The document discusses energy flow and clustering algorithms used to reconstruct physics objects in the ATLAS experiment at the Large Hadron Collider. It first provides background on the LHC, ATLAS detector, and jet physics. It then describes an energy flow algorithm that combines calorimeter energy measurements with tracking information to improve jet energy resolution. This is important because two-thirds of a jet's energy is from charged particles. The document also discusses using Atlfast software to simulate jet reconstruction and analyze particle composition within jets using different clustering algorithms and radius parameters.

1. Energy Flow and Clustering
algorithms for the reconstruction
of physics objects in ATLAS
Tesis Doctoral
Dpto. Física Atómica, Molecular y Nuclear
Carmen Iglesias Escudero

2. OUTLINE
 LHC and ATLAS
 ATLAS Calorimetry
 Jet Physics in ATLAS
I. Energy Flow algorithm in ATLFAST
 Underlying Events, Minimum Bias & Pile Up
II. Clustering Algorithms for VLE particles
(simulated)
III. Clustering Algorithms for VLE data of
Combined TB

3. LHC and ATLAS
 LHC Physics
 LHC Setup
 LHC Experiments
 ATLAS
 Particle detection

4. LHC Physics
The LHC will allow to explore the structure of matter at energy frontier and
at the energy density frontier.
 The physical origin of electroweak symmetry breaking and the origin of
mass
 Higgs boson
 The physical origin of CP violation
 Unitary triangle
 Searches beyond the standard model
 supersimmetry, new gauge bosons, compositeness,…
 Precision measurements of Standard Model parameters
 Top. Beauty, tau, QCD,…
 The physics of strongly interacting matter at extreme energy densities
 quark-gluon plasma

5. LHC Setup
CMS
LHCb
ALICE
ATLAS
Over 1000 superconductive 8.36 Tesla
(at 1.9 Kelvin) dipoles are needed to
bend the 7 TeV protons in the 27 Km
LHC circumference

6. Detector dedicated to the study of General purpose detectors which will be
heavy ions. Focused in the study of the p-p interactions.
They will be used in further test of SM (Higgs
Boson search) and in new physics search
ATLAS
(supersimmetries, extra dimensions…).
Detector dedicated to the study of
B-physics (CP violation)

7. ATLAS (A Toroidal LHC Apparatus)
 The design considerations for ATLAS detector are:
 good EM-calorimetry for e, γ identification and measurement.
 Hermetic jet and Emiss calorimetry.
 Efficient tracking at high luminosity for lepton measurements, b-quark tagging
and e, γ identification.
 τ and heavy flavour vertexing and reconstruction capability of some B decays.

8. Particle Detection
The photons and electrons deposit almost all their
energy in EM Calorimeter
The hadrons deposit their energy in HAD
calorimeter
The muons as has little interaction with the matter,
arrive until the spectrometer
The moment from charged particles is measured
from the curvature of the tracks in the inner
detector
Each layer identifie and measure the
energy non defined in the previous one
Only one detector can not measure
the Energy/momentum of all particles

9. ATLAS Calorimetry
 Electromagnetic shower
 Hadronic cascades
 EM calorimeter
 Hadronic Calorimeter: TileCal
 Physics issues for Calorimetry
 Energy resolution

10. Electromagnetic showers
 A high energy e or γ initiates a cascade of e
and γ’s via
 bremsstrahlung and
 pair production
until they fall below critical energy Ec
 Characteristic length X0≡ radiation length
Mean distance in the absorver over wich a high-energy
e- reduces its energy by a factor 1/e only due to bremst.
Shower can be fully measured or sampled.
 Needs a depth of > 25 X0to contain a high
energy em shower
 The lateral development is governs by the
Moliere Radius (average lateral deflection
of critical energy electrons after 1 X0).
RM = X0/EC

11. Hadronic cascades
 Similar to em shower but with
strong interaction responsible for
cascading effect :
 Multi-particle production (π0,
π±, K etc..)
 nuclear break up until π
production threshold
 Characteristic length λ≡nuclear
interaction length
Mean distance between inelastic
collision of hadrons with nuclei
 About 10λ necessary to contain
99% of energy of 200 GeV pion
 High pt quarks/gluons hadronize
giving narrow JETS

12. EM Calorimeter
Provide a very precise energy reconstruction of e- and γ
Powerful tool for the particle identification due to its high granularity
Accordion geometry benefits: No cracks in ϕ
The detection element is liquid Argon.
The EM shower emit electrons in the
Argon which are collected and register.

13. Hadronic Calorimeter: TileCal
Extended Barrel Sampling calorimeter:
modules - Scintillators (active mat.)
Barrel - Iron (absorber mat)
modules
64 modules
The tiles are
placed in the
perpendicular
plane to the
beam axis and
the read out is
performed by
optical fibres and
routing them to
the PMTs.

14. Physics issues for Calorimetry
ATLAS calorimetry: Crucial role
at the LHC:
Detectors are required to measure
the energy and direction of:
photons and electrons
isolated hadrons and jets,
the missing transverse energy
(ET).
Electromagnetic calorimeter Hadronic calorimeter
• Rapidity coverage up to |η|=5.
• Dynamic range: From few MeVs to TeVs • Energy resolution:
• Good energy resolution:
• Good electron/jet and γ/jet separation • Linearity better than 2% up to 4TeV.
• High granularity : • Granularity
– At least ∆ηx∆φ=0.03x0.03 for |η|<2.5 • ∆ηx∆φ =0.1x0.1 for |η|<3 ∆ηx∆φ =0.2x0.2 for 3<|η|<5
– Longitudinal segmentation for particule ID • Jet tagging efficiency > 90%
• Tolerance to radiation • Tolerance to radiation

15. Energy Resolution

16. Jet Physics in ATLAS
 Jet definition
 Fragmentation
 Initial parton to jet
 Hard scattering and Underlying Events
 Jet measurement

17. JET definition
Jet : Group of energetic particles which
are emitted spatially collimated.
Jets are manifestations of scattered
sub-nuclear 'partons' (quarks & gluons)
so due to partons cannot be isoleted,
jets gives information about them.
A jet constains mainly hadrons: tens of
neutral and charged pions, a lesser extent
of kaons and very few light baryons
(such protons and neutrons)

18. Fragmentation
 Hard Scattering
 Elementary hard process: p-p interaction produces fundamental objets: quarks
and gluons (they can be seen as free particles).
 Parton shower: primary partons generate a shower
of partons because color forces will organize them
into colorless hadrons involving the creation of
many quark-antiquarks pairs.

 Hadronization
 Hadronization: parton shower is transformed into the observed set of short-life
hadrons. Phenomenological models are used.
 Decay of unstable primary particles into stable hadrons and leptons according
to the lifetimes and braching ratios for each unstable particle.

19. Initial Parton to Jet
The definition of a jet is not unique and the corresponence between parton energy
and direction and measured jet characteristic is influenced by many factors: parton
fragmentation, FSR, Underlying Events, detector response and by the jet algorithm

20. Hard Scattering & Underlying Events
 The 'Hard Scattering' components consists of
the outcoming two 'jets‘ which come from a hard
2 parton scattering which interact at short
distance with large pT transfer.
 The ‘underlying events’ is everything except the 2 hard scattered jets and consist of:
-the beam-beam renmants: because protons are not
elementary particles bur are formed by 3 quarks.
- ISR and FSR: interaction between quark and gluons
before and after the hard scattering.
- multiple interaction: a second, a third parton
scattering...softer than hard scattering
 Finally, in high luminosity, it is possible to have several collision between beam particles in
the same beam crossing, ie, pile-up events.

21. Jet Measurement
 Each jet is characterized by :
 a charged fraction: mainly π±
 a neutral electromagnetic fraction:
mainly photons from π0γγ decays
 a neutral hadronic one: mainly KL
and neutrons.
 The calorimeter is segmented in ϕ (azimuthal ang.)
and η (pseudo-rapidity).
 Jets are observed as
Jets used to be reconstructed
clusters of ET located
with a cone centered in the
in adjacent cells with
cell with max ET and a radius
0.1x0.1 in η-ϕ
R= √∆η2+∆φ2 around the
center (usually R=0.4-0.7)

22. I. Energy Flow in
ATLFAST
 Energy Flow algorithm
 Overlapping
 Resolution in ATLFAST
 Jet Generation (Pythia) and Reconstruction (Atlfast)
 Particle composition of the jets
 Analysis by Cell
 Applying Energy Flow

23. Energy Flow Algorithm
 About 2/3 of the jet energy are carried by charged particles (p±,K±...)
However jet algorithm makes no use of tracking information
 Energy Flow algorithm make an optimal use of the detector information combining
the measurement of the energy deposition in calorimeter cells with the reconstructed
track in the inner detector to improve jet energy resolution and ETMiss.
Introduced first by LEP experiments .
 For low momentum charged particles, the tracking error is much smaller than the
calorimetric energy error. In example,
for the Central Barrel in ATLAS (η=0):
Track: σpT/pT = 0.036%pT⊕1.3%
Cal: σE/E = 50%/√E⊕3%
where pT and E are in GeV. We can see, i.e
for one π± of 10 GeV E resolution is 16 %
while for PT is 1.3%.
Energy Flow must be applied at pT<140 GeV.
 So for charged particles, their energy resolution will be sustituted by the track
momentum resolution  better resolution in jet ET.

24. Energy Flow: Overlapping
 The use of the track momentum improves the resolution only works if cluster is
isolated. If the track shares a cluster with a neutral particle, the gain in resolution
from track will be limited by loss of resolution from remaining cluster.
 Efficiency of algorithm is limited by the overlapping between neutral and
charged particles in the cell of the calorimeter. We need to know more about this
effect and its influence in the analysis
 Typical multi-jet event :
 64% charged energy
 25% photons
 11% neutral hadron

25. Resolution in Atlfast
 ATHENA: Framework of ‘offline’ Software in ATLAS
 Atlfast: C++ Object Oriented implementation which provides a fast particle-level
simulation of the detector response and its later reconstruction, and allow:
 define the 4-momentum of the particles
 reconstruct clusters and jets inside the calorimeters
 characterize the tracks
In Atlfast  no detailed simulation of particle shower
neither of the tracks in the inner detector
only a parametrisation of calorimeter E resolution
and a simulation of efficiency and Pt resolution in Si detector.
Parametrisations were derived from Full Simulation studies:
EM Cal resolution HAD Cal resolution Si Detect resolution
( γ and electrons) (hadrons :π± and k±) (track of e ±, µ ± , π± )
0.245/√Pt ⊕0.007 at η<1.4 0.5/√Pt ⊕0.03 at η<3.2 0.0005(1+ η10/7000)Pt ⊕0.012
0.306((2.4- η)+0.228) /√Pt ⊕0.007 η>1.4 1.0/√Pt ⊕0.07 at η>3.2
Effects as overlap of particles inside the cell can be studied by Atlfast,
HOWEVER when the influence of the shower is relevant  Full Simulation.

26. Generation with PYTHIA 6.2
Generate 1000 events of QCD jets, applying in Pythia the next conditions:
- for differents range of PT:
20-40, 40-80 , 80-160, 160-320, 320-640 and 640-1280 (GeV)
- Without include Underlying Events and Minimum Bias effects
- ISR and FSR are taken into account
- |ηparton| < 5.0, to use only parton insider the calorimeter coverage
Jet Reconstruction with Atlfast
Release 6.2.0 is used for the reconstruction of QCD jets:
- Cone algoritm is used with different values of radius R=0.4 and 0.7
- |ηjet| < 2.0, to ensure the completed containment of the cone jet
inside Inner coverage (calo+track info used later)
- Minimum Pt of the jet, to prevent excessive merging of noise and energy
not associated with hard scattering. Different values depending on R
(multiplicity of jets still significant)
Ptmin=20GeV if R=0.7
Ptmin=15GeV if R=0.4

27. Particle composition of jets
To reconstruct jet ET from particle energy into the cone,
we select:
 only stables particles deposited in Calorimeter
 mainly charged hadrons (π ± and k ± )
 Similar ammount of photons (from π0γγ)
 a too lesser extend of neutral hadrons (kLO & n)
 and very few leptons (e ± ,µ± and ν)
 ET>0.5GeV for charged particles
 |ηpartc| < 2.5, only particles inside inner coverage
R=0.4
R=0.4 Multiplicity
Et deposited by particles R=0.7
R=0.7
So, Charged had Neutral had Photons
2 important Charged had had Neutral had Photons
TotalforCharged hadhadron we have Photons
charged Neutral had Total Charged Neutral had
results: Photons
in jet per jet (%) per jet (%) perjet (%) inper jet per(%) (%) jet jet (%) jet jet (%)
jet
1) Their number is ~ 47% of the total particles
per jet (%) per jet (%) per jet (%) jet per per (%) per per (%)
40-80 13.2 2) 22.6
40-80 Their deposited ET7.1 ~ 61% of45.5 total energy
6.2 61.2
46.6 0.9
4.6 is
12.5 6.0
9.2 the 24.15 6.4
25.2 13.4 61.1 46.6
4.88 0.9
12.4 7.0 9.2 6.0 25.245.5
80-16080-160
17.2 8.2
40.3 47.1
61.3 1.1
7.8 6.4
11.8 7.9
16.9 45.7
25.6 17.7
42.62 8.4
61.3 47.1
8.19 1.1
11.8 6.311.7 8.2 25.745.7
160-320 Energy10.0 61.4applied
Flow 47.3 13.1
20.9 69.1
160-320 is 1.3 to 11.9 charged hadrons, BUT not to 13.98 1.3 to the 9.9 25.745.7
the 28.9 45.7 21.7 10.3 47.3 only 6.130.7
6.1 9.6 25.7 73.50 61.4 all 11.7
mainly charged hadrons andcell without sharing with neutral particles,
charged hadrons which hit photons
ET deposited by particles increase as the ET of jet is bigger
the ammount of leptons ishad (2/3 parts), it is ∼double that photons ET
most of ET from charged negligible (<0.5%)
Et per jet in R=0.7 is bigger with the
Number of particle increasethan 0.4 E

28. Analysis by Cells
a) define the calorimeter CELL that the particles hits
Grid of 81 cells with 0.1 x 0.1 granularity in η-φ plane around deposition point of jet
b) classification of the cell based on the type of particle
(charged or neutral) that fell in it
CHARGED CELLS: only charged partic (π ± and k ± )
NEUTRAL CELLS: only photons
MIXED CELLS: mixed charged and neutral particles
in this last case it’s analyzed the overlapping between
charged and neutral particles

29. ET deposited in cells
Et jet Charged Cells Neutral Cells Mixed Cells
(GeV)
per jet (%) per jet (%) per jet (%)
40-80 35.50 16.3 45.8 6.7 18.9 12.6 35.3
80-160 65.94 21.8 33.8 8.7 13.4 35.3 54.6
160-320 94.20 23.7 25.2 9.6 10.2 60.7 64.4
Up to 45% of total ET, in the best case, come from charged had in Charged
cells. For this ET a gain in resolution will be done by Energy Flow
This proportion decrease quickly with the jet ET, as the same time as the
energy in Mixed Cell increase.
So, the overlapping will be bigger with the E, and the gain in resolution applying
Energy Flow will be worse.

30. Improvement in ET of the jet
(Range 40-80GeV and DR=0.4)
Aplying HAD Cal smearing
to the CHARGED CELLS:
0.5/√Pt ⊕0.03 at η<3.2
resolution in the jet energy
~8%
Aplying INNER smearing
0.0005(1+ η10/7000)Pt ⊕0.012 at η<2.5
resolution in the jet energy
~4.8%
much better result than with HAD Cal
Resolution of the jet energy have been improved in ~40%

31. Variation of gain in resolution
RMS RMS (%)
R=0.4 HAD INNER
40-80 0.079 0.048 39.0
80-160 0.062 0.042 31.0
160-320 0.051 0.039 23.6
320-640 0.041 0.034 16.9
640-1280 0.032 0.029 9.6
RMS RMS (%)
R=0.7 HAD INNER
40-80 0.076 0.049 35.7
80-160 0.062 0.043 30.7
160-320 0.049 0.039 20.4
320-640 0.039 0.033 16.6
640-1280 0.031 0.029 9.5
- Very optimistic result: high gain in resolution using Energy Flow at low Pt~40 %
- The improvement decrease with E.
- At few 100 GeV the overlap of particles gets higher and the gain in resolution is marginal

32. Underlying Events,
Minimum Bias & Pile Up
 Soft physics processes
 The Underlying Event
Multiple Scattering with Pythia
 Influence in the multiplicity
 Ocupancy and Density
 Applying Energy Flow
 Minimum Bias Event and Pile-Up
Number & ET of particles

33. Soft physics processes
There is no observable high-pt signature
Physically a combination of several physical
processes: mainly non-diffractive inelastic double
Minimum diffractive
Experimentally depends on the experiment-trigger:
bias Collider expts usually measure non-single
diffractive(NSD)
Soft physics
Underlying
event Associated with high PT events:
Beam remnants
ISR
More difficult to define experimentally and
theoretically

34. The underlying event
•Underlying event is everything
High PT scatter
except the two outgoing hard
scattered jets.
Beam remnants
ISR
•In a hard scattering process, the underlying event has a hard component
(initial + final-state radiation and particles from the outgoing hard scattered
partons) and a soft component (beam-beam remnants).

35. Influence in the multiplicity
When we add Underlying events :
- Increase the multiplicity of charged hadrons (10%)
Mean=7.0 Mean=7.8
- Increase the multiplicity of photons (14%)
Mean=7.1 Mean=8.1

36. Occupancy and Density
OCCUPANCY: number of particles which
hit in each cell with a granularity 0.1x0.1.
PTcut: ET>0.5GeV for charged particles
The occupancy is more than the double when we consider UE+QCDjets
DENSITY: number of particles which
hit in ∆η =1, i.e. dN/dη
When we apply the pT cut to charged hadrons
the density decrease.
UE Density = dN/dη ∼ (38-15) = 23 Similar results than previous analysis
A. Moraes studies:
ATL-PHYSICS-2003-020

37. Applying Energy Flow
Only QCDjets QCDjets+UnderlyingEvents
RMS(Had)=0.079 RMS(Had)=0.079
RMS(Inner)=0.048 Similar results: RMS(Inner)=0.049
Gain(%)=39 Underlying can be negligible Gain(%)=38

38. A minimum bias event
Forward production
Low multiplicity
Large Enegy
Central production
High multiplicity
Small Energy
MB consists of 4 processes: non diffractive, single diffractive,
double diffractive and elastic Most popular models takes MB
Events as non-diffractic inelastic.

39. Minimum Bias Events
Multiplicity of particles in Δη
~ 7 charged partc/Δη
~ 8 neutral partc/Δη
Similar results to the shown
in Calorimeter Performance of
ATLAS and TDR
Pile-Up Events in PYTHIA
Pile-up events are taken by PYTHIA to be of the MinimumBias type.
PYTHIA can generate several events and put one after the other in the event record,
knowing the assumed luminosity per bunch crossing expressed in mb-1.

40. Number of
Particles QCDjets+UE
QCDjets
Pile Up
Min bias
Although at occupancy level Pile Up at low luminosity is of the order of QCDjets,
the ET deposited by Pile-up Events is much smaller than the come from jets
ETof
Particles QCDjets+UE
with cut ET>0.5 QCDjets
for charged had
Pile Up
Min bias
So if we applied Energy Flow, the influence of the Pile-up events at low
luminosity can be negligible.

41. Conclusions
 The application of the Energy Flow algorithm at particle level in
ATLAS can potentially improve the jet energy resolution.
 This improvement is better at lower pT reaching values up to ∼40%
of relative gain in resolution. Nevertheless, around 100 GeV the
overlap between particles is higher anf the gain in resolution of the
jet energy is marginal.
 Respect to the soft process, the influence of the Underlying Events
and the Pile-Up events at low luminosity can be neglegible for
Energy Flow resolutions.

42. II. Clustering Algorithms
for VLE particles
(simulated)
 Why clustering algorithms…?
 Samples used
 Clustering algorithms in ATLAS
 TopoCluster analysis:
 EM Noise
 Lower threshold for Seed and Neighbor cells
 Cone algorithms
 Clustering comparison
 TopoCluster with electronic Noise

43. Why Clustering is useful for EFlow?

44. Samples used
 DC1 samples of pions and neutrons (the main components of jets)
at very low ET (pT =1-30 GeV).
 Used to generate ntuples with 1000 events at η=0.3 (central barrel) and φ=1.6 of :
 π’0s, to understand the behavior of photons inside the EM calorimeter.
 π’+s and neutrons, to know more about the hadronic shower.
First, without electronic noise applied and later with it.
Shower composition
The shower of the π0 has only e.m. components!!!
neutron
electrons photons
π0 e-and q γ
π0
positrons proton
π-
π+

45. Total energy deposited
 For the π0’s, as there are only e.m. particles we expect
having all the ET deposited in the E.M calorimeter
 For π+’s and neutrons the situation is different. Although,
at high pT their ET is usually deposited only in HAD calo,
at very low energy, they also deposited their ET in EM calo.
This deposition decrease with the ET of the particles.

46. Clustering Algorithms in ATLAS
 Sliding Window (SW) Clustering
 Simple search for local maxima of ET deposit on a grid using a
fixed-size “window” of adjacent cells in η-φ space.
 Default value is 5 x 5 cells in each cluster. Another values:
 3x5 cells (unconverted photons)
 3x7 cells (e- and converted γ).
 EGAMMA Clusters
 Combines Inner detector tracks information
with calorimeter clusters (SW) using 5 x 5 cells for cluster
 Useful for the identification of the e.m objects
(photons and electrons).
 TopoCluster Algorithm
To reconstruct hadronic shower, the ET depositions from closed cells is merged to clusters
Seed Cell
Cluster is built around a Seed Cell which has an ET
phi
above a certain threshold (Seedcut). The neighbours are
scanned for their ET and are added to the cluster if this
ET is above the neighborcut. Then the neighbors of the
neighbors are scanned and so on.
The cuts depend on the noise in each cell eta
Neighbour Cell

47. Clustering comparison
First, calculate the ET deposited in all CELLs of the calorimeter and consider it as the
“reference Energy Flow”, i.e., the best resolution that could be reach for the most
sophisticated algorithm taking into account the whole ET in all the calorimeter.
 For π0’s, compare the resolution of “reference Energy Flow” with the
resolution of:
 Sliding Window Cluster/EGAMMA cluster
 TOPOcluster in EM calrim
 ∆R cone around seed
 For neutrons, compare the resolution of “reference Energy Flow” with :
 TOPOcluster in EM and Tile
 ∆R cone around seed
 For π+’s, compare the resolution of “reference Energy Flow” with :
 TOPOcluster in EM and Tile
 ∆R cone around seed
 PT of TRACKS from XKalman

48. TopoCluster Analysis: EM Noise
Compare different ways of reconstructing TopoCluster at VLE particles, to find:
the best ET resolution
the largest amount of ET deposited inside the cluster.
 Use these thresholds:
And checking different thresholds for EM Noise:
 EM Noise=10 MeV (lower than realistic case, only useful for checking
VLE particles)
 EM Noise=70 MeV (Fix Value by default for EM cal)
 CaloNoiseTool=true (package with a model for the electronic noise)

49. π+’s resolution
•Resolution from pT of TRACKS
is the best result, but it get worse
as the ET of particle increases.
•Respect to the calorimeter ET,
the best resolution comes from
the ET deposited in all calo cells.
•Around 30 GeV, ET resolution
get better than pT resolution 
limit of Energy Flow algo
neutrons resolution
The worst result is at 1 GeV:
•ET very similar to the mass of
neutron~940MeV.
 For the TOPOclusters CaloNoiseTool is the most realistic simulation of Electronic Noise.
The rest of the analysis will be done using it.

50. π0’s resolution
 π0’s have better resolution
than π+’s and neutrons
 For Sliding-Window clusters,
always are obtained the same
results as EGamma.
 Best result for all calo cells,
and next for EGamma cluster.
• For all TopoClusters at 1, 3 and 5 GeV their multiplicity is very low. Results have
non-sense -> ET resolution increase instead of decreasing with ET.

51. Lower threshold for TopoCluster
 Loss of ET deposited in TOPOcluster due to the low multiplicity of these clusters
It’s needed to move for lower threshold for Seed and Neighbor cells:
 Seed_cut: E/σ= 30  6, 5, 4…
 Neigh_cut: E/σ= 3  3, 2.5, 2…
The low efficiency of TopoClusters has been practically eliminated, mainly in π0’s case.
The worst results is for neutrons at 1 GeV, but it also improves with the changed cuts.

52. For π+’s and neutrons, the best
resolution for TOPOcluster using
Seed_cut=4 and Neigh_cut=2.
The TOPOcluster resolution is
more similar to the resolution of the
ET deposited by all calorimeter cells
For π0’s, the resolution of TOPOclusters using
any of these new cuts is even better than the
resolution of EGamma.

53. Deposited Energy
For π+’s and neutrons,
changing the Seedcut from
30 to 4, a large increase in
the deposited ET is
obtained, mainly at 1-5 GeV
(the ET is almost the double)
For π0’s, with the new cuts, the
Values of deposited ET for Topo
are very similar to the EGamma
one and competitive respect to
the ET in all the cells.

54. Cone algorithms
 The ET of the clusters is reconstructed from the ET of all cells inside
a cone with a radius ∆R=√∆η2+∆φ2
 Different strategies are followed for the different type of particle
Neutral pions
- Cone’s centred in η-φ coord of EGAMMA cluster
- Cone’s centred in η-φ coord of TOPO cluster in EM cal
- Cone’s centred in η-φ coord of TRUTH generated π0
Charged pions
- Cone’s centred in η-φ of TRUTH generated π±
- Cone’s centred in η-φ of TRACK position at 2nd layer
Neutrons
- Cone’s centred in η-φ of TRUTH generated neutrons
 In principle, it’s used a cone with ∆R<1.0  in this first contact, only it’s
required to select the cone algorithm with the best resolution.
For π0’s and neutrons:
- Cone’s centered in η-φ coord of TRUTH generated partc
For π±’s:
- Cone’s centered in η-φ of TRACK position at 2nd layer
But with ∆R<1.0 I’m taking into account more than one shower in the same cluster.
It’s needed to defined a smaller ∆R, different for each type of particle

55. Defined ∆R of the cone algorithm
 For π0’s:
From “Calorimeter Performance” analysis the cluster size are (for E<100GeV):
 Unconverted photons: 5x3 cells  ∆φ= 0.0625 ∆η=0.0375 (∆R<0.073)
 Converted photons and electrons : 7x3cells  ∆φ= 0.0875 ∆η=0.0375 (∆R<0.095)
For the reconstruction of the clusters from π0’s, will be used:
 ∆R <0.1 for starting, because I’m using very low ET
 ∆φ= 0.0875 ∆η=0.0375 : 7x3cells
 ∆φ= 0.0625 ∆η=0.0375 : 5x3 cells
 ∆R<0.0375: 3x3 cells
 For π±’s:
From LAr TestBeam analysis, the cluster size for pions:
 7x7 cells (∆R<0.12),
 9x7 cells (∆R<0.16),
 11x11 cells (∆R<0.20)…
For the reconstruction of the clusters from π±’s:
 ∆R <0.4
 ∆R<0.2
 ∆R <0.1
 For neutrons: as their shower will be as wide as the π±'s ones, the same values
for ∆R will be checked:
 ∆R>0.1, ∆R<0.2 and ∆R<0.4

56. ET Resolution
The best resolution is for ∆R<1.0, but it includes
more than the shower of one particle.
For π±’s the best resolution for TRACK-cone
with ∆R<0.4, but with ∆R<0.2. I have also a
good resolution and it let me a better definition
of the shower of only one π±.
For neutrons: the best resolution with ∆R<0.4,
but ∆R<0.2 is still very good resolution.
In both cases, ∆R<0.1 is too strict to
defined hadronic particles.
For π0’s: Resolution with ∆R<0.1 is the better.
Clusters with 7x3 and 5x3 cells gives us good
resolution but not enough.3x3 is too strict. They
could be useful when elect noise will be applied

57. Clustering Algorithms Comparison
 The previous results from Cone
algorithm are the best of all.
 Anyway, the results from TOPO
algorithm with Seed_cut=4 and
Neigh_cut=2 are very competitive
with them.
 EGAMMA-cluster give worse
resolution, in general, than TOPO
and Truth-cone, except at 1 GeV.

58. Topocluster with Electronic Noise
The values of have increased, now the ET deposited in TopoCluster comes
from the generated particles, but also from the electronic noise
π±’s neu π0’s
Asking for a minimum value of ET in Seed Cell and Neighbor cells:
Seed Cell >200MeV
Neighbor cells >80MeV
a similar value of without noise is obtained.
After these cuts, the size of the
Topocluster is up to 14 times
smaller.
This difference is more important
for the EM calo because there the
level of noise with respect to the
signal is bigger.

59. π±’s neu π0’s
The ET resolution get worse with the application of these cuts there is a loss in
energy reconstruction of the clusters. WHY?
Because we have applied a general threshold to the ETcell for all calorimeter, and the
electronic noise contribution is different in each layer of LAr and Tile.
Seed Cell >200MeV
Neighbor cells >80MeV

60. Conclusions
 WITHOUT NOISE:
 The best E resolution for VLE particles is obtained with cone
algorithms
 TopoCluster is a very competitive algorithm but doing the changes:
 Using CaloNoiseTool to model the EM Noise
 Applying lower thresholds to Seed and Neighbor cells:
 SeedCut=4 and NeighborCut =2
TopoClusters is event better than EGamma cluster for π0’s.
 WITH NOISE:
 The E resolution get worse for TopoCluster
 If we try to remove electronic noise, we get a loss in ET from particles
 It will be needed to applied ET thresholds in each layer of LAr and Tile

61. III. Clustering Algorithms for
VLE data of CombinedTB
 Combined TestBeam Setup
 Physcics Samples
 Energy reconstruction
 Particle Selection
The electron sample
 Separate pions from muons
First method: using sample D as a muon veto
Second method: Using the longitudinal profile
Third method: using MDT information
 Clustering info in CBT ntuples
 ET resolutions

62. A full slice of the ATLAS experiment has been tested with beams of different particles
(π’s, µ’s, γ, electrons and protons), at different energies (1-350 GeV) and polarities.
Inner Detector: 3 layers of Pixel, 4 layers of SCT and 2 modules-barrel slice of TRT
Barrel EM and HAD calorimeter: 2 barrel modules of EM LAr calo and 3 barrel
modules of HAD TileCal + 3 extended barrel modules of HAD calo
Muon spectrometer:

64. Physics samples
 events from 1 to 9 GeV at eta=0.35, with Calo info (LAr+Tile) and
the tracks info from TRT only (pixels have problems)
 100 k events for each point
 Mixture of e, π and µ
 Reconstruction with release 9.1.1
 Separate the different kind of particles
 Evaluate the fraction of e, π and µ
 Apply clustering algorithms
Ntuples were generated by Vincent with the default values of RecExTB:
castor/cern.ch/atlas/ctb/test/real_data/reconstruction/Combined/
Energy #Run Energy #Run Energy #Run
1 GeV 2101077 4 GeV 2101080 7 GeV 2101085
2 GeV 2101078 5 GeV 2101047 8 GeV 2101048
3 GeV 2101079 6 GeV 2101084 9 GeV 2101049

65. Energy Reconstruction
 E = Sum of cells with
|Ecell| >σpedestal
 Only cells in a small volume around the beam axis
 For LAr
0.25 < η < 0.45
­0.15< ϕ < 0.15
 For TileCal
0.20 ≤η≤ 0.50
­0.1< ϕ < 0.1
(cells A3, A4, A5, BC3, BC4, BC5, D1, D2)
Because the hadronic shower is wider than the electronic one,
and the most of the deposition comes from pions in Tile.

66. Particle selection
 Selection of good tracks
 trk_nTracks==1Only 1 track
 trk_nTrtHits[0]≥20 More than
20 hits per track
 to separate e from π/µ
 Cherenkov2 counter cut
 for electrons: sADC_C2>650
 for π/µ: sADC_C2<650
 high-level hits (improves the
Cherenkov efficiency)
 for π/µ: nHL>5
 for π/µ: nHL≤2

67. The Electron sample
Electrons are selected requesting:
sADC_C2>650 Cherenkov2 counter cut
nHL>5 number of high-level hits
No energy in TileCal sample D : to remove
the µ contamination

68. Separate pions from muons
Both pions and muons are:
 sADC_C2<650 Cherenkov2 counter cut
 nHL≤2 number of high-level hits
First method: using sample D as a muon veto
Assuming that only muons can reach
sample D and π signal is only coming
from pedestal, we put the cut:
ADVANTAGE: method very efficient and
easy to reproduce with MC
DISADVANTAGE: we can reject pions that reach the sample D, getting a bias.
In order to avoid it, different strategies are followed depending on ET:
a) below 6 GeV : using TileCal last sample as a muon veto. It is supposed that
there is no ET in Sample D from pions (only pedestal)
b) above 6 GeV : use another method longitudinal profile in TileCal

69. Second method: Using the longitudinal profile
Using the fact tha muons leave their ET uniformly in the detector
(normalizing by the path lenght)
E ∝path in matter
For ET>6GeV, different conditions are applied to , and

70. In LAr total
The contamination of muons increase when E decreases electrons
The number of electrons and pions decrease at low energies muons
pions

71. Clustering info in CBT ntuples
 Emcluster: clusters from the sliding window algorithm
 Tbemclusters: clusters from an algorithm used in
previous test beam. It has been added to allow comparison.
It’s a window of 3x3 cells.
 Emclusters and tbemclusters use only cells from the LAr calorimeter.
 Cmbclusters: sliding window clusters but they are done on towers (LAr+Tile) and
not anymore on cells. It is not working for the moment because of a coordinate
problem between LAr and Tile.
 Topo_EM and Topo_Tile cluster: Finds a seed
cell, then cluster expands by checking energy in
neighboring cells. Thresholds for seed and
neighbors can be changed. The default values are:
 seed threshold is E/σnoise>6
 neighbor threshold is E/ σnoise>3
(Hadronic TopoCluster is the sum of Topo_EM
and Topo_Tile)

72. e- in Lar: Energy distribution
For electrons at 9 GeV
For electron it seems as the cuts on
TRT works good

73. e- in LAr: Number of Clusters
#particles and
#cluster is very similar
 #clusters is very
similar between them
for each ET value.
#clusters is very low
 #clusters defined
increase with the energy.
(*) There is a cut (E>2 GeV) in this algorithm by definition

74. e- in LAr: Resolutions
In general, the E resolution is better when E increases
SW SW_TB TOPO_EM
9 GeV 7.57 8.92 10.48
8 GeV 8.51 10.04 11.64
7 GeV 7.85 6.93 8.51
6 GeV 8.83 7.81 9.62
5 GeV 13.07 15.47 17.34
4 GeV 11.04 11.47 14.78
3 GeV 9.59 (*) 14.38 20.39
2 GeV ---(*) 20.51 34.99 E resolution slightly better than it’s expected, WHY?
1 GeV ---(*) 80.75 48.38 Maybe problems in the reconstruction chain
The best resolution is for SW, but at 1-3 GeV we have bad results.
(*) There is a cut (E>2 GeV) in this algorithm by definition
TOPO obtain the worst resolutions
maybe it will be needed to change the thresholds for seed and neighbor cells.

75. Improvement in the resolution of electrons
New release of Athena is used:
Optimal Filtering is applied in LAr signal
Problems in the reconstruction chain have been solved.
Now the TopoCluster is the global cluster for Lar+Tile calo:”super3D”, as well as
new values are used for the thresholds:
There is a important improvement of the resolution
The values are of the order that are expected for VLE particles

76. Results for pions and muons
Results are very difficult to interpert, because there is still a mixing of µ’s and π’s
at energies above 7 GeV

77. New method to separate µ’s and π’s
Third method: using MDT information
Using the variable nMDTdig
to count the number of hits
in the different MDT stations
We can assume that events with more
than 8 digits in a MDT stations are
muons (because we have 8 plans
tubes per station)

78. After applying these cuts, the correct separation of π’s from µ’s above 7 GeV it’s possible

79. #TopoClusters is very similar to
#particles, so the clustering method
seems to works well.
The resolution from π’s is rather similar,
nevertheless the most important results
is the improvement in resolution for µ’s.

80. Conclusions
 The reconstruction of very low energy particles it’s possible with the
tools available in the reconstruction package for the Combined TB inside
Athena.
 For the recostruction of 1-9 GeV e-, the two Sliding Windows algo are
usefull, and the Topocluster results are very competivie with them. The
energy resolutions obtained are of the order that it is expected
 Nevertheles, it will be necessary to apply some changes in the ET thresholds of SW to
can apply them at 1-3 GeV e.m. particles
 The reconstruction of π’s and µ’s, first nedeed of a very accuracy
separation of them. We conclude to use the Sample D as muon veto for
E<6GeV and the MDT cuts for larger energies.
The values of E resolutions obtained are inside the expected ones.
 However, it will be interesting a tunning work to adapt the E threshold more properly to
VLE particles (as in the previous simulation analysis)