1. Atlas TTbar to Tau Analysis
William P. Edson, Teeba Rashid
Adviser: Mohammad Sajjad Alam
State University of New York at Albany
Anirvan Sircar
Adviser: Dick Greenwood
Loisianna Tech
April 14, 2013

2. Contents The Large Hadron Collider (LHC)
ATLAS (A Toroidal LHC Apparatus) Detector
ttbar Production and Decay

3. The LHC
Proton–proton particle accelerator located at CERN (European Organization for Nuclear Research)
Composed of two accelerating rings with cross-over points where collisions occur (collision points).
CERN aerial view from Geneva airport (© CERN) Photograph: Jean-Luc Caron Date: Jun 1986

4. ATLAS Detector
A detailed computer-generated image of the ATLAS detector and it's systems [15]

5. Top Pair Branching Fractions for ttbar decays[5]
LHC’s high energy will cause gluons to be the major source for ttbar production (~87%)
Immediately decays due to large mass
Decay is via weak force to Wb (BR = ).[4]
W and b decay producing the particles found in the detectors
Top Pair Branching Fractions for ttbar decays[5]

6. Interaction Overview
What we see in the detector for a ttbar → τ event:
2 b-jets
Missing transverse energy (ETmiss)
the presence of a lepton (muon or electron) or hadronic jets
tau jet t W+ b
tbar W-
bbar ντ
τ -
l+, q’
νl, q

7. Tau Hadronic Decays Hadronic decays compose approximately 64.79%
Two types of jets based on the number of charged hadron tracks (π’s or K’s) [7]
1-prong (77%)
3-prong (23%)
Highly collimated resulting in a narrow particle shower recorded in the calorimeter π0 π+ π-
Example 3-prong jet

8. Boosted Decision Trees
Multivariate machine learning algorithm
Makes cuts on a given set of variables
does not discard samples which fail a cut
Trained using MC and data samples identified as signal or background
samples split into background and signal based on purity (P)
[27]
Events are assigned a score between 0 (background) and 1 (signal)
Analysis uses two BDTs
BDTe – remove electrons faking taus
BDTj – remove jets faking taus

9. Backgrounds QCD multi-jet W + jets Z + jets
reduced via ETmiss selection
estimated contribution determined via data driven techniques
W + jets
reduced via HT selection and b-tag selection
Z + jets
ttbar decay (muons and electrons)
reduced via Lepton veto cut and TauID
QCD multi-jet not modeled well in MC following τ selection criteria

10. Conclusions

11. Thanks Dr. Sajjad Alam Dr. Pat Skupic Dr. Dick Greenwood
Catrin Bernius
ttbar to tau group

12. References https://twiki.cern.ch/twiki/bin/viewauth/Atlas/WorkBook
[1] Marion Lambacher, Study of fully hadronic ttbar decays and their separation from QCD multijet background events in the first year of the ATLAS experiment. July
[2]CTEQ.
[3] V. Chekelian, Standard model physics at HERA hep-ex/
[4]
[5]Neil Collins, TopCross Section (Current Status and Early LHC Prospects). 10 March
[6]Zofia Czyczula, Search for New Physics in Tau-pair Events in ATLAS at the LHC. June
[7]Noel Dawe, Tau Identification with Boosted Decision Trees. 18 March 2010.
[8]Bjorn Gosdzik, Tau Reconstruction at the ATLAS experiment. 22 January
[9]
[10]
[11]
[12]
[13]
[14] The ATLAS Liquid Argon Calorimeter at the LHC: Overview and Performance – Mathieu Aurousseau,
[15]
[16] ATLAS Tile calorimeter commissioning status and performance – Ana Henriques,
[17] Muon Spectrometer Technical Design Report,
[18]
[19] The ATLAS Experiment at the CERN Large Hadron Collider
[20] CERN faq LHC the guide.
[21] Jane T. Bromley.,Investigation of the Operation of Resistive Plate Chambers. September
[22] Joel Feltesse, Introduction to Parton Distribution Functions. Scholarpedia. 6 November
[24] CMS collaboration, First Measurement of the top quark pair production in the dilepton channel with tau leptons in the final state in pp collisions at √s = 7 TeV. CMS-PAS-TOP-11-06
[25] Interactions.org, Particle Physics Glossary
[26] David Griffiths, Introduction to Elementary Particles Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
[27] Noel Dawe, Dugan O’Neil, Tau Identification with Boosted Decision Trees. 15 July
[28] Michael Wright, Exclusive Jet Study with Kt, AntiKt and Cambridge Aachen Jet Finders. 10 June
[29] Mark R. Leach, The INTERNET Database of Periodic Tables
[30] Arghir, S; Behar, S; Cavaliere, V; Coadou,Y ; Czodrowski, P; Flechl, M; Kopp, A; McCarn, A; zur Nedden, M ;Neubauer, M; Randle-Conde, A; Rozen, Y; Sekula, S; Takahashi, Y; Tomoto, M; Vickey, T; Xella, S, Data-driven estimation of the background to H+ searches with hadronic-tau final states. 31 March

13. Back-ups

14. LHC continued Internal pressure is kept at 10-13 atm [10]
Helps prevent collisions within beam-pipe between protons and gas molecules
1232 dipole magnets [10]
Bends particle beams around circular path
392 quadrupole magnets [10] to
Squeezes the proton bunches together to improve the probability of collision

15. Semi-Conductor Tracker
Also called Silicon Microstrip Trackers
Spatial resolution ~16μm in R-φ [19]
Covers Region |η| < 2.5 [19]
Operating Temperature ~-5 to -10˚C [19]
Cooling Gas C3F8
Eight strip layers crossed by each track
Barrel SCT Module [19]

16. Silicon Pixel Detector
Pixel Size [19]
(90%) 50 x 400μm
(10%) 50 x 600μm
Cover region |η| < 2.5 [19]
Cooling Temperature ~-5 to -10˚C [19]
Operating Gas C3F8
Typically 3 layers crossed by each track in Barrel region
Barrel Pixel Module Schematic [19]

17. How Silicon Detectors Work
Particles pass through the silicon (depletion) layer
Electrons within the silicon are knocked away (leaving a free electron and a “hole”)
Electrons and holes are pulled into contacts via an electric field produced by doping areas of the silicon
Charge builds up on the contacts producing a current E - +
n-type
p-type E + -
n-type
p-type

18. Transition Radiation Tracker
Gaseous straw tubes interleaved with transition radiation material
Diameter 4mm [19]
Intrinsic accuracy of ~130μm per straw [19]
Covers Region |η| < 2.0 [19]
Typically 36 hits per track
Operates at room temperature
Operating Gas [19]
70% Xe
27% CO2
3% O2
End-cap TRT module inner and outer ends [19]

19. Electromagnetic Calorimeter
Electrodes – three copper layers separated by polyimide sheets
Lead absorber plates
Accordion-shaped geometry
provides φ symmetry with no cracks
Covers Region |η| < 3.2 (|η| < 2.5 precision physics region) [19]
Liquid Argon:
stability of response over time
intrinsic radiation hardness
intrinsic linear behavior
Barrel Electromagnetic LAr Calorimeter [19]

20. Hadronic End-Cap Calorimeter
32 identical wedge-shaped modules per wheel
Gap between copper plates 8.5 mm [19]
three electrodes per gap
Covers Region 1.5 <|η| < 3.2 [19]
Outer wheel radius 2.03 m [19]
Inner wheel radius [19]
0.475 m
in overlay region with FCal m
Shares End-Cap cryostat with EMEC and FCal
Liquid Argon chosen instead of tile due to high radiation levels in End-Cap
Hadronic End-Cap Module [19]

21. Hadronic module absorber matrix [19]
Forward Calorimeter
Metal absorber matrix interposed with copper tubes centered around metallic rods
Three modules per End-Cap:
copper rods – EM measurements
tungsten rods – hadronic interaction energy
tungsten also
Covers Region 3.1 < |η| < 4.9 [19]
copper – optimizes resolution and heat removal
tungsten – provides containment and minimizes spread of hadronic showers
Hadronic module absorber matrix [19]

22. How the Tile Calorimeter Works
Particles induce the production of ultraviolet light in the scintillation material
Doping material (fluors) shifts the wavelength to that of visible blue light
Wavelength-shifting fibers collect the light for the photomultipliers
Tile calorimeter module [19]

23. Cross-section and longitudinal view of MDT [19]
Monitored Drift Tubes
Chambers of three to eight layers of drift tubes
Average Resolution of 80 μm per tube (30 μm per chamber) [19]
Covers Region [19]
|η| < 2.0 innermost end-cap layer
|η| < 2.7
Operating Gas Ar/CO2 (93/7) at 3 bar [19]
Maximum drift time 700 ns [19]
Cross-section and longitudinal view of MDT [19]

24. Cathode Strip Chambers
“X” - Strips
Wire
Fiducial Mark
Gas Seal Bar
Seal Rubber
5 Composit Panels
Lower Density Foam
“X” – Readout Connector
Wire Fixation Bar
Assembly Holes
Gas Inlet (Outlet)
“Y” - Strips
Multi-wire proportional chambers with 2 strip segmented cathodes
perpendicular strips
parallel strips
Combine high spatial, time and double track resolution with high rate capability
Cover Region 2 < |η| < 2.7 [19]
Operating Gas Ar/CO2 [19]
Four η and φ measurements per track
CSC Design [19]

25. Resistive Plate Chambers
Gaseous parallel electrode plate detector
Electric field between the plates 4.9 kV/mm [19]
Covers Region |η| ≤ 1.05 [19]
Operating Gas Mixture [19]
C2H2F4 → 94.7%
Iso-C4H10 → 5%
SF6 → 0.3%
Low-pT (6 - 9 GeV) trigger signal
hits inner two RPC layers
High-pT (9 – 35 GeV) trigger signal
hits in all three RPC layers
Cross Section of an RPC [19]

26. TGC triplet cross-section [19]
Thin Gap Chambers
Multi-wire proportional chambers in either doublet or triplet form
Four layers each end-cap
Covers Region 1.05 ≤ |η| ≤ 2.4 [19]
Distances [19]
wire to cathode 1.4 mm
wire to wire 1.8 mm
Operating Gas mixture CO2, n-C5H12
TGC triplet cross-section [19]

27. Proton Interactions
The primary interaction for ttbar production is the strong interaction.
As the energy of the reaction increases the amount of gluons within the proton (the sea) increases.
Gluon distribution at different energies where x is the momentum fraction carried by the gluon within the proton.[3]

28. Anti-kt Jet Algorithm Clustering algorithm
Distances are calculated between particles or pseudo-jets (dij) and between particles and the beam (diB)
if dij minimum and less than some dcut, then i and j are combined
if diB minimum, then i is labeled jet and removed from list
Highest pT objects grouped first
Depiction of the clustering conducted by the anti-kt algorithm
dij = min (1/p2Ti, 1/p2Tj)∆R2ij/R
diB = 1/p2iB

29. 1-prong Cut Flow
Cut Flow for 1-prong ttbar to tau + muon analysis with statistical uncertainties [23]

30. 3-prong Cut Flow
Cut Flow for 3-prong ttbar to tau + muon tau analysis with statistical uncertainties [23]

31. Systematic Uncertainty Determination
Scale Factors
selection event variation
Muon momentum scale and resolution
reconstructed μ pair invariant mass distributions of Z/γ*
Jet Energy Scale (JES)
combined information from test-beam, LHC collision and simulation
Jet Reconstruction/Identification
comparison between simulation and data
Initial and Final-State Radiation (ISR/FSR)
AcerMC generator interfaced with PYTHIA shower model
varying parameters in range consistent with data
MC Generator
compare predictions with those from POWHEG
PDF
range of current PDF sets
b-tag scale factor
calibration studies using inclusive lepton and multi-jet final states
TauID
compare MC and data Z → ττ events with our selection requirements except Njets < 2 and MT < 20GeV
Luminosity
TauID difference in selection to remove W + jets events

32. The Inner Detector Inner Detector is composed of three subdetectors:
Silicon Pixel Detector:
precise measurement of vertices
Semi-Conductor Tracker:
measures particle momentum
Transition Radiation Tracker:
electron identification and path tracking
Detailed computer generated view of the ATLAS Inner Detector [13]

33. Liquid Argon Calorimeters
Liquid Argon Calorimeters (LAr Calorimeter)
Electromagnetic Calorimeter (EMCal):
precise measurements for photons and electrons (also used in their ID)
Hadronic End-Cap Calorimeter (HEC):
precise measurements for hadronic jets, as well as Missing Transverse Energy (ETiss).
Forward Calorimeter (FCal):
electromagnetic and hadronic energy measurements
Tile barrel
Tile extended barrel
LAr hadronic end-cap (HEC)
LAr electromagnetic end-cap (EMEC)
Lar electromagnetic barrel
LAr forward (FCal)
Detailed computer generated view of the ATLAS Liquid Argon Calorimeter [14]

34. Tile Calorimeter Doped polystyrene used as scintillating tiles
Steel absorber plates
Over 460,000 scintillating tiles [19]
Covers Region |η| < 1.7 [19]
Measurements for both ETmiss and hadronic jets
Jet information (from HEC, FCal and Tile Calorimeter) is used for jet reconstruction and identification
doped – 1.5% PTP primary fluor, 0.044% POPOP secondary fluor
Detailed computer generated view of the ATLAS Tile Calorimeter [16]

35. Muon Spectrometer
The Muon spectrometer is constructed of four different chamber types:
Monitored Drift Tubes (MDTs)
precise measurements of track position
Cathode Strip Chambers (CSCs)
Resistive Plate Chambers (RPCs)
triggering and measurement of track coordinates
Thin Gap Chambers (TGCs)
same as RPCs but in end-cap region
3-D view of the muon system [17]

36. ATLAS Magnet System Three superconducting magnet sub-systems:
Barrel Toroid
3-8 Tm over the central region
End-cap Toroids:
3-8 Tm over the forward regions
Solenoid:
2 Tesla within the central tracking volume
Computer generated view of the ATLAS Magnet System [18]

37. Standard Model Depiction (28)

38. Quarks Elementary particles which come together to form hadrons
Fermions (have spin ±1/2)
Color charged (red, green and blue)
Fractional elementary charge of +2/3 or -1/3
Have associated antiparticles with the same mass and spin but opposite charge
Antiparticles noted with the use of a bar above their symbol or following their name (tbar).
Top Quark (t)
Most massive of the known quarks (172.0±0.9±1.3GeV/c2 [4])
Third generation of quarks (with bottom)
Charge of +2/3 [4]

39. Leptons
Elementary particles that are similar to quarks but vary in some very drastic ways.
Fermions
Do not carry color charge
Integral elementary charge (-1 or 0)
Have corresponding antiparticles
Antiparticles noted by adding a + signifying their positive charge (e+) or using a bar above (υe)
Taus (τ)
Most massive of the known leptons ( ±0.16MeV/c2 [4])
Third generation of leptons
Charge of -1
Each Lepton generation includes a corresponding neutrino with neutral charge and extremely small mass (~0).

40. Particle Interactions and Gauge Bosons
Three major interactions and their mediating particles (the gauge bosons):
Strong force
Only occurs between color charged particles
8 gluons
massless and elementary chargeless
doubly color charged
Electromagnetic force
Only occurs between elementary charged particles
photon
Weak force
Occurs between all quarks and leptons
intermediate vector bosons W±, and Z
Have mass and the W carries charge

41. Protons Most stable baryons which exist
Positive elementary charge of +1
Mass of ± MeV/c2 [4]
First generation quarks uud
Parton distributions from CTEQ6M plotted at Q = 100 GeV. x is the fraction of proton momentum carried by the particle, and f(x) is the probability density. [22]

42. Cross Section
Measure of the likelihood that some specific interaction will occur
Units barns (b) ~ m2 [25]
Theoretically:
σ = ∫dσ = ∫|b/sinθ(db/dθ)|dΩ [26]
where b is the impact parameter (distance traveling particle moves in relation to the scattering particle’s center), θ is the polar angle and dΩ is the solid angle
Equation used for measuring:
σ = NEvents/L
where NEvents are the number of events while L is the measured luminosity (number of particles per unit time per area) b dΩ

43. Tau Importance
Tau is an expected final state of previously undiscovered particles both within and beyond the Standard Model (SM):
Higgs Boson: H → τ+τ-
extra gauge bosons: Z’ → τ+τ-
Minimal Supersymmetric Standard Model (MSSM) Higgs (charged Higgs): H+ → τ+ντ H τ+ τ Z’ τ+ τ

44. ttbar Decays Continued
b quark decays
form jets which can be readily identified
W decays
quarks (seen as jets)
leptons and neutrinos
Top Pair Branching Fractions for ttbar decays[5]

45. ttbar Production
LHC’s high energy will cause gluons to be the major source for ttbar production
87% by gluon fusion [1]
~13% mainly by qqbar annihilation [1] g t g
tbar g g t g t
tbar g
tbar g q t g q
tbar

46. Tau Decays
μ-νμντ 17.36±0.05% e-νeντ 17.85±0.05% π-ντ 10.91±0.07% π-π0ντ 25.51±0.09% π-2π0ντ 9.29±0.11% π-3π0ντ 1.04±0.07% h-h-h+ντ 9.80±0.08% h-h-h+π0ντ 4.75±0.06%
Extremely short lifetime of (290±1)*10-15s [4] due to large mass
Decays via the weak interaction
Reconstructed from the daughter particles of its decay
Main branching ratios of tau decays[4]. h± can stand for either K or π and

47. Tau Decay Hadronic decays compose approximately 64.79%
Leptonic decays are almost impossible to distinguish as being produced via the tau decay versus any other process τ- W- ντ l νl τ- W- ντ d
ubar

48. Tau Reconstruction
Calo-seeded: Begins with the detection of a particle shower within the calorimeter and then searches for associated tracks within a narrow cone.
Track-seeded: Selects a good quality track and then looks for other tracks around the seed track within an isolated cone. The energy deposited in the calorimeter within this cone is also collected.
Calo shower
Calo seed
Seed track
Track cone
Isolation cone

49. ttbar to tau + mu t W+ b tbar W- bbar ντ τ μ+ νμ t W+ b tbar W- bbarτ+ ντ

51. BDTj Template Fit Results corresponding to a, b and c [23]
BDTj Contd
χ2 fit of OS – SS BDTj distribution determined:
(a * OS – b * SS) + c * Signal [23]
OS, SS and Signal → templates
a, b and c → free parameters to set the normalization of each template, also equivalent to the resultant number of events
OS and SS templates weighted using scale factors determined from MC
BDTj Template Fit Results corresponding to a, b and c [23]
Weighting of OS and SS necessary due to the pT spectrum of tau candidates being different between ttbar and W + jets events

52. OS – SS Plots
BDTj (OS – SS) before ETmiss selection (a) τ1 (b) τ3 [23]
BDTj (OS – SS) prior to b-tag selection (a) τ1 (b) τ2
BDTj (OS – SS) following b-tag selection (a) τ1 (b) τ3

53. Signal Template Composition derived from MC [23]
Remaining Signal
Signal Template Composition derived from MC [23]
Contamination remains from:
Z → ττ
ttbar → μ + e
Small predicted contributions are subtracted from signal prior to cross section calculation

54. ttbar to tau + jets t W+ b tbar W- bbar ντ τ qbar q’ t W+ b tbar W-τ+ ντ

55. To Do Determine the Sources of possible background
Decide on the necessary MC samples
Determine the Event Selection
Make plots of the Data
Determine how you will estimate the Background
Plot the Background using MC
Estimate the Uncertainty caused by Background Estimation
Estimate the Uncertainty caused by Event Selection Criteria

56. Background QCD multi-jet Single-top W → μ + jets W → e + jets
Z/γ* → μ + jets
Z/γ* → e + jets
Light jets (u and d)