Kinematics Comparison of Monte Carlo to Real
Data for Z
Kunal Garg
NISER
School of Physical Sciences
Abstract
This project involved an extensive usage of ROOT framework, an Object Oriented C++ framework. As a first step gaining familiarity with ROOT had been
essential. Using the ROOT framework, Z events are analysed from the
4.7fb1 of proton-proton collision data recorded by the ATLAS experiment. In
such events, detailed kinematics comparison between the data and Monte-Carlo
(simulated and reconstructed with the ATLAS detectors and software) are presented here.
0.1
Introduction
The basic physical process we study here is called Drell-Yan process. It is an
electromagnetic effect that takes place in high energy hadron-hadron scattering.
When a quark from one hadron and anti-hadron from another hadron annihilate,
in the process creating a virtual Z0 boson which then decays into a pair of
oppositely charged leptons. This process was first suggested by Sidney Drell and
Tung-Mow Yan in 1970 to describe the production of leptonanti-lepton pairs in
high-energy hadron collisions. Experimentally, this process was first observed by
J.H. Christenson et al. in protonuranium collisions at the Alternating Gradient
Synchrotron.
Figure 1: Drell-Yan Process
The ATLAS experiment identifies and reconstructs muons with two high
precision tracking systems, the Inner Detector (ID) and the Muon Spectrometer
(MS) which provide independent measurements of the muon momentum. In this
report, we compare the performance of the reconstruction algorithm and data as
derived from the dataset corresponding to an integrated luminosity of 4.7 fb1
of 7 TeV pp collisions recorded in 2011. Also there is described the corrections
to be applied to the simulation to reproduce the momentum resolution and scale
observed in experimental data.
0.2
Drell-Yan Process
The partons
The strongly interaction particles such as mesons and nucleons are not elementary. They are made up from simpler particles: quarks, antiquarks and gluons,
known collectively as partons. The quarks are point like, spin- 21 fermions, properties they share with other class of matter particles, the leptons. Quarks differ
from leptons in that they interact strongly via gluon exchange, whilst leptons
2
only feel electroweak force. Evidence of internal structure of hadrons has come
from deep inelastic lepton scattering of nucleons and from hadron spectrum
(Gell-Mann 1964, Zweig 1964).
Colour
Electroweak and gravitational forces are known to be gauge interactions transmitted by gauge fields. In this case of electroweak interactions, the fields are
quantised as spin-1 particles, namely photons and intermediate vector bosons.
It is expected that the quark-quark strong interaction is also a gauge force.
Greenberg (1964) deduced evidence for a new symmetry called colour, which
turned out to be the gauge symmetry. He pointed out that if the account is
taken of the spin, flavour and relative motion, then the quarks inside baryons
+
with spin parity 32 are in totally symmetric state. This contradicts the Pauli
exclusion principle. In order to resolve this [Link] proposed that
a further three-fold symmetry existed, SU (3) of colour, with the quarks assigned to the fundamental three-representation (red, blue and green). Then if
the quarks in a baryon are arranged in an antisymmetric colour state with no
net colour, the difficulty disappears. The solution is to have one red, one blue
and one green quark inside each; in the mesons a quark of one colour is paired
with an antiquark of the same colour. The state of the + meson in this scheme
becomes
+ u(b)d(b)
+ u(g)d(g)
(1/ 3) u(r)d(r)
The parton model
The interactions of leptons with hadrons lend themselves to a treatment by
an impulse approximation when the relative four-momentum squared |q|2 is
large. Figure 2 shows the form this takes for (a) the Drell-Yan process, (b)
e+ e annihilation and (c) deep inelastic lepton scattering. The lepton (line) is
supposed to interact instantaneously with a single quark (line), and any partonparton interactions are ignored. On a longer time scale the partons interact
to give the hadronic final state. This impulse approximation leads to simple
electromagnetic calculations of the cross sections. Bjorken originally proved
the model for lepton scattering at large energies and introduced the Bjorken
x variable (Bjorken 1969); x is the fraction of longitudinal momentum of the
hadron that is carried by a given parton, and must lie in the range 0-1. The
asymptotic freedom limit gives a theoretic basis to the impulse approximation
inside QCD. When the inelastic momentum transfer to a hadron is large enough,
it will appear to be made up of quasi-free point-like partons.
Structure Functions
The structure functions are the covariant scalar quantities used to describe
deep inelastic scattering from nucleons. Most determinations of the structure
functions of nucleons have been made in deep inelastic scattering experiments
for the reactions:
l + nucleon l + hadrons
+ nucleons l + hadrons
and their analogues. Under the assumption that a single vector boson exchange dominates, the dynamics of the reaction is described by the three structure functions F1 , F2 and F3 , which are in turn functions of the two covariant
kinematic quantities:
x = q 2 /2p.q
and
q2
where q is the four-momentum transfer between the leptons and the target
nucleons and p is the incident nucleon four-momentum.
Of central interest to us here is the interpretation of x and the structure
functions in the parton model; x is to be interpreted as the fraction of longitudinal momentum of the parent hadron (Bjorken x) that is carried by the active,
struck quark in the overall centre-of-mass frame. If n(x)dx overall centre-ofmass frame. If n (x) dx is the number of quarks carrying longitudinal momentum
fractions between x and x + dx, then
X
F2l (x) =
Q2i xni (x)
(charged leptons)
i
F2 (x) =
xni (x)
(neutrinos)
where the sums are taken over all quark flavours. The difference between
charged lepton and neutrino scattering arises from the difference between the
weak and electromagnetic forces at usual experimental energies. The electromagnetic coupling requires the appearance of the quark charge, namely Q in
units of e. In addition, the model requires exact relationships to hold between
the structure functions:
F2l = 2xF1l
F2 = xF3 = 2xF1
and
where the sign is positive for neutrino-quark and antineutrino-antiquark scattering and negative for the other choices. Bjorken (1969) argued that the structure
functions at large q 2 and finite x are functions of x alone (Bjorken scaling).
Dynamics of Drell-Yan process
Drell and Yan (1970a, 1971) proposed a simple electromagnetic process to account for the production of the continuum of dilepton pairs from hadron-hadron
collisions:
A + B l+ + l + anything
The essential steps are the annihilation of a single antiquark from one hadron
on a single quark from the other hadron to produce a virtual photon. Subsequently the virtual photon converts to a pair of leptons. If the mass M of the
dilepton is large compared to the nucleon mass, then the Heisenberg uncertainty principle tells us that the time of interaction is short on the nuclear scale.
There is thus no opportunity for annihilating quark-antiquark pair to interact
with other components of the parent hadrons. The other partons can be treated
purely as spectators. Subsequently, on a time scale that is long compared to the
annihilation, the spectators rearrange themselves into outgoing hadrons. Generally the leptons alone are detected and momentum-analysed. This procedure
measures the inclusive cross section for the dilepton production and sums over
all possible combinations of the final-state hadrons. Thus in the calculation of
the cross section the factor expressing the probability that the spectators convert to hadrons us simply unity. The dilepton kinematic variables and xF are
directly related to the Bjorken x values (longitudinal momenta) of the parent
quark and anti-quark. Suppose the active quark (antiquark) in the beam particle has a Bjorken x of x1 , and the active antiquark (quark) in the target has
a Bjorken x of x2 . Then in the CM frame
of beam plus
target the annihilat
s/2 and x2s/2, respectively. If the
ing pair have longitudinal momenta x1
quarks are massless their energies are x1 s/2 and x2 s/2. Then equating the
four-momentum of the annihilating pair to the dilepton four-momentum (E, pl )
gives
E = (x1 + x2 ) s/2
(1)
and
pl = (x1 x2 ) s/2
(2)
The dilepton mass squared is
M 2 = E 2 p2l = sx1 x2
(3)
Two basic relations between quark variables and the observables now emerge
= M 2 /s = x1 x2
(4)
xF = 2pl / s = x1 x2
(5)
and
0.3
ATLAS detectors
ATLAS (A Toroidal LHC Apparatus) is one of the seven particle detector experiments (ALICE, ATLAS, CMS, TOTEM, LHCb, LHCf and MoEDAL) constructed at the Large Hadron Collider (LHC). ATLAS is 45 metres long, 25
metres in diameter, and weighs about 7,000 tons. The detector consists of a
series of ever-larger concentric cylinders around the interaction point where the
proton beams collide. It can be divided into four major parts: the Inner Detector, the calorimeters, the Muon Spectrometer and the magnet systems. Of
special interest to us are the Inner Detector (ID) and the Muon Spectrometer
(MS) and to a lesser extent the calorimeter.
MDT chambers
12 m
Resistive plate chambers
10
Barrel toroid coil
8
Thin gap
chambers
End-cap
toroid
2
Radiation shield
Cathode strip
chambers
0
20
18
16
14
12
10
2m
Figure 2: ATLAS design
0.3.1
Inner Detector
The Inner Detector begins a few centimetres from the proton beam axis, extends
to a radius of 1.2 metres, and is 6.2 metres in length along the beam pipe. Its
basic function is to track charged particles by detecting their interaction with
material at discrete points, revealing detailed information about the types of
particles and their momentum. The magnetic field surrounding the entire inner
detector causes charged particles to curve; the direction of the curve reveals
a particles charge and the degree of curvature reveals its momentum. The
starting points of the tracks yield useful information for identifying particles as
well.
In the barrel region the high-precision detectors are arranged in concentric
cylinders around the beam axis, while the end-cap detectors are mounted on
disks perpendicular to the beam axis. The barrel TRT straws are parallel to
Figure 3: Inner Detector
the beam direction. All end-cap tracking elements are located in planes perpendicular to the beam direction. ID has three parts:
Pixel Detector - innermost part of the detector, contains three concentric layers and three disks on each end-cap. It has minute pixel size
designed for extremely precise tracking very close to the interaction point.
Semi-Conductor Tracker (SCT) - It is similar in concept and function
to the Pixel Detector but with long, narrow strips rather than small pixels,
making coverage of a larger area practical. The SCT is the most critical
part of the inner detector for basic tracking in the plane perpendicular
to the beam, since it measures particles over a much larger area than the
Pixel Detector.
Transition Radiation Tracker (TRT) - Its the outermost component
of the detector. The detecting elements are drift tubes (straws). Each
straw is filled with gas that becomes ionised when a charged particle passes
through. Between the straws, materials with widely varying indices of
refraction cause ultra-relativistic charged particles to produce transition
radiation and leave much stronger signals in some straws. Xenon gas
is used to increase the number of straws with strong signals. Since the
amount of transition radiation is greatest for highly relativistic particles,
and because particles of a particular energy have a higher speed the lighter
they are, particle paths with many very strong signals can be identified as
belonging to the lightest charged particles: electrons and its antiparticles,
positrons.
0.3.2
Calorimeters
The calorimeters are situated outside the solenoidal magnet that surrounds the
Inner Detector. Their purpose is to measure the energy from particles by absorbing it. There are two basic calorimeter systems: an inner electromagnetic
calorimeter and an outer hadronic calorimeter. Both are sampling calorimeters; that is, they absorb energy in high-density metal and periodically sample
the shape of the resulting particle shower, inferring the energy of the original
particle from this measurement.
0.3.3
Muon Spectrometer
Muons are particles just like electrons, but 200 times heavier. They are the
only detectable particles that can traverse all the calorimeter absorbers without
being stopped. The muon spectrometer surrounds the calorimeter and measures muon paths to determine their momenta with high precision. The Muon
Spectrometer is an extremely large tracking system, consisting of 3 parts: (1) a
magnetic field provided by three toroidal magnets, (2) a set of 1200 chambers
measuring with high spatial precision the tracks of the outgoing muons, (3) a
set of triggering chambers with accurate time-resolution. Its tremendous size is
required to accurately measure the momentum of muons, which first go through
all the other elements of the detector before reaching the muon spectrometer.
Figure 4: Muon Spectrometer
It functions similarly to the Inner Detector, with muons curving so that their
momentum can be measured, albeit with a different magnetic field configuration,
lower spatial precision, and a much larger volume. It also serves the function
8
of simply identifying muons very few particles of other types are expected
to pass through the calorimeters and subsequently leave signals in the Muon
Spectrometer.
0.4
- scale and Momentum Resolution
Efficient and accurate muon identification and reconstruction is of primary importance for the physics program of the ATLAS experiment. The experiment
uses the information of MS and ID sub-detectors, and to a lesser extent, of the
calorimeter, to identify and precisely reconstruct muon produced in the pp collisions.
The momentum resolution and scale are additional important parameters used
in the evaluation of the muon reconstruction performance. Di-muon decays
of the Z, J/ and resonances are used to determine the muon momentum
resolution and scale. This allows a validation of the MC prediction for these
quantities.
0.5
Monte Carlo and Data
The results presented here are obtained from the analysis of the s = 7 TeV
pp collision events corresponding to an integrated luminosity of 4fb1 and collected by the ATLAS detector in 2011. Online event selection is performed by
a three level trigger system. Events were accepted for the analysis documented
only if the ID and MS detectors were in good data-taking conditions and both
solenoidal and toroidal magnet systems were on.
In the following, experimental data are compared to MC simulations. The
Z signal MCs are generated with ALPGEN. The generated signal events
are passed through the full Geant4 simulation of the ATLAS detector, the trigger simulation and same reconstruction chain used for data. The simulation includes a realistic evaluation of the MS detectors alignment obtained by studying
straight muon tracks from cosmic ray events and from special runs performed
with toroidal magnetic field off.
We selected the bins during our analysis of the two data sets from the detector and MC based on the detector design. And there are two peaks at = 1.05
and = 1.9 and end caps at = 0 and = 2.5 So we selected three regions
based on these numbers as
|| < 1.05
1.05 < || < 1.9
1.9 < || < 2.5
There are several different parameters which contribute to the resolution of
the detector at different energy scales.
Contribution to resolution (%)
12
Total
Spectrometer entrance
Multiple scattering
Chamber Alignment
Tube resolution and autocalibration (stochastic)
10
Energy loss fluctuations
8
6
4
2
0
10
102
103
Pt (GeV/c)
Figure 5: Contribution to Resolution
0.6
Results
First we plot the Invariant Mass of the Z boson calculated from both MC as
well as Data and then fitted to a double Gaussian for different bins. Since there
are two muons and three bins, well have six different combinations of the bins.
10
Number of events/GeV
Invariant Mass (Muon1 ||<1.05, Muon2 ||<1.05)
ATLAS Preliminary
s=7 TeV;
50000
L dt = 4.7 fb
-1
: 90.244.29
40000
Data
Data Fit
2/NDF: 13.28/8
MC
: 91.011.825
30000
MC Fit
2/NDF: 67.59/9
20000
10000
0
80
82
84
86
88
90
92
94
96
98
100
M [GeV]
Figure 6: Invariant Mass of Z boson
Number of events/GeV
Invariant Mass (Muon1 ||<1.05, Muon2 1.05<||<1.9)
70000
ATLAS Preliminary
s=7 TeV;
60000
L dt = 4.7 fb
-1
: 90.86-2.008
50000
Data
Data Fit
2/NDF: 9.639/5
MC
40000
: 90.524.33
2/NDF:
MC Fit
11.47/7
30000
20000
10000
0
80
82
84
86
88
90
92
94
Figure 7: Inner Detector
11
96
98
100
M [GeV]
Number of events/GeV
Invariant Mass (Muon1 ||<1.05, Muon2 1.9<||<2.5)
ATLAS Preliminary
s=7 TeV;
25000
: 91.182.26
20000
-1
Data
Data Fit
2/NDF: 4.212/1
MC
: 90.592.882
15000
L dt = 4.7 fb
MC Fit
2/NDF: 0.4708/0
10000
5000
0
80
82
84
86
88
90
92
94
96
98
100
M [GeV]
Figure 8: Invariant Mass of Z boson
Number of events/GeV
Invariant Mass (Muon1 1.05<||<1.9, Muon2 1.05<||<1.9)
ATLAS Preliminary
18000
s=7 TeV;
16000
14000
12000
10000
L dt = 4.7 fb
-1
: 90.483.164
Data
Data Fit
2/NDF: 7.019/1
MC
: 91.112.457
2/NDF:
MC Fit
3.557/2
8000
6000
4000
2000
80
82
84
86
88
90
92
94
96
98
100
M [GeV]
Figure 9: Invariant Mass of Z boson
12
Number of events/GeV
Invariant Mass (Muon1 1.05<||<1.9, Muon2 1.9<||<2.5)
22000
ATLAS Preliminary
20000
s=7 TeV;
18000
: 90.54.884
16000
2/NDF: 6.401/8
14000
: 91.753.26
12000
L dt = 4.7 fb
-1
Data
Data Fit
MC
MC Fit
2/NDF: 7.954/1
10000
8000
6000
4000
2000
80
82
84
86
88
90
92
94
96
98
100
M [GeV]
Figure 10: Invariant Mass of Z boson
Number of events/GeV
Invariant Mass (Muon1 1.9<||<2.5, Muon2 1.9<||<2.5)
ATLAS Preliminary
7000
6000
5000
s=7 TeV;
-1
: 90.015.844
Data
Data Fit
2/NDF: 29.51/15
MC
: 91.43.022
4000
L dt = 4.7 fb
2/NDF:
MC Fit
57.38/5
3000
2000
1000
80
82
84
86
88
90
92
94
96
98
100
M [GeV]
Figure 11: Invariant Mass of Z boson
Next we look the value comparison for different muons in different bins
and combined Lepton pT for MC and data.
13
Muon1_phi_bin1
Number of events/GeV
Number of events/GeV
Muon1_phi_bin1
15500
15000
14500
7000
6500
6000
14000
13500
5500
13000
12500
-3
-2
-1
5000
-3
3
Muon1_
-2
-1
(a) MC
3
Muon1_
3
Muon1_
3
Muon1_
(b) Data
Figure 12: Muon 1 in bin 1
Muon1_phi_bin2
Number of events/GeV
Number of events/GeV
Muon1_phi_bin2
6600
6400
6200
2900
2800
2700
2600
6000
2500
5800
2400
5600
2300
5400
-3
-2
-1
3
Muon1_
-3
-2
-1
(a) MC
(b) Data
Figure 13: Muon 1 in bin 2
Muon1_phi_bin3
Number of events/GeV
Number of events/GeV
Muon1_phi_bin3
2850
2800
2750
2700
1180
1160
1140
1120
1100
1080
2650
1060
2600
1040
2550
1020
2500
-3
1000
-2
-1
3
Muon1_
-3
-2
(a) MC
-1
(b) Data
Figure 14: Muon 1 in bin 3
14
Muon2_phi_bin1
Number of events/GeV
Number of events/GeV
Muon2_phi_bin1
16000
15500
15000
14500
7000
6500
6000
14000
13500
5500
13000
12500
-3
-2
-1
5000
-3
3
Muon2_
-2
-1
(a) MC
3
Muon2_
3
Muon2_
3
Muon2_
(b) Data
Figure 15: Muon 2 in bin 1
Muon2_phi_bin2
Number of events/GeV
Number of events/GeV
Muon2_phi_bin2
6600
6400
6200
2900
2800
2700
6000
2600
5800
2500
5600
2400
5400
-3
2300
-2
-1
3
Muon2_
-3
-2
-1
(a) MC
(b) Data
Figure 16: Muon 2 in bin 2
Muon2_phi_bin3
Number of events/GeV
Number of events/GeV
Muon2_phi_bin3
2800
2750
2700
1160
1140
1120
1100
1080
2650
1060
2600
1040
2550
1020
1000
2500
-3
-2
-1
3
Muon2_
-3
-2
(a) MC
-1
(b) Data
Figure 17: Muon 2 in bin 3
15
Now for pT comparison
(a) MC
(b) Data
Figure 18: Combined Lepton pT
0.7
Conclusions
As we can see from the Invariant mass data and MC fitting, our theoretical
predictions match the data quite well up to a certain statistical limit.
Dimuon invariant Mass (all muon )
Number of events/GeV
10
200
180
160
140
120
100
80
60
40
20
0
80
82
84
86
88
90
92
94
96
Figure 19: Combined statistics
16
98
100
M [GeV]
The comparison between the data and MC for Z 0 events are presented
here. The M distributions show reasonable agreement between the data and
MC. However in data peak and width of the Z - pole mass distribution are
slightly different from the ones predicted by the MC. Such mismatches between
the data and MC may have originated from the potential mis-modelling of multiple scattering, detector alignment etc. which require further investigation.
17
0.8
Acknowledgements
I would like to thank very heartily Dr. Prolay Mal for all his support and help in
completion of this project. I would never have been able to complete it without
his guidance and input.
18
0.9
References
The Drell-Yan Process - I.R Kenyon (Rep. Prog. Phys., Vol. 45, 1982)
ATLAS Note, ATLAS CONF-2013-088 - Prelimanry results on the
muon reconstruction efficiency, momentum resolution , and momentum
scale in ATLAS 2012 pp collision data
[Link]
[Link]
[Link]
19