction w.r.t. all the other MC predictions. The predictions of Mad-
Graph and sherpa without MPI exhibit the need of the missing DPS contribution
in this region.
Figure 9.2 (right column) shows the absolute real cross section as a function of
ΔS. pythia 8 and MadGraph give compatible predictions to the measurement.
pythia 6 still gives lower prediction than pythia 8, MadGraph in the whole region.
And sherpa still shows significantly overestimated prediction w.r.t. all the other MC
predictions in the whole region.
The normalized differential cross section as a function of Δφ23 is shown in Fig. 9.2
(left column) and is well described by all MC predictions, with some slight fluctua-
tions. The DPS contribution shows very little sensitivity at Δφ ∼ π , as illustrated by
the predictions of MadGraph and sherpa without MPI. In Fig. 9.2 (center column),
the normalized differential cross section as a function of Δrel pT,23 is shown to be in
agreement with all MC prediction for Δrel pT,23 > 0.4 but visible differences arise
78 9 Results

CMS Preliminary 2010, s = 7 TeV, L = 36.14 pb-1, pp → γ +3jets CMS Preliminary 2010, s = 7 TeV, L = 36.14 pb-1, pp → γ +3jets CMS Preliminary 2010, s = 7 TeV, L = 36.14 pb-1, pp → γ +3jets
σ [pb] 102

σ [pb]

σ [pb]
γ -jet: p > 75 GeV DATA γ -jet: p > 75 GeV DATA γ -jet: p > 75 GeV DATA
T T T
2jets: p > 20 GeV 2jets: p > 20 GeV 2jets: p > 20 GeV
T T T
γ : |η|<1.4442, 1.566<|η|< 2.5 Pythia8 4C γ : |η|<1.4442, 1.566<|η|< 2.5 Pythia8 4C 103 γ : |η|<1.4442, 1.566<|η|< 2.5 Pythia8 4C
jets: |η|<2.4 jets: |η|<2.4 jets: |η|<2.4
Pythia6 Z2 Pythia6 Z2 Pythia6 Z2
102
Stat. + syst. uncertainty Stat. + syst. uncertainty Stat. + syst. uncertainty
102

10
10

10

1
2 2.5 3.5
MC/DATA

MC/DATA

MC/DATA
1.8
1.6
1.4
2 ΔrelPT(jet2,jet3) 3
2.5
1.2 1.5 2
1
0.8 1 1.5
0.6 1
0.4 0.5
0.2 0.5
0 0 0
0 0.5 1 1.5 2 2.5 3 0 0.2 0.4 0.6 0.8 1 0 0.5 1 1.5 2 2.5 3
Δφ(jet2,jet3) (rad) ΔrelPT(jet2,jet3) ΔS (photonjet,dijet) (rad)
3
CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j 3
CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j
10 10
σ dΔφ [1/rad]

σ dΔS [1/rad]
T,23
γ -j: p ,p > 75GeV DATA γ -j: p ,p > 75GeV DATA 104 γ -j: p ,p > 75GeV DATA
Tγ T jet1 Tγ T jet1 Tγ T jet1
MADGRAPH+PY6 Z2* σ dΔrelp MADGRAPH+PY6 Z2* MADGRAPH+PY6 Z2*
1 dσ
2j: p ,p > 20GeV MADGRAPH+PY6 Z2* no MPI 2j: p ,p > 20GeV MADGRAPH+PY6 Z2* no MPI 2j: p ,p > 20GeV MADGRAPH+PY6 Z2* no MPI
T jet2 T jet3 T jet2 T jet3 T jet2 T jet3
γ : |η|<1.4442, 1.566<|η|<2.5 MADGRAPH+PY8 4C γ : |η|<1.4442, 1.566<|η|<2.5 MADGRAPH+PY8 4C γ : |η|<1.4442, 1.566<|η|<2.5 MADGRAPH+PY8 4C
23

1 dσ
MADGRAPH+PY8 4C no MPI MADGRAPH+PY8 4C no MPI 3 MADGRAPH+PY8 4C no MPI
1 dσ

jets: |η|<2.4 jets: |η|<2.4 10 jets: |η|<2.4

SHERPA CT10 SHERPA CT10 SHERPA CT10
2 2
10 SHERPA CT10 no MPI 10 SHERPA CT10 no MPI SHERPA CT10 no MPI
Total Uncertainty Total Uncertainty Total Uncertainty
102

10
10 10

1 1 10-1
2.5 2.5 3
MC / DATA

MC / DATA

MC / DATA
MADGRAPH+PY6 Z2* MADGRAPH+PY6 Z2* 2.5
MADGRAPH+PY6 Z2*
2 2
2
1.5 1.5
1.5
1 1
1
0.5 0.5 0.5
2.5 2.5 3
MC / DATA

MC / DATA

MC / DATA
MADGRAPH+PY8 4C MADGRAPH+PY8 4C
T,23 2.5
MADGRAPH+PY8 4C
2 2
2
1.5 1.5
1.5
1 1
1
0.5 0.5 0.5
2.5 3
MC / DATA

MC / DATA

SHERPA CT10 SHERPA CT10

T,23
MC / DATA 2.5
SHERPA CT10
2 2
2
1.5 1.5
1.5
1 1
1
0.5 0.5 0.5
0 0.5 1 1.5 2 2.5 3 0 0.2 0.4 0.6 0.8 1 0 0.5 1 1.5 2 2.5 3
Δφ (rad)
23
Δ relpT,23 Δ S (rad)

Fig. 9.1 Absolute real cross sections as a function of Δφ23 (left), Δrel pT,23 (center), and ΔS
(right) compared to different MC predictions. The comparisons are shown for two MC classes:
comparisons between the data and LO MC predictions, pythia 6 and pythia 8, (upper row),
comparisons between the data and multijet-improved MC predictions, MadGraph interfaced with
pythia 6 or pythia 8, and sherpa, (lower row). The lower panels show the ratio of the MC
prediction to the data. For the MC predictions, three kinds of contributions are included: direct
photon + 3 jets events, fragmentation photon + 3 jets events and misidentified photon + 3 jets
events. The band represents the total uncertainty of the data

in the low value region. For Δrel pT,23 < 0.2, pythia 6 has lower prediction than
pythia 8, MadGraph and sherpa. The predictions of MadGraph and sherpa
without MPI exhibit the need of the missing DPS contribution in this region.
Figure 9.2 (right column) shows the normalized differential cross section as a
function of ΔS. In the region of ΔS < 2.5, pythia 6 shows insufficient activity,
while the predictions of pythia 8, MadGraph and sherpa are compatible with the
measured distribution. Different pythia interfaces do not affect the predictions of
MadGraph. sherpa is slightly higher than the other predictions. Switching off the
MPI simulation for MadGraph and sherpa causes about 5–10% differences which
9 Results 79

CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j
1
γ -j: pT γ , pT jet1 > 75GeV DATA γ -j: pT γ , pT jet1 > 75GeV DATA γ -j: pT γ , pT jet1 > 75GeV DATA
2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV
PYTHIA8 4C PYTHIA8 4C 10 PYTHIA8 4C
γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
1
jets: | η|<2.4 jets: | η|<2.4 jets: | η|<2.4
σ dΔφ [1/rad]

PYTHIA6 Z2 PYTHIA6 Z2 PYTHIA6 Z2

σ dΔS [1/rad]
T,23
Total Uncertainty Total Uncertainty Total Uncertainty
1

σ dΔrelp
1 dσ
23
1 dσ

1 dσ
-1
10
10-1
10-1

10-2

20
10-22
0.5 1 1.5 2 2.5 3 0 0.2 0.4 0.6 0.8 1 20 0.5 1 1.5 2 2.5 3
MC / DATA

MC / DATA

MC / DATA
1.8
1.6
Δφ (rad) 1.8
1.6 Δ p
rel 1.8
1.6 ΔS (rad)
23 T,23 1.4
1.4 1.4 1.2
1.2 1.2 1
1 1 0.8
0.8 0.8 0.6
0.4
0.6 0.6 0.2
0.4 0.4 0
0 0.5 1 1.5 2 2.5 3 0 0.2 0.4 0.6 0.8 1 0 0.5 1 1.5 2 2.5 3
Δφ23 (rad) ΔrelpT,23 ΔS (rad)

CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j
1
γ -j: p ,p > 75GeV DATA γ -j: p ,p > 75GeV DATA γ -j: p ,p > 75GeV DATA
Tγ T jet1 Tγ T jet1 Tγ T jet1
MADGRAPH+PY6 Z2* MADGRAPH+PY6 Z2* MADGRAPH+PY6 Z2*
2j: p ,p > 20GeV MADGRAPH+PY6 Z2* no MPI 2j: p ,p > 20GeV MADGRAPH+PY6 Z2* no MPI 2j: p ,p > 20GeV MADGRAPH+PY6 Z2* no MPI
T jet2 T jet3 T jet2 T jet3 T jet2 T jet3
MADGRAPH+PY8 4C MADGRAPH+PY8 4C 10 MADGRAPH+PY8 4C
γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
MADGRAPH+PY8 4C no MPI 1 MADGRAPH+PY8 4C no MPI MADGRAPH+PY8 4C no MPI
jets: |η|<2.4 SHERPA CT10 jets: |η|<2.4 SHERPA CT10 jets: |η|<2.4 SHERPA CT10
σ dΔφ [1/rad]

σ dΔS [1/rad]
T,23

SHERPA CT10 no MPI SHERPA CT10 no MPI SHERPA CT10 no MPI

σ dΔrelp
1 dσ

Total Uncertainty Total Uncertainty Total Uncertainty

1
23

1 dσ
1 dσ

10-1
10-1
10-1

10-2

10-2
2 2 2
MC / DATA

MC / DATA

MC / DATA
1.8 MADGRAPH+PY6 Z2* 1.8 MADGRAPH+PY6 Z2* 1.8 MADGRAPH+PY6 Z2*
1.6 1.6 1.6
1.4 1.4 1.4
1.2 1.2 1.2
1 1 1
0.8 0.8 0.8
0.6 0.6 0.6
0.4 0.4 0.4
20 0.5 1 1.5 2 2.5 3 20 0.2 0.4 0.6 0.8 rel 1 20 0.5 1 1.5 2 2.5 3
Δ p4C
MC / DATA

MC / DATA

MC / DATA
1.8 Δφ (rad)
MADGRAPH+PY8 4C 1.8 MADGRAPH+PY8 1.8 Δ S (rad)
MADGRAPH+PY8 4C
1.6 23 1.6 T,23 1.6
1.4 1.4 1.4
1.2 1.2 1.2
1 1 1
0.8 0.8 0.8
0.6 0.6 0.6
0.4 0.4 0.4
20 0.5 1 1.5 2 2.5 3 20 0.2 0.4 0.6 0.8 rel 1 20 0.5 1 1.5 2 2.5 3
SHERPAΔCT10
MC / DATA

MC / DATA

MC / DATA
1.8 Δφ CT10
SHERPA (rad) 1.8 pT,23 1.8 Δ SCT10
SHERPA (rad)
1.6 23 1.6 1.6
1.4 1.4 1.4
1.2 1.2 1.2
1 1 1
0.8 0.8 0.8
0.6 0.6 0.6
0.4 0.4 0.4
0 0.5 1 1.5 2 2.5 3 0 0.2 0.4 0.6 0.8 1 0 0.5 1 1.5 2 2.5 3
Δφ (rad)
23
Δ relpT,23 Δ S (rad)

Fig. 9.2 Normalized differential cross sections as a function of Δφ23 (left), Δrel pT,23 (center),
and ΔS (right) compared to different MC predictions. The comparisons are shown for two MC
classes: comparisons between the data and LO MC predictions, pythia 6 and pythia 8, (upper
row), comparisons between the data and multijet-improved MC predictions, MadGraph interfaced
with pythia 6 or pythia 8, and sherpa, (lower row). The lower panels show the ratio of the
MC prediction to the data. For the MC predictions, three kinds of contributions are included: direct
photon + 3 jets events, fragmentation photon + 3 jets events and misidentified photon + 3 jets events.
The band represents the total uncertainty of the data

is lower than the results observed in Reference [23, 24]. The reason the MPI causes
only small changes in ΔS is due to various background components as discussed in
Chap. 5. They reduce the possible room for the DPS signal contribution.
Chapter 10
Conclusion

Measurements of observables for the production of one photon and at least three
jets have been performed based on the data collected with the CMS experiment in
2010 with an integrated luminosity of 36 pb−1 . The cross section is measured to be
σ = 124.9 ± 8.9 (stat.) ± 22.6 (syst.) pb for a final state with a photon and a jet of pT
> 75 GeV together with 2 jets of pT > 20 GeV, where the photon is found within η <
1.4442 or 1.566 <η < 2.5 and the jets lie within η < 2.4. The normalized differential
cross sections, as a function of Δφ23 , Δrel pT,23 , and ΔS, are compared with various
theoretical predictions: pythia 6 tune Z2, pythia 8 tune 4C, MadGraph interfaced
with pythia 6 tune Z2*, MadGraph interfaced with pythia 8 tune 4C, and sherpa
tune CT10.
The DPS signal and various background components are studied with the help of
the Δφ23 , Δrel pT,23 , and ΔS observables. The DPS signal distribution shows different
features with respect to each background distribution and is distinguishable in the
DPS-sensitive region, at high Δφ23 values, low Δrel pT,23 values, and low to medium
ΔS values. Model dependence can be observed from comparisons of background
distributions produced by different MC simulations. It indicates that higher order
calculations including merging between matrix elements and parton showers affect
the contributions from SPS events and misidentified photon events. Implementing
different fragmentation scales in the calculation of MC simulation changes the SPS
contribution of fragmentation photon + 3 jets events.
From the comparison of the normalized differential cross sections as a function of
Δφ23 , Δrel pT,23 , and ΔS, the predictions of pythia 8, MadGraph interfaced with
pythia 6 or pythia 8, and sherpa with MPI describe the measured distributions
better than pythia 6 within the uncertainties except for some fluctuations in a few
bins. The MC predictions with and without a simulation of MPI do not differ sig-
nificantly even in the DPS-sensitive region due to various background components
which make the DPS signal contribution less pronounced. Hence, not only must
the predictions including MPI be validated with underlying event measurements, but

© Springer Nature Singapore Pte Ltd. 2017 81

Y.-H. Chang, Study of Double Parton Scattering in Photon + 3 Jets Final State,
Springer Theses, DOI 10.1007/978-981-10-3824-2_10
82 10 Conclusion

also the models used to extrapolate the hardest interaction need to be well understood
before the direct extraction of the underlying DPS fraction.
To sum up, the data are reasonably well described by MC predictions with or
without MPI and any conclusion on a DPS component is not possible within the
given precision of the current data and simulations. And this analysis demonstrates
the importance of understanding the background contribution which is crucial for
the extraction of the DPS fraction in proton-proton collisions at LHC energies.
Appendix A
K-Factor Correction for the LO
Cross-Section of P YTHIA, M AD G RAPH
and S HERPA MCs

In order to get the k-factor of isolated prompt photon production (direct photon +
fragmentation photon) used in this analysis, we compare the LO Pythia, MadGraph
and Sherpa predictions with the NLO pQCD predictions from JETPHOX
1.3.0 [68, 85, 86] using CT10 PDFs [65]. The k-factors of Pythia 6, Pythia 8,
MadGraph and Sherpa are calculated by the following formula with ET and η
dependences:

NLO(ET , η) of JETPHOX
K factor(ET , η) = (A.1)
LO(ET , η) of Monte − Carlo

For the photon-jet samples which contains direct photon production, the selected
photons have to be a stable particle and have at least one mother particle. Its mother
particle should be a photon, which is corresponding to the definition of direct photon
production. Besides, the selected photons are requested to have generator-level iso-
lation less than 5 GeV. The pseudorapidity range of selected photons is from −2.5
to 2.5 rad. For the QCD di-jet samples which contain fragmentation photon produc-
tion, the selected photons have to be a stable particle and have at least one mother
particle. Its mother particle should be a quark or gluon, which is corresponding to
the definition of fragmentation photon production. Besides, the selected photons
are requested to have generator-level isolation less than 5 GeV. The pseudorapidity
range of selected photons is from −2.5 to 2.5 rad. The isolation sum is calculated
as the sum of the ET ofall charged and neutral particles, after removing the photon,
within a cone of ΔR≡ (η − ηγ )2 + (φ − φ γ )2 , ηγ and φ γ being the coordinates of
the photon. The 5 GeV threshold at the generator level was chosen to ensure greater
than 95% efficiency for direct photons and minimise dependence of the efficiency
on the variation of underlying event models.
Figure A.1, Tables A.1 and A.2 show the details of k-factor values for Pythia 6,
Pythia 8 and MadGraph.
In order to take the assumption that the k-factors for photon + 1 jet and photon + 3
jets should be similar based upon Z + 1 jet and Z + 3 jets, we study the BLACKHAT’s
predictions of Z + N-Jet(s) events ratio [87, 88] with MSTW2008 NLO and LO PDF

© Springer Nature Singapore Pte Ltd. 2017 83

Y.-H. Chang, Study of Double Parton Scattering in Photon + 3 Jets Final State,
Springer Theses, DOI 10.1007/978-981-10-3824-2
84 Appendix A: K-Factor Correction for the LO Cross-Section …

104
|ηγ |<0.9 JETPHOX 0.9<|ηγ |<1.4442 JETPHOX

104
PYTHIA-6 PYTHIA-6
103
MADGRAPH+PYTHIA-6 MADGRAPH+PYTHIA-6
3
10
PYTHIA-8 PYTHIA-8

102

102

σ (pb)

σ (pb)
10

10

1
1

2 2
JETPHOX

JETPHOX
1.8 1.8
1.6 1.6
MC

MC
1.4 1.4
1.2 1.2
1 1
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
50 100 150 200 250 300 350 400 50 100 150 200 250 300 350 400
ET (photon) (GeV) ET (photon) (GeV)
104 104
1.566<|ηγ |<2.1 JETPHOX
2.2<|ηγ |<2.5 JETPHOX

PYTHIA-6 PYTHIA-6
3 103
10
MADGRAPH+PYTHIA-6 MADGRAPH+PYTHIA-6

PYTHIA-8 102 PYTHIA-8

102

10
σ (pb)

σ (pb)
10

1
10-1

2 2
JETPHOX
JETPHOX

1.8 1.8
1.6 1.6
MC

MC

1.4 1.4
1.2 1.2
1 1
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
50 100 150 200 250 300 350 400 50 100 150 200 250 300 350 400
ET (photon) (GeV) ET (photon) (GeV)

Fig. A.1 K-factors of NLO (JETPHOX) to LO (Monte-Carlo) ratios with respect to different ET
and η regions. a |η| < 0.9 b 0.9 < |η| < 1.4442 c 1.566 < |η| < 2.1 d 2.1 < |η| < 2.5

Table A.1 The k-factors of Pythia 6(PY-6) and Pythia 8(PY-8) are shown with ET and η
dependences
|η| < 0.9 0.9 < |η| < 1.4442 1.566 < |η| < 2.1 2.1 < |η| < 2.5
ET bin PY-6 PY-8 PY-6 PY-8 PY-6 PY-8 PY-6 PY-8
( GeV)
70−80 1.267 0.916 1.233 0.922 1.174 0.888 1.108 0.832
80−100 1.235 0.961 1.225 0.953 1.151 0.881 1.137 0.835
100−120 1.265 1.016 1.229 0.947 1.163 0.947 1.170 0.952
120−200 1.310 1.025 1.261 0.986 1.216 0.967 1.160 0.936
200−300 1.337 1.073 1.298 1.032 1.210 1.014 1.090 0.957
300−400 1.342 1.140 1.281 1.081 1.207 1.010 1.111 0.976

Table A.2 The k-factors of Sherpa, MadGraph, showered and hadronized by

Pythia 6,(MG+Py6) and MadGraph, showered and hadronized by Pythia 8,(MG+Py8)
are shown with η dependences
η bin MG+Py6 MG+Py8 Sherpa
|η| < 0.9 1.18 1.46 1.03
0.9 < |η| < 1.4442 1.17 1.44 1.04
1.566 < |η| < 2.1 1.15 1.40 1.07
2.1 < |η| < 2.5 1.20 1.48 1.08
Appendix A: K-Factor Correction for the LO Cross-section … 85

Table A.3 Summary of jet Selection criteria for lepton

and lepton selection criteria
for Z + N-Jet(s) events pT > 20 GeV
|η| < 2.1
76 < M(di-lepton) < 106
Selection criteria for jet
Anti-Kt algorithm with cone size 0.5
pT > 30 GeV
|η| < 2.4
ΔR(lepton, jet)>0.5

3
10
10
LO of Z+1jets, lo LO of Z+1jets, lo
1 102

-1
10

Events / 0.1 rad

Events / 20 GeV

NLO of Z+1jets, nlo 10 NLO of Z+1jets, nlo

10-2

-3 1
10

10-4 10-1
-5
10

-6
10-2
10

10-7 -3
10 4-3
4 -2 -1 0 1 2 3
NLO_nlo

NLO_nlo

η (Z) (rad)
LO_lo

LO_lo

3.5 3.5
3 3
2.5 2.5
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 200 400 600 800 1000 1200 1400 -3 -2 -1 0 1 2 3
PT (Z) (GeV) η (Z) (rad)
3
10
10
LO of Z+1jets, lo LO of Z+1jets, lo
1 102

10-1
Events / 0.1 rad
Events / 20 GeV

NLO of Z+1jets, nlo 10 NLO of Z+1jets, nlo

10-2
-3
10 1
10-4
-5 10-1
10
-6
10
10-2
10-7
-8 -3
1010 10 4-3 -2 -1 0 1 2 3
NLO_nlo

NLO_nlo

9 η (jet1) (rad)
LO_lo

LO_lo

3.5
8
7 3
6 2.5
5 2
4 1.5
3 1
2
1 0.5
0 0
0 200 400 600 800 1000 1200 1400 -3 -2 -1 0 1 2 3
PT (jet1) (GeV) η (jet1) (rad)

Fig. A.2 LO and NLO distribution of pT and η of Z boson and leading jet for Z + 1 jet events

sets [89]. Table A.3 shows the selection criteria for Z + N-Jet(s) event. The related
results are shown from Figs. A.2, A.3 and A.4. Basically, an additional correction
factor of 0.73 for photon + 3 jets events is included through studying the BLACK-
HAT’s predictions of Z + N-Jet(s) events ratio. This additional factor is applied to
MadGraph samples due to that they contain 2 → 5 processes (γ + 4 jets).
The associated systematic uncertainty is studied by varying the factorization and
renormalization scale of Z + N-Jet(s) events. The original and recommended scale
is 0.5 Ht. By varying both of the the factorization and renormalization scale to 0.25
Ht and Ht, around 21% uncertainty is estimated and assigned for all MadGraph
samples in this analysis.
86 Appendix A: K-Factor Correction for the LO Cross-Section …
3
1 10

LO of Z+3jets, lo 102 LO of Z+3jets, lo

10-1

Events / 0.1 rad

Events / 20 GeV
10
10-2
NLO of Z+3jets, nlo NLO of Z+3jets, nlo
-3 1
10

10-1
10-4

-5
10-2
10
-3
10
-6
10
10-4
10-7
-5
4
10 4
NLO_nlo

NLO_nlo
LO_lo

LO_lo
3.5 3.5
3 3
2.5 2.5
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 200 400 600 800 1000 1200 1400 -3 -2 -1 0 1 2 3
PT (Z) (GeV) η (Z) (rad)
3
10
1 LO of Z+3jets, lo LO of Z+3jets, lo
102

10-1

Events / 0.1 rad

Events / 20 GeV

10
NLO of Z+3jets, nlo NLO of Z+3jets, nlo
10-2
1
-3
10 10-1

10-4 10-2
-5 -3
10 10

-6
10 10-4

-5
10-74 10 4
NLO_nlo

NLO_nlo
LO_lo

LO_lo

3.5 3.5
3 3
2.5 2.5
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 200 400 600 800 1000 1200 1400 -3 -2 -1 0 1 2 3
PT (jet1) (GeV) η (jet1) (rad)

Fig. A.3 LO and NLO distribution of pT and η of Z boson and first leading jet for Z + 3 jets events
4.5 4.5

4 NLO/LO, Z+3jets 4 NLO/LO, Z+3jets

3.5 3.5

3 NLO/LO, Z+1jet 3 NLO/LO, Z+1jet

2.5 2.5
NLO_nlo

NLO_nlo
LO_lo

LO_lo

2 2

1.5 1.5

1 1

0.5 0.5

0 0
1.4 1.4
K-factor Z3j
K-factor Z1j

K-factor Z3j
K-factor Z1j

1.2 1.2
1 1
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 200 400 600 800 1000 1200 1400 -3 -2 -1 0 1 2 3
PT (Z) (GeV) η (Z) (GeV)
10 4.5

9 NLO/LO, Z+3jets 4 NLO/LO, Z+3jets

8 3.5
7 NLO/LO, Z+1jet NLO/LO, Z+1jet
3
NLO_nlo

NLO_nlo
LO_lo

LO_lo

6
2.5
5
2
4
1.5
3
1
2

1 0.5

0 0
1.4 1.4
K-factor Z3j
K-factor Z1j

K-factor Z3j
K-factor Z1j

1.2 1.2
1 1
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 200 400 600 800 1000 1200 1400 -3 -2 -1 0 1 2 3
PT (jet1) (GeV) η (jet1) (GeV)

Fig. A.4 k-factor comparisons between Z + 3 jets event and Z + 1 jet event in a pT spectra of
Z boson, b η distribution of Z boson, c pT spectra of first leading jet and d η distribution of first
leading jet
Appendix B
Study of Low Acceptance of Photon + 3 Jets
Events

In order to figure out why the acceptance of photon + 3 jets events is so low in this
analysis, we try to investigate the effect of different selection cuts on the accep-
tance. A comparison between original study and a special study by removing the
photon identification selection, described in Sect. 4.3, except for cuts of pT and η
region. As already mentioned before, photon isolation cuts are introduced to reduce
the background rate, caused by any possible fake photons or photons come from
the decay of π 0 and η. In this analysis, high photon purity is necessary due to the
multiple background can make the observation of DPS hard. It’s still worth to use
such selection criteria even we will loss much events which maybe are actually DPS
events. On the other hand, a selection cut on the 2nd vertex described before will
make the analysis focus on those events with only one vertex or events whose pile-
up is relatively soft. It’s side effect will make the acceptance of final photon + 3 jets
events becomes quite low. But this is unavoidable if we want to reduce the influence
of SPS events. In this appendix, we only perform the side effect caused by the photon
selection criteria.
It should be noted that we only studied difference of the original event yield
after selection by combining different processes of MCs together. That is, only the
weighting factors for combining MCs would be introduced to scale the distributions.
The results can be found in Fig. B.1. The acceptances would be much higher if we
remove the photon identification selection from our original selection criteria. But in
the same time, the background rate also increased and would be even much higher
than acceptance. This results are just consistent with our understanding described
before.

© Springer Nature Singapore Pte Ltd. 2017 87

Y.-H. Chang, Study of Double Parton Scattering in Photon + 3 Jets Final State,
Springer Theses, DOI 10.1007/978-981-10-3824-2
88 Appendix B: Study of Low Acceptance of Photon + 3 Jets Events

CMS Preliminary 2010 CMS Preliminary 2010

s=7TeV, L=36 pb-1 Acceptance s=7TeV, L=36 pb-1 Acceptance
1.4 pp→γ +3jet
Purity 1.4 pp→γ +3jet
Purity
Stability Stability
Background Background
1.2 1.2
Events / 0.315 rad

Events / 0.101
1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.5 1 1.5 2 2.5 3 0 0.2 0.4 0.6 0.8 1
Δφ(jet2,jet3) (rad) ΔrelPT(Jet2,jet3)

CMS Preliminary 2010

s=7TeV, L=36 pb-1 Acceptance
1.4 pp→γ +3jet
Purity
Stability
Background
1.2
Events / 0.315 (rad)

0.8

0.6

0.4

0.2

0
0 0.5 1 1.5 2 2.5 3
Δ
S (photonjet,dijet) (rad)

Fig. B.1 Study of low acceptance: acceptance, purity, stability and background for a Δφ(jet2,jet3)
and b Δrel pT(jet2,jet3) and c ΔS
Appendix C
More DPS Discriminating Observables

To further differentiate between the SPS and DPS processes, more discriminating
observables with effective distinguishing power are used to the independence and
pairwise momentum balance of the two hard scatterings in DPS events. Basically,
all of observables included in this thesis are designed based on the study described
in [16, 17]. As already mentioned before, a final state arising from a single chain tends
to have a strongly correlated configuration in the azimuthal angle and pT -balance
distribution between the two sub-systems, while a DPS event has a preferred back-
to-back topology for the separated systems that are not correlated. The additional
four discriminating observables are listed below. Relevant detector-level results are
also included in this appendix.
• The difference in azimuthal angle between the photon and jet belonging to the
photon-jet pair,
Δφ(γ ,jet1) = φγ − φjet1 ; (C.1)

• a combination of the difference in azimuthal angle between the photon and jets of
the whole photon + 3 jets system,

1 
Sφ = √ (Δφγ ,jet1 )2 + (Δφjet2,jet3 )2 ; (C.2)
2π 2

Different configurations for the photon and 3 jets in the final state translate in
different region for angular variables. Basically, these variables have good dis-
tinguishing power. Sφ accounts for the difference in azimuthal angle between the
selected photon and jets. Δφ(γ ,jet1) computes the difference of azimuthal angle in
photon-jet pair. SPS events lead to a broad distribution for these quantities, while
DPS events contribute most at values close to π , meaning the correlation between
the photon-jet pair and di-jet pair is very low. It should be noted that the variable
Sφ have been normalized by a factor of √ 1 2 .
2π
• the transverse momentum balance of the photon and jet belonging to the photon-jet
pair,

© Springer Nature Singapore Pte Ltd. 2017 89

Y.-H. Chang, Study of Double Parton Scattering in Photon + 3 Jets Final State,
Springer Theses, DOI 10.1007/978-981-10-3824-2
90 Appendix C: More DPS Discriminating Observables

|pT γ + pT jet1 |
Δrel pT(γ ,jet1) = ; (C.3)
|pT γ | + |pT jet1 |

Δrel pT(γ ,jet1) is sensitive to the level of momentum balance in photon-jet pair.
A back-to-back for the two separate sub-systems contributes at low value of
Δrel pT(γ ,jet1) , while correlated pairs of jets result in a broader distribution over
the whole phase space. The balance level of transverse momentum is not only
analyzed in separate sub-systems but also in a compound scale.
• compound information of transverse momenta from all selected photon and 3 jets,
 2  2
1 pT γ + pT jet1 pT jet2 + pT jet3
SpT =√ + ; (C.4)
2 |pT γ | + |pT jet1 | |pT jet2 | + |pT jet3 |

The distribution of SpT is similar to the one expected for the single variables
Δrel pT(γ ,jet1) and Δrel pT(jet2,jet3) for considering the transverse momentum balance
of correlated and uncorrelated systems but it gives a better overview of the process,
by taking into account information of the whole photon + 3 jets final state. The
better pT -balance the system have, the smaller value SpT will be. It should be noted
the variable SpT have been normalized by a factor of √12 .
Figures C.1, C.2, C.3 and C.4 represent the measurement of discriminating observ-
ables Δrel pT(γ ,jet1) and Δrel pT(jet2,jet3) , respectively, for both event distribution and
normalized shape comparison between data and different MC simulation. It can
be found the results are quite consistent to those shown in Sect. 6.1. In real scale,
MadGraph samples predict data quite well within statistic uncertainty. Pythia-
6 and Pythia-8 samples predicts a slightly smaller total number of events about
20–30% for both Pythia-6 and Pythia-8 but with larger statistic uncertainties,
especially for Pythia-8. There is an overall good agreement of shapes of all the
distribution for uncorrected data and different MCs in normalization scale.

CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j+X CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j+X
105 γ -j: pT γ , pT jet1 > 75GeV γ -j: pT γ , pT jet1 > 75GeV
DATA DATA
2j: pT jet2, pT jet3 > 20GeV 10 2j: pT jet2, pT jet3 > 20GeV
PYTHIA8 4C PYTHIA8 4C
4 γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
10
Events / 0.315 rad

jets: | η|<2.4 PYTHIA6 Z2 jets: | η|<2.4 PYTHIA6 Z2

MADGRAPH+PY6 Z2 1 MADGRAPH+PY6 Z2
103
dN/N

102
10-1

10

10-2
1

4 5
MC / DATA

3.5 4.5
MC / DATA

3 4
3.5
2.5 3
2 2.5
1.5 2
1 1.5
1
0.5 0.5
0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
Δφ(γ jet1) (rad) Δφ(γ jet1) (rad)

Fig. C.1 The detector level comparison of Δφ(γ ,jet1) between uncorrected data and different com-
bined MC samples. a Event distribution. b Shape comparison
Appendix C: More DPS Discriminating Observables 91

CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j+X CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j+X
γ -j: pT γ , pT jet1 > 75GeV DATA γ -j: pT γ , pT jet1 > 75GeV DATA
2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV
PYTHIA8 4C 1 PYTHIA8 4C
γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
103 jets: | η|<2.4 PYTHIA6 Z2 jets: | η|<2.4 PYTHIA6 Z2

Events / 0.101
MADGRAPH+PY6 Z2 MADGRAPH+PY6 Z2

dN/N
10-1

102

10-2

10
3.5 1 3.5

MC / DATA
3 3
MC / DATA

2.5 2.5
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
ΔPT(γ jet1) ΔPT(γ jet1)

Fig. C.2 The detector level comparison of Δrel pT(γ ,jet1) between uncorrected data and different
combined MC samples. a Event distribution. b Shape comparison

CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j+X CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j+X
102
γ -j: pT γ , pT jet1 > 75GeV DATA γ -j: pT γ , pT jet1 > 75GeV DATA
105
2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV
PYTHIA8 4C 10 PYTHIA8 4C
γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
4
10
jets: | η|<2.4 PYTHIA6 Z2 jets: | η|<2.4 PYTHIA6 Z2
Events / 0.101

1
103
MADGRAPH+PY6 Z2 MADGRAPH+PY6 Z2
dN/N

10-1
102

10-2
10

1 10-3

10-1
10-4
3 4
3.5
MC / DATA

2.5
MC / DATA

3
2 2.5
1.5 2
1 1.5
1
0.5 0.5
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Sφ Sφ

Fig. C.3 The detector level comparison of Sφ between uncorrected data and different combined
MC samples. a Event distribution. b Shape comparison

CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j+X CMS Preliminary, s = 7 TeV, L = 36 pb-1, pp → γ +3j+X
104
γ -j: pT γ , pT jet1 > 75GeV DATA γ -j: pT γ , pT jet1 > 75GeV DATA
2j: pT jet2, pT jet3 > 20GeV 2j: pT jet2, pT jet3 > 20GeV
PYTHIA8 4C 1 PYTHIA8 4C
γ : |η|<1.4442, 1.566<|η|<2.5 γ : |η|<1.4442, 1.566<|η|<2.5
PYTHIA6 Z2 PYTHIA6 Z2
Events / 0.101

jets: | η|<2.4 jets: | η|<2.4

103
MADGRAPH+PY6 Z2 MADGRAPH+PY6 Z2
dN/N

10-1

102

10-2
10
2 2.5
MC / DATA

MC / DATA

1.8
1.6 2
1.4
1.2 1.5
1
0.8 1
0.6
0.4 0.5
0.2
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
SPT SPT

Fig. C.4 The detector level comparison of SpT between uncorrected data and different combined
MC samples. a Event distribution. b Shape comparison
Curriculum Vitae

Dr. You-Hao Chang

Personal Information
Date of Birth: 02.08.1985
Nationality: Taiwan
Phone: 886-912-592559
Email: [email protected]
Address: 1F., No. 49, Chongsheng St.,
Yilan City, Yilan Country 260, Taiwan

Work Experience
June 2015 to date Principal Engineer of Data Analysis at TSMC, Taiwan
Algorithm optimization, modeling and numerical simulation

Sep 2009–Jul 2015 Research Assistant at NTU-HEP Lab., Taipei, Taiwan

CERN, Geneva, Swissland
Developed and programmed for numerical analysis and simulation
Collaborated with multiple developer on several complex projects

Nov 2011–Jul 2015 Manager Assistant at Grid Center of NTU, Taipei, Taiwan
Assisted with the management of Unix/Linux server
Assisted with the establishment of cluster system
Monitored network and system operations
Designed and set up the website

© Springer Nature Singapore Pte Ltd. 2017 93

Y.-H. Chang, Study of Double Parton Scattering in Photon + 3 Jets Final State,
Springer Theses, DOI 10.1007/978-981-10-3824-2
94 Curriculum Vitae

Education
July 2015 Ph.D – Experimental High Energy Physics
National Taiwan University, Taipei, Taiwan
GPA: 4.0
Thesis: Study of Double Parton Scattering in Photon + 3 Jets
Final State in Proton-proton Collisions at 7 TeV at the CMS at the LHC.

July 2009 Master of Science – Physics

National Taiwan University, Taipei, Taiwan
GPA: 3.7
Thesis: Monte-Carlo Data Validation of the Minimum Bias and Under-
lying Events at the LHC.

July 2007 Bachelor of Science – Geography

National Taiwan University, Taipei, Taiwan
GPA: 3.7

Publications
Apr 2015 Study of Double Parton Scattering in Photon + 3 Jets
Final State in Proton-proton Collisions at 7TeV
The CMS Collaboration, CMS-PAS-FSQ-12-017

Jul 2011 Measurement

√ of the Underlying Event Activity
√ at the LHC
with s = 7 TeV and Comparison with s = 0.9 TeV
The CMS Collaboration, JHEP 9 (2011) 109

Jun 2010 Measurement of the Underlying Event Activity in Proton-Proton

Collisions at 0.9 TeV
The CMS Collaboration, Eur. Phys. J. C70:555-572, 2010

Accomplishments and Talks at Conferences

Jun 2015 NTU Dean’s Award of College of Science

Apr 2015 Measurement of Double Parton Scattering at LHC with

the CMS experiment, Deep Inelastic Scattering (DIS)
Conference 2014, Dallas, Texas, US
Curriculum Vitae 95

Jan 2014 Double Parton Scattering Study in Photon + 3 Jets Final State
at the CMS Experiment, the Physical Society of the Republic
of China (PSROC) Annual Meeting 2014, Taichung, Taiwan

Feb 2010 Monte-Carlo Data Validation of the Minimum Bias and

Underlying Events at LHC, the Physical Society of the Republic
of China (PSROC) Annual Meeting 2010, Tainan, Taiwan

Jan 2009 The Poster Award of PSROC 2009 Annual Meeting

Title: Minimum Bias and Underlying Event Studies at CMS

Languages
Taiwan Mandarin: Mother tongue
English: Fluent
Japanese: Intermediate
French: Basic Knowledge
German: Basic Knowledge

Technical Skills and Computer Knowledge

Word Processing: Word, Latex, Excel, Powerpoint, Vim
Programming Languages: C, C++, Python, JavaScript, jQuery, Qml, Shell script
Development Tools: Git
Operation Systems: Linux, Mac OS, Windows
Generator Software: Pythia6, Pythia8, Herwig6, Herwig++, Sherpa, Mad-
graph
Simulation Software: Geant4
Analysis Software: Root, CMSSW, Crab

Interests and Activities

Cycling, Running, Swimming, Baseball, Basketball
Traveling, Reading, Painting, Language learning
Technology, Open-Source, Programming
References

1. CMS Collaboration,√ V. Khachatryan et al., “First Measurement of the Underlying Event Activ-
ity at the LHC with s = 0.9 TeV,” Eur. Phys. J. C 70 (2010) 555–572, arXiv:1006.2083.
2. CMS Collaboration,
√ S. Chatrchyan et al., “Measurement
√ of the Underlying Event Activity at
the LHC with s = 7 TeV and Comparison with s = 0.9 TeV,” JHEP 1109 (2011) 109,
arXiv:1107.0330.
3. CMS Collaboration, S. Chatrchyan et al., “Measurement
√ of the underlying event in the Drell-
Yan process in proton-proton collisions at s = 7 TeV,” Eur. Phys. J. C 72 (2012) 2080,
arXiv:1204.1411.
4. C. Goebel, F. Halzen, and D. Scott, “Double Drell-Yan Annihilations in Hadron Collisions:
Novel Tests of the Constituent Picture,” Phys. Rev. D 22 (1980) 2789.
5. N. Paver and D. Treleani, “Multi - Quark Scattering and Large pT Jet Production in Hadronic
Collisions,” Nuovo Cim. A 70 (1982) 215.
6. B. Humpert, “ARE THERE MULTI - QUARK INTERACTIONS?,” Phys. Lett. B 131 (1983)
461.
7. B. Humpert and R. Odorico, “Multiparton Scattering and QCD Radiation as Sources of Four
Jet Events,” Phys. Lett. B 154 (1985) 211.
8. T. Sjöstrand and M. van Zijl, “A Multiple Interaction Model for the Event Structure in Hadron
Collisions,” Phys. Rev. D 36 (1987) 2019.
9. G. Calucci and D. Treleani, “Double parton scatterings in high-energy hadronic collisions,”
Nucl. Phys. Proc. Suppl. 71 (1999) 392–399, arXiv:hep-ph/9711225.
10. M. Drees and T. Han, “Signals for Double Parton Scattering at the Fermilab Tevatron,” Phys.
Rev. Lett. 77 (Nov, 1996) 4142–4145.
11. P. Bartalini, E. L. Berger, B. Blok, G. Calucci, R. Corke, et al., “Multi-Parton Interactions at
the LHC,” ((2011)), no. ANL-HEP-PR-11-65. CMS-CR-2011-048. DESY 11-185. KA-TP-
32-2011. TTK-11-52, arXiv:1111.0469.
12. M. Mekhfi, “Multiparton Processes: an Application to Double Drell-Yan,” Phys. Rev. D 32
(1985) 2371.
13. L. Ametller, N. Paver, and D. Treleani, “Possible Signature of Multiple Parton Interactions in
Collider Four Jet Events,” Phys. Lett. B 169 (1986) 289.
14. R. Godbole, S. Gupta, and J. Lindfors, “Double Parton Scattering Contribution to W + Jets,”
Z. Phys. C 47 (1990) 69.
15. J. R. Gaunt, C.-H. Kom, A. Kulesza, and W. J. Stirling, “Same-sign W pair production as a
probe of double parton scattering at the LHC,” Eur. Phys. J. C 69 (2010) 53, arXiv:1003.3953.
√
16. CDF Collaboration, F. Abe et al., “Double parton scattering in pp collisions at s = 1.8 TeV,”
Phys. Rev. D 56 (Oct, 1997) 3811.
© Springer Nature Singapore Pte Ltd. 2017 97
Y.-H. Chang, Study of Double Parton Scattering in Photon + 3 Jets Final State,
Springer Theses, DOI 10.1007/978-981-10-3824-2
98 References

17. D0 Collaboration,
√ V. Abazov et al., “Double parton interactions in photon+3 jet events in pp
collisions s = 1.96 TeV,” Phys. Rev. D 81 (2010) 052012, arXiv:0912.5104.
18. D0 Collaboration, V. M. Abazov et al., “Azimuthal decorrelations√ and multiple parton interac-
tions in photon+2 jet and photon+3 jet events in pp̄ collisions at s = 1.96 TeV,” Phys. Rev.
D 83 (2011) 052008, arXiv:1101.1509.
19. D0 Collaboration, V. M. Abazov et al., √ “Double parton interactions in γ + 3 jet and γ +
b/cjet + 2 jet events in pp̄ collisions at s = 1.96 TeV,” Phys.Rev. D89 (2014), no. 7, 072006,
arXiv:1402.1550.
20. Axial Field Spectrometer
√ Collaboration, T. Akesson et al., “Double Parton Scattering in pp
Collisions at s = 63 GeV,” Z. Phys. C 34 (1987) 163.
21. UA2 Collaboration, J. Alitti et al., “A Study of multi-jet events at the CERN pp collider and a
search for double parton scattering,” Phys. Lett. B 268 (1991) 145.
22. CDF Collaboration, F. Abe et√ al., “Study of four jet events and evidence for double parton
interactions in pp̄ collisions at s = 1.8 TeV,” Phys. Rev. D 47 (1993) 4857.
23. CMS Collaboration,
√ S. Chatrchyan et al., “Measurement of four-jet production in proton-proton
collisions at s=7 TeV,” Phys. Rev. D 89 (2014) 092010, arXiv:1312.6440.
24. CMS Collaboration, S. Chatrchyan et√al., “Study of double parton scattering using W + 2-jet
events in proton-proton collisions at s = 7 TeV,” JHEP 03 (2014) 032, arXiv:1312.5729.
25. ATLAS Collaboration, G. Aad √ et al., “Measurement of hard double-parton interactions in
W (→ lν)+ 2 jet events at s=7 TeV with the ATLAS detector,” New J. Phys. 15 (2013)
033038, arXiv:1301.6872.
26. M. Glück, L. Gordon, E. Reya, and W. Vogelsang, “High P(T ) photon production at pp̄ collider,”
Phys. Rev. Lett. 73 (1994) 388.
27. M. Cacciari, G. P. Salam, and G. Soyez, “The Anti-k(t) jet clustering algorithm,” JHEP 04
(2008) 063, arXiv:0802.1189.
28. M. Cacciari, G. P. Salam, and G. Soyez, “FastJet User Manual,” Eur. Phys. J. C 72 (2012)
1896, arXiv:1111.6097.
29. O. S. Bruning, (Ed.) et al., “LHC design report. Vol. I: The LHC main ring,”. CERN-2004-
003-V-1.
30. CMS Collaboration, “CMS, the Compact Muon Solenoid: Technical proposal,”
CERN/LHCC 94-38 (1994).
31. ATLAS Collaboration, W. W. Armstrong et al., “ATLAS: Technical proposal for a general-
purpose pp experiment at the Large Hadron Collider at CERN,”. CERN-LHCC-94-43.
32. ALICE Collaboration, S. Ahmad et al., “ALICE: Technical proposal for a Large Ion collider
Experiment at the CERN LHC,”. CERN-LHCC-95-71.
33. LHCb Collaboration, S. Amato et al., “LHCb technical proposal,”. CERN-LHCC-98-04.
34. TOTEM Collaboration, V. Berardi et al., “TOTEM: Technical design report. Total cross
section, elastic scattering and diffraction dissociation at the Large Hadron Collider at CERN,”.
CERN-LHCC-2004-002.
35. LHCf Collaboration, O. Adriani et al., “Technnical proposal for the CERN LHCf experiment:
Measurement of photons and neutral pions in the very forward region of LHC,”. CERN-LHCC-
2005-032.
36. CMS Collaboration, S. Chatrchyan et al., “The CMS experiment at the CERN LHC,” JINST 3
(2008) S08004.
37. CMS Collaboration, CMS, “Jet Energy Corrections Determination at 7 TeV,” CMS Physics
Analysis Summary CMS-PAS-JME-10-010, CERN, 2010.
38. M. Cacciari and G. P. Salam, “Pileup subtraction using jet areas,” Phys. Lett. B 659 (2008) 119,
arXiv:0707.1378.
39. CMS Collaboration, “Particle-Flow Event Reconstruction in CMS and Performance for Jets,
Taus, and ETmiss ,” CMS Physics Analysis Summary CMS-PAS-PFT-09-001, CERN, 2009.
40. CMS Collaboration, “Commissioning of the Particle-flow Event Reconstruction with the first
LHC collisions recorded in the CMS detector,” CMS Physics Analysis Summary CMS-PAS-
PFT-10-001, CERN, 2010.
References 99

41. CMS Collaboration, V. Khachatryan et al., “Performance of photon √ reconstruction and iden-
tification with the CMS detector in proton-proton collisions at s = 8 TeV.” Submitted to
JINST, 2015.
42. CMS Collaboration, S. Chatrchyan et al., “Commissioning of the CMS High-Level Trigger
with Cosmic Rays,” JINST 5 (2010) T03005, arXiv:0911.4889.
43. CMS Collaboration, “Absolute Calibration of Luminosity Measurement at CMS: Summer
2011 Update,” CMS Physics Analysis Summary CMS-PAS-EWK-11-001 (2011).
44. S. van der Meer, “Calibration of the effective beam height in the ISR,” CERN-ISR-PO/68-31
(1968).
45. CMS Collaboration, D. Spiga, “CMS workload management,” Nucl. Phys. Proc. Suppl 172
(2007) 141–144.
46. T. Sjöstrand, S. Mrenna, and P. Skands, “A Brief Introduction to PYTHIA 8.1,” Comput. Phys.
Comm. 178 (2007) 852, arXiv:hep-ph/0710.3820.
47. T. Sjöstrand, S. Mrenna, and P. Skands, “PYTHIA 6.4 physics and manual,” JHEP 05 (2006)
026, arXiv:hep-ph/0603175.
48. F. Maltoni and T. Stelzer, “MadEvent: Automatic event generation with MadGraph,” JHEP 02
(2003) 027, arXiv:hep-ph/0208156.
49. A. Johan, H. Michel, M. Fabio, M. Olivier, and S. Tim, “MadGraph 5: Going Beyond,” JHEP 06
(2011) 128, arXiv:hep-ph/1106.0522.
50. J. Alwall, R. Frederix, S. Frixione, V. Hirschi, F. Maltoni, O. Mattelaer, H.-S. Shao, T. Stelzer,
P. Torrielli, and M. Zaro, “The automated computation of tree-level and next-to-leading order
differential cross sections, and their matching to parton shower simulations,” JHEP 07 (2014)
079, arXiv:1405.0301.
51. A. Buckley, H. Hoeth, H. Lacker, H. Schulz, and J. E. von Seggern, “Systematic event generator
tuning for the LHC,” Eur. Phys. J. C 65 (2010) 331–357, arXiv:0907.2973.
52. S. Agostinelli, J. Allison, K. Amako, J. Apostolakis, H. Araujo, P. Arce, M. Asai, D. Axen,
S. Banerjee, G. Barrand, F. Behner, L. Bellagamba, J. Boudreau, L. Broglia, A. Brunengo,
H. Burkhardt, S. Chauvie, J. Chuma, R. Chytracek, G. Cooperman, G. Cosmo, P. Degtyarenko,
A. Dell’Acqua, G. Depaola, D. Dietrich, R. Enami, A. Feliciello, C. Ferguson, H. Fesefeldt,
G. Folger, F. Foppiano, A. Forti, S. Garelli, S. Giani, R. Giannitrapani, D. Gibin, J. G. Cadenas,
I. González, G. G. Abril, G. Greeniaus, W. Greiner, V. Grichine, A. Grossheim, S. Guatelli,
P. Gumplinger, R. Hamatsu, K. Hashimoto, H. Hasui, A. Heikkinen, A. Howard, V. Ivanchenko,
A. Johnson, F. Jones, J. Kallenbach, N. Kanaya, M. Kawabata, Y. Kawabata, M. Kawaguti,
S. Kelner, P. Kent, A. Kimura, T. Kodama, R. Kokoulin, M. Kossov, H. Kurashige, E. Lamanna,
T. L. V. Lara, V. Lefebure, F. Lei, M. Liendl, W. Lockman, F. Longo, S. Magni, M. Maire,
E. Medernach, K. Minamimoto, P. M. de Freitas, Y. Morita, K. Murakami, M. Nagamatu,
R. Nartallo, P. Nieminen, T. Nishimura, K. Ohtsubo, M. Okamura, S. O’Neale, Y. Oohata,
K. Paech, J. Perl, A. Pfeiffer, M. Pia, F. Ranjard, A. Rybin, S. Sadilov, E. D. Salvo, G. Santin,
T. Sasaki, N. Savvas, Y. Sawada, S. Scherer, S. Sei, V. Sirotenko, D. Smith, N. Starkov,
H. Stoecker, J. Sulkimo, M. Takahata, S. Tanaka, E. Tcherniaev, E. S. Tehrani, M. Tropeano,
P. Truscott, H. Uno, L. Urban, P. Urban, M. Verderi, A. Walkden, W. Wander, H. Weber,
J. Wellisch, T. Wenaus, D. Williams, D. Wright, T. Yamada, H. Yoshida, and D. Zschiesche,
“Geant4–a simulation toolkit,” NIM A 506 (2003) 250.
53. J. Allison, K. Amako, J. Apostolakis, H. Araujo, P. Dubois, M. Asai, G. Barrand, R. Capra,
S. Chauvie, R. Chytracek, G. A. P. Cirrone, G. Cooperman, G. Cosmo, G. Cuttone, G. G.
Daquino, M. Donszelmann, M. Dressel, G. Folger, F. Foppiano, J. Generowicz, V. Grichine,
S. Guatelli, P. Gumplinger, A. Heikkinen, I. Hrivnacova, A. Howard, S. Incerti, V. Ivanchenko,
T. Johnson, F. Jones, T. Koi, R. Kokoulin, M. Kossov, H. Kurashige, V. Lara, S. Larsson,
F. Lei, O. Link, F. Longo, M. Maire, A. Mantero, B. Mascialino, I. McLaren, P. Lorenzo,
K. Minamimoto, K. Murakami, P. Nieminen, L. Pandola, S. Parlati, L. Peralta, J. Perl, A. Pfeif-
fer, M. Pia, A. Ribon, P. Rodrigues, G. Russo, S. Sadilov, G. Santin, T. Sasaki, D. Smith,
N. Starkov, S. Tanaka, E. Tcherniaev, B. Tome, A. Trindade, P. Truscott, L. Urban, M. Verderi,
A. Walkden, J. P. Wellisch, D. Williams, D. Wright, and H. Yoshida, “Geant4 Developments
and Applications,” IEEE Trans. Nucl. Sci. 53 (2006) 270.
100 References

54. R. Field, “Early LHC Underlying Event Data - Findings and Surprises,” (2010)
arXiv:1010.3558.
55. J. Pumplin, D. Stump, J. Huston, H. Lai, P. M. Nadolsky, et al., “New generation of
parton distributions with uncertainties from global QCD analysis,” JHEP 07 (2002) 012,
arXiv:hep-ph/0201195.
56. P. Z. Skands and D. Wicke, “Non-perturbative QCD effects and the top mass at the Tevatron,”
Eur. Phys. J. C 52 (2007) 133–140, arXiv:hep-ph/0703081.
57. R. Corke and T. Sjöstrand, “Interleaved Parton Showers and Tuning Prospects,” JHEP 03
(2011) 032, arXiv:1011.1759.
58. R. Corke and T. Sjöstrand, “Multiparton Interactions and Rescattering,” JHEP 01 (2010) 035,
arXiv:0911.1909.
59. CMS Collaboration, √ S. Chatrchyan et al., “Study of the underlying event at forward rapidity
in pp collisions at s = 0.9, 2.76, and 7 TeV,” JHEP 04 (2013) 072, arXiv:1302.2394.
60. T. Gleisberg, S. Höche, F. Krauss, M. Schonherr, S. Schumann, et al., “Event generation with
SHERPA 1.1,” JHEP 02 (2009) 007, arXiv:0811.4622.
61. S. Schumann and F. Krauss, “A Parton shower algorithm based on Catani-Seymour dipole
factorisation,” JHEP 03 (2008) 038, arXiv:0709.1027.
62. T. Gleisberg and S. Höche, “Comix, a new matrix element generator,” JHEP 12 (2008) 039,
arXiv:0808.3674.
63. S. Höche, F. Krauss, S. Schumann, and F. Siegert, “QCD matrix elements and truncated show-
ers,” JHEP 05 (2009) 053, arXiv:0903.1219.
64. M. Schönherr and F. Krauss, “Soft Photon Radiation in Particle Decays in SHERPA,” JHEP 12
(2008) 018, arXiv:0810.5071.
65. H.-L. Lai, M. Guzzi, J. Huston, Z. Li, P. M. Nadolsky, et al., “New parton distributions for
collider physics,” Phys. Rev. D 82 (2010) 074024, arXiv:1007.2241.
√
66. CMS Collaboration, “Photon reconstruction and identification at s = 7TeV,” CMS Physics
Analysis Summary CMS-PAS-EGM-10-005, CERN, 2010.
67. CMS Collaboration, “Electromagnetic calorimeter commissioning and first results with 7 TeV
data,” CMS Note 2010/012, CERN, 2010.
68. CMS Collaboration, “Measurement of the Differential Cross Section for Isolated Prompt
Photon Production in pp Collisions at 7 TeV,” Phys. Rev. D 84 (2011) 052011. 45 p,
arXiv:hep-ex/1108.2044.
69. CMS Collaboration, V. Khachatryan et al.,√“Measurement of the Isolated Prompt Photon
Production Cross Section in pp Collisions at s = 7 TeV,” Phys. Rev. Lett. 106 (2011) 082001,
arXiv:1012.0799.
70. CMS Collaboration, S. Chatrchyan et al., “Measurement of the Differential Cross Section for
Isolated Prompt Photon Production in pp Collisions at 7 TeV,” Phys. Rev. D 84 (2011) 052011,
arXiv:1108.2044.
71. CMS Collaboration, “Electromagnetic calorimeter calibration with 7 TeV data,” CMS Physics
Analysis Summary CMS-PAS-EGM-10-003, CERN, Geneva, 2010.
72. CMS Collaboration, S. Chatrchyan√et al., “Measurement of the Inclusive W and Z Production
Cross Sections in pp Collisions at s = 7 TeV,” JHEP 10 (2011) 132, arXiv:1107.4789.
73. CMS Collaboration, “ECAL 2010 performance results,” Tech. Rep. CMS-DP-2011-008,
CERN, Geneva, 2011.
74. CMS Collaboration, “Commissioning of the Particle-Flow reconstruction in Minimum-Bias
and Jet Events from pp Collisions at 7 TeV,” Tech. Rep. CMS-PAS-PFT-10-002, CERN, Geneva,
2010.
75. CMS Collaboration, “Jet Energy Corrections determination at 7 TeV,” Tech. Rep. CMS-PAS-
JME-10-010, CERN, Geneva, 2010.
76. CMS Collaboration,
√ S. Chatrchyan et al., “Measurement of the Inclusive Jet Cross Section in
pp Collisions at s = 7 TeV,” Phys. Rev. Lett. 107 (2011) 132001, arXiv:1106.0208.
77. G. D’Agostini, “A Multidimensional unfolding method based on Bayes’ theorem,” Nucl.
Instrum. Meth. A 362 (1995) 487.
78. T. Adye, “Unfolding algorithms and tests using RooUnfold,” (2011) arXiv:1105.1160.
References 101

79. A. Höcker and V. Kartvelishvili, “SVD approach to data unfolding,” Nucl. Instrum. Meth. A 372
(1996) 469, arXiv:hep-ph/9509307.
80. CMS Collaboration, S. Chatrchyan et al., “Measurement of √ the triple-differential cross section
for photon+jets production in proton-proton collisions at s=7 TeV,” JHEP 1406 (2014) 009,
arXiv:1311.6141.
81. CMS Collaboration, S. Chatrchyan et al., “Energy √ calibration and resolution of the CMS
electromagnetic calorimeter in pp collisions at s = 7 TeV,” JINST 8 (2013) P09009,
arXiv:1306.2016.
82. CMS Collaboration, S. Chatrchyan et al., “Determination of jet energy calibration and trans-
verse momentum resolution in CMS,” JINST 6 (2011) P11002, arXiv:1107.4277.
83. CMS Collaboration, “CMS twiki page: Pile-up Systematic Errors,” tech. rep., CERN, 1900.
Geneva, 2012.
84. CMS Collaboration, S. Chatrchyan et al., “Description and performance of track and primary-
vertex reconstruction with the CMS tracker,” JINST 9 (2014), no. 10, P10009, arXiv:1405.6569.
85. S. Catani, M. Fontannaz, J. Guillet, and E. Pilon, “Cross-section of isolated prompt photons in
hadron hadron collisions,” JHEP 0205 (2002) 028, arXiv:hep-ph/0204023.
86. P. Aurenche, M. Fontannaz, J.-P. Guillet, E. Pilon, and M. Werlen, “A New critical study of pho-
ton production in hadronic collisions,” Phys.Rev. D73 (2006) 094007, arXiv:hep-ph/0602133.
87. C. Berger, Z. Bern, L. J. Dixon, F. Febres Cordero, D. Forde, et al., “Next-to-Leading Order
Jet Physics with BlackHat,” PoS RADCOR2009 (2010) 002, arXiv:0912.4927.
88. Z. Bern, L. Dixon, F. F. Cordero, S. Hoeche, H. Ita, et al., “Ntuples for NLO Events at Hadron
Colliders,” arXiv:1310.7439.
89. A. Martin, W. Stirling, R. Thorne, and G. Watt, “Parton distributions for the LHC,”
Eur.Phys.J. C63 (2009) 189–285, arXiv:0901.0002.