Transcript Corporate 2 Template - Istituto Nazionale di Fisica Nucleare

Study of The Diffractive
Component of the Inclusive
Z->e+e- and Z->m+m- Cross Section
Candidato: Marone Matteo
Relatori: Dott.sa Arcidiacono Roberta
Dott. Cartiglia Nicolo’
Scuola di dottorato in Scienza ed Alta Tecnologia,
Indirizzo Fisica ed Astrofisica
Ciclo XXIII, Ph.D. final dissertation
Outline
• Introduction
– LHC & CMS
– ECAL
• Measurement of
ECAL Thermal Stability
– DCU
– Results
• Study of the
Diffractive Component
– Pile-up Removal
– Results
2/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
My Activity during Ph.D.
My activity in ECAL:
• Installation and Commissioning
• Readout Software Development
• Detector Thermal Stability
Analysis work:
• Diffractive Z Production
2008
3/34
2009
Matteo Marone –Ph.D. Final Dissertation
2010
Torino- June 20th 2011
2011
LHC
4/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
CMS Detector
CMS physics goals:
• Perform precision
measurements in the
electroweak sector
• Higgs search
• Supersimmetry and new
Physics
5/34
Matteo Marone –Ph.D. Final Dissertation
• Very good muon identification system
• Excellent electromagnetic calorimeter
to resolve the energy of the
electrons/photons
• Efficient tracker system to reconstruct
the tracks and measure the
momentum of the charged particles
Torino- June 20th 2011
ECAL
ECAL is an homogeneous calorimeter made of PbWO4 crystals:
• 36 SuperModules, 1700 Crystal each
• 4 Endcap Dees, 3662 Crystals each
• 8 meters long
• 90 Tons of Crystal
• More than 75000 channels
Barrel crystals
Barrel
crystals
Endcap
Preshower
Pb/Si
Barrel
Supermodule
Barrel
Supermodule
Energy

Light
Light

Current
Current

Voltage
Voltage

Bit
Physics reach of the ECAL, in
particular the H->gg discovery
potential, depends on its excellent
energy resolution.
Requires high precision
calibrations
Bit

Light
DAQ
Optical Fiber
Crystal
AP
D
MB
MGPA
VFE
6/34
Matteo Marone –Ph.D. Final Dissertation
Trigger
ADC
FE
Torino- June 20th 2011
Forward Calorimeters @ CMS
Hadronic Forward Calorimeter
• @ 11 m from IP
• Coverage 3 < |h| < 5
• Steel absorbers and embedded
radiation-hard quartz fibers for fast
collection of Cherenkov light
• Two calorimeters (minus and plus
side)
7/34
Matteo Marone –Ph.D. Final Dissertation
CASTOR Calorimeter
• W absorber & quartz plates sandwich
• @14m from IP
•coverage -5.2 < h < -6.6
• signal collection through Cherenkov
photons
• 16 azimuthal segments in φ and 2
(EM) + 12 (HAD) long. segments.
• available on only one side
Torino- June 20th 2011
ECAL Thermal stability:
Hardware installation,
calibration and
commissioning
2008
2010
2009
Commissioning
Read Out Software Development
ECAL Thermal Stability
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Why Measure the Temperatures?
ECAL response sensitive to variations of:
• Crystal transparency (irradiation)
• Intercalibration
• Temperature: ∂(LY)/∂T ~ -2%/oK
1/M(∂M/∂T) ~ -2%/oK
• High voltage: 1/M(∂M/∂V) ~ 3%/V
Temperature monitoring
system is needed
affect the constant term
M= Photodetector
gain
LY= Light Yeld
Temperature stability within 0.05/0.1oC
9/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Detector Control Units (DCU)
The DCUs are special ASIC chips able to
read the following quantities:
Temperature
0.012 °C
Currents
340 nA
Voltages
~ mV
Energy

Light
Crystal
Light

Current
APD
Current

Voltage
MB
Voltage

Bit
MGPA
ADC
Basic Read-out Geometry: 5X5 crystals (TT)
VFE
FE
Very high granularity:
8 DCUs per TT ~ 20000 (1 each VFE
and 3 in LVR boards)
VFE
LVR
MB
Optical Fiber
Trigger
Useful tool to deeply investigate
the status of the calorimeter
10/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
ECAL Thermal Stability
• A detailed study of temperature stability has been carried on during
each collision period.
• DCU system provides one temperature reading every 10 (25) crystals.
Temperature estimation obtained driving a known internal current
through an external thermistor.
• The analysis has been performed using two independent monitoring
system: DCU and Precision Temperature Monitoring (PTM)
Poor granularity: 4 sensor per SM
Useful to calibrate the DCU sensors and
to double check the results
Results have been published in:
• CMS Paper (CFT-09-004) “Performance and Operation of the CMS
Electromagnetic Calorimeter” Published on Jinst
• R.Arcidiacono, M.Marone, “Ecal thermal stability during Cosmic Rays Run
2008”, CERN Detector Note number DN2010/003 , 2010.
11/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Results
• The RMS distribution of
every temperature sensors
estimates the detector
thermal stability
EB
EE
Period
RMS EE (°C)
RMS EB (°C)
2010 BEAMII 0.009
0.007
2010 BEAMI
0.015
0.008
CRAFT09
0.011
0.006
CRAFT08
0.017
0.009
• Very good spatial uniformity and stability in time.
12/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Results (2)
• Integration of the DCU in the readout (online) software
• Calibration of detector temperature thermistors
• Measured the Barrel and Endcaps temperature stability to
be within the specification (0.05/0.1oC). Measured the
detector thermal time constant (in the “turn on” transition)
to be ~2 hours in the barrel and ~6 in the Endcaps
• Help the ECAL community to investigate front end
problems (APD leakage, dead channels,.. ) using the DCU
data
“ECAL Front-End Monitoring in the CMS experiment” presented at
CHEP09: “International Conference On Computing In High Energy
Physics And Nuclear Physics”, March 2009
13/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Data Analysis:
Measurement of the Inclusive
Z->e+e- and Z->m+m- Cross
Section
2008
2010
2009
Diffractive Z study
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
2011
Diffractive Physics at LHC
• The study of hard diffraction at LHC is feasible and it will
offer the possibility to explore and test the ideas and
models developed at much lower energies.
• Diffraction: inherently present in p-p collisions (30% of tot)
• Pomeron (IP): successful description within Regge theory
of diffractive scattering
15/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Data Samples
• The data are divided in two periods:
LHC Run A
Period (2010) 04-08
BX Inst.Lumi.
LHC Run B
09-10
0.1-0.2 1030cm-2s-1 0.2-0.6 1030cm-2s-1
• Pythia 6 (tune D6T and Z2) has been
used to simulate the Drell-Yan (DY) events
decaying into ee (μμ)
• PomPyt has been used to simulate:
– Single Diffractive Z boson production
– Dissociative (or Double Diffractive)
Matteo Marone –Ph.D. Final Dissertation
X
X
How do we select the diffractive
over the non diffractive part?
16/34
X
Torino- June 20th 2011
Rapidity Gaps
• In diffraction the hadronization of the
final states X and Y happens
independently. If s is large enough, then
there is a gap in rapidity in between X-Y
• Since gaps are exponentially suppressed in QCD
fragmentation, a cut on rapidity gap increases the relative
fraction of diffractive events.
• @ LHC, s, MX and My are very large
The particles can easily cover a large zone
of the CMS detector total acceptance
We select diffractive events requiring visible rapidity gap
17/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Z Candidates Selection
•
•
•
•
•
•
18/34
HLT trigger muon pt>9 GeV
h <2.1
X2/NDOF < 10
Two muon stations fired
10 hit in the tracker and 2 in the
pixel detector
Transverse parameter < 2mm
EWK standard Isolation Criteria
Matteo Marone –Ph.D. Final Dissertation
Z -> mm
•

•
•
•
Pass HLT trigger (Cluster Et>15
GeV)
Reconstructed within the fiducial
region
Track trajectory, estrapolated to
match the ECAL Cluster
Reject Barrel Spikes
EWK standard isolation criteria
Z -> ee
•
Known problem in the ECAL calibration.
No further conditions on the Z mass are
requested
Torino- June 20th 2011
Definition of the Variables
We use the following variables:
Particle
Position
Threshold
(GeV)
Charged
Particle
CMS
Pt > 0.5
Neutral
Barrel
Et > 1.5
Neutral
Endcap
Et > 2.0
Neutral
HF
Et > 4.0
• SumHF: the energy deposit in the HF
 hMax: max η of energy deposits in the detector
 z: fractional momentum loss of the scattered proton in the diffractive
event
• MinHF: the minimum deposit in one HF side (+/−)
19/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Diffractive Selection with MC
The conventional way to recognize a diffractive event is to look for rapidity gap in its particle flow. Since gaps are exponentially suppressed in
QCD fragmentation5, the cut on rapidity gap increases the relative fraction of diffractive events.
• We have studied which was the best size of the rapidity gap to
reject the background and select signal
We select events
requiring HF=0 (2 units
of gap)
Ln(M2x)
CMS
In the data, LRG suppressed by the presence of the Pile-up
20/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Pile-up
• The number of PU events follows a
Poisson distribution
(L  ) pileup
P(n pileup ) 
 e -(L  )
n pileup!
n
• A possible way to remove PU can be
to require only one vertex in the

event. The number of events having
one vertex decreases when
luminosity increases.
• PU interaction can be classified into:
•“hard” PU. Visible interactions
(2.4< h). Can be removed requiring 1
vertex
•“soft” PU. Interaction not detected
and therefore not removed by the
one vertex selection
To correct for this loss of selection efficiency a method is presented
21/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Event reweight
The conventional way to recognize a diffractive event is to look for rapidity gap in its
particle flow. Since gaps are exponentially suppressed in QCD fragmentation5, the
cut on rapidity gap increases the relative fraction of diffractive events.
One vertex only
Events collected at higher
luminosity have less probability of
being selected.
Fit the fraction of
events with no energy in HF as a
function of the BX inst. luminosity.
assign to each event a weight
22/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
z distribution in diffractive events
The conventional way to recognize a diffractive event is to look for rapidity gap in its
particle flow. Since gaps are exponentially suppressed in QCD fragmentation5, the
cut on rapidity gap increases the relative fraction of diffractive events.
Using PomPyt, we simulate the z distribution with and without the
HF=0 cut
HF=0
HF=0 &
z<0.03
PomPyt
0.19
0.18
Pythia D6T
0.67 10-3
0.40 10-3
Pythia Z2
2.33 10-3
1.53 10-3
The simulations show that the diffractive signal is contained within the
kinematic region [0-0.03] z.
Limiting the analysis to this kinematic region will also produce a good
signal enhancement.
23/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Final Selection
Diffractive events have been
selected requiring:
• energy below a minimum
threshold in HF- or HF+
calorimeters
• only one vertex with a
quality cut to avoid
reconstruction of fake
vertices
• Value of ζ within 0 < ζ < 0.03
To measure the signal, the kinematic region has to be split
in a certain number of bins
24/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Migration
The reconstructed ζ is almost always underestimated if compared with
the true value, because of:
• incomplete detector coverage
• particle thresholds.
Consequently a migration from
high ζgen values to small ζrec value
is expected.
To evaluate the impact of the migration effect, we have studied the
resolution, purity and the migration maps. We chose then number of bins
requiring the following limits:
25/34
Resolution
< 30-40 %
Purity
> 50%
Efficiency
[50-150] %
Matteo Marone –Ph.D. Final Dissertation
Influence the number of
z bins
Torino- June 20th 2011
Resolution
Relative
Resolution
Absolute
Resolution
ζ measured is, on average 30% lower than the generated value, and its
resolution is 28%.
kinematic region divided in two equal bins (0≤ ζ ≤0.015 and 0.015≤ ζ
≤0.03).
Migration maps, purity and efficiency have been checked to be good
26/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Unfolding of data distributions
We have used the Pythia 6 D6T and
Z2 Monte Carlo samples, generated
without pile-up events: necessary
to remove the pile-up contribution
from the data events before being
able to compare
Example: MinHF Unfolding
1)Divide the distribution in energy
bins
2)For each bin, calculate the
fraction of events as a function of
BX Instantaneous Lumi
3)Extrapolate to zero Lumi to
obtain the pile-up free number of
events
27/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Which MC fits better?
• Discrepancy between data and Monte Carlo in the description of the
energy flow in the forward region.
• Impossible to choose one single Monte Carlo model for the description
of the non diffractive part
Forced to use two Pythia tunes, D6T and Z2
28/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Selected Events
Data and MC events which pass the above selection:
• Different behavior of the two Pythia tunes.
• The number of selected data events is small, especially if compared to
the Z2 tune prediction.
• Diffractive PomPyt events which pass the diffractive selection cuts is
very large compared to data
29/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Signal Significance
ni
Significance defined as:
Range
D6T
Z2
0-0.015
1.38
-0.89
0.015-0.03
2.26
0.38
TOTAL
2.62
-0.07
Assuming D6T to be the correct background description, then we
would have a significance of about 2.6 σ. Considering the Z2 tune,
this value drops down to ∼ 0 σ.
To assess at 3 σ the presence of a signal, we would need ∼ 11 pb−1.
The 5 σ signal is instead assessed with ∼ 29 pb−1.
30/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Cross Section Measurement
Cross Section evaluated as:
Where,
A is the acceptance
L the (effective) integrated Lumi
eZ the efficiency of the Z boson
selection
eD efficiency of the diffractive
selection
31/34
MonteCarlo
Z->ee (pb)
Z->mm(pb)
Combined (pb)
D6T
33±12
9±8
42±15
Z2
14±12
-9+8
5±15
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Prospects for 2011
The request of no energy in both CASTOR (-6.6≤ η ≤-5.2) and HF
calorimeters corresponds to a gap of ∼ 3.5 units, which makes
this selection virtually background-free.
MonteCarlo
PomPyt
Pythia D6T
Pythia Z2
HF=0
9.6%
0.002%
0.005%
0.0002%
0.0006%
HF=0 + CASTOR 8.0%
CASTOR calorimeter has suffered
of intermittent calibration
problem during 2010.
This study shows the possibility
to use this cut to obtain a cross
section measurement during 2011
32/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Conclusions
• In this thesis we have proposed and employed a novel
method to select diffractive events.
• We have derived a weight function that weights
diffractive events on the probability of having a rapidity
gap at a given luminosity
• The extraction of the diffractive signal from the events
that pass our selection criteria is further complicated
by the current discrepancy between data and Monte
Carlo in the description of the energy flow in the
forward region.
• This mismatch, which is actually quite important, did
not allow us to choose one single Monte Carlo model
for the description of the non diffractive part but has
forced us to use two Pythia tunes, D6T and Z2, which
bracket the range of uncertainties.
33/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Conclusions (2)
• Within these constrains, and due to the quite low
luminosity, we were not able to establish the presence
of diffractive Z production, but only to see a production
excess over one of the two Pythia tunes prediction.
• We are confident that the tools developed for this
analysis can be applied to the much larger sample of
the 2011 data, and we are looking forward to do the
analysis in the next few months.
34/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Spares
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Read-out detector software
The digitized data from the FE are read by the
the off-detector electronics, consisting of
54 Readout Units each comprising three type of
VME boards:
Clock and Control System (CCS)
Trigger Concentrator Card (TCC)
Data Concentrator Card (DCC).
Data reduction is achieved using a Selective
Readout algorithm based on the classification of
the detector in high al low interest regions (SRP)
The ECAL Online software is responsible for the operation of the ECAL
detector during data taking. The system is built on top of the CMS data
acquisition frameworks (XDAQ) and interfaced with the run control (RCMS).
In parallel, other relevant front end parameters are
read out by the DCU system, heavily used during
the commissioning phase
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Off-Detector Electronics
CCS (clock and control system) :
LHC clock and control signals +
front-end initialization
DCC (data
TCC (trigger
concentration
concentration card):
card):
Encoding of TT
Data reduction
Regional
Transmission to
Calorimeter TT TT
central DAQ (at
importance
Level 1 rate)
transmission to SRP
(at Level 1 rate)
SRP (Selective Readout Protocol):
send to the DCC the list of trigger towers
to be read out
Overall the off-detector electronics is made by 18 VME-9U and 1 VME-6U
crates controlled by 28 crate mounted PCs
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
What is monitored
APD:
currents (1 DCU for xtal = 1700/SM)
temperatures (1 DCU every 10 xtals = 170 values/SM):
VFE & LVR:
DCU internal temperatures (8x68 values /SM)
MEM box:
VDD_1, VDD_2, 2.5 V, Vinj (4X2 values / SM)
DCU internal temperatures (1x2 values /SM)
LVR:
3 thermistors
2.5 V (12x68 values / SM)
4.3 V (2X68 values / SM)
0.1 V – inhibit (1X68 values /SM)
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
DCU Software Architecture
XDAQ
DCUConverter
CondDB
DCU Reader
Calibrations
DCS – Detector Control
System
Soap
PC
Storage
Data
Files
Converter
Write
CondDB
8/22
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Detector Calibration
Calibrations aim at the best estimate of the energy of e and g’s
Energy deposited over multiple crystals:
Ee/g = Fe/g G i ci Ai
•
•
•
•
Amplitude in ADC counts Ai
Intercalibration: uniform single channel response to a reference ci
Global scale calibration G
Particle-specific corrections (containment, clustering for e/g’s)
Fe/g
Intercalibration together with global scale feeds directly
into the constant term
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
DCU graphical interface
14/28
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Z-> ee In situ Intercalibration
The electromagnetic shower spreads over several
crystals. linear system associated to a huge matrix have
to be inverted in order to get the single inter-calibration
factor
• Single region intercalibration
coefficient can be obtained with an
iterative method
Can be used to tune Barrel/Endcap
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Intercalibration
Problem1: the same photon (or electron) gives a
different answer (in ADC counts) depending upon
the crystals it hits.
• each crystal has a specific light yield
• each photodetector has its specific gain
Solution: find 75848 coefficients which make every
crystal answer in the same way
2100 ADC
2000 ADC
Intercalibration has been achieved in several ways, with different
precision:
EXAMPLE:BARREL
- Using data collected in the laboratories : 4.5-6%
- Cosmic ray (all): expose each SM to cosmic rays: 1-2 %
- TestBeam (9 SM): electrons at a given E in each crystal ~ 0.3 %
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Z->ee events selection
(Leptonic)
• At the nominal LHC c.m. energy, the
leptonic Z cross section is ~2nb:
• Decreasing to 0.9nb at 7 TeV
• Main background is due to QCD
Dijets and γ + Jet:
Channel
Cross section (nb)
QCD Dijets
~5x105
γ + Jet
~2x102
• High transverse momentum leptons
are the strong signature for Z decay
24/28
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Global scale
Problem2: the ECAL response depends on the energy of
the incoming particles itself.
The “linearity” of the calorimeter must be studied at the
level of the per mille.
Solution: find absolute references to tune the energy
scale
• Z and W decays, J/Psi, pi Zero and others.
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Energy reconstruction in ECAL
The measurement of the electron E is
hampered by the amount of tracker material
and by the strong magnetic field.
Electrons radiate brem. photons in the
azimuthal direction Φ
Brem
~ 35% of the photons
radiate more than
70% of their energy
γ
γ
The ECAL “superclustering” is designed to
take into account the spread and the brem
Clustering
ε ~ 99% for p>7GeV
26/28
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Temperature Measurements
This chip drives an internal (known)
current across a thermistor glued on the
back of the crystals
The thermistor temperature response
has been studied prior in laboratory
The in situ read-out circuit differs from
the one used in calibration
Another calibration has been performed
using an independent monitoring
system: Precision Temperature
Monitoring PTM
5/22
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Z->ee variables
1276
1271
353
H/E < 0.1
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
pb
ECAL Dead Channels
ECAL shows a certain number of problems ( 1% of dead
channels, DAQ related errors).
Any missing channel directly affects the energy
reconstruction.
Therefore systematic studies are necessary to tune the
official reconstruction algorithm with the real data.
27/28
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Cross Section Measurement
• We measure the inelastic pp cross section using pile-up (PU)
events:
(L )
P(n pileup ) 
 e-(L  )
n pileup!
n pileup
The probability of having npileup
depends on the total (pp)
cross section.
•The pile-up depends on the “Luminosity per bunch crossing
(Lbx)”: max. during 2010 = ~0.6 1030 cm-2 s-1
 Cross checked using the number of triggers in each bunch (L

*  = Nevents)
•Pile up events are recorded by a high efficient stable trigger
(e.g. Double ee, pt > 10GeV)
• The goal of the analysis is to count the number of vertices
as a function of luminosity
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Result - fits
Using the correction functions, we unfold the measured vertex
distributions to obtain the correct distributions which we fit with a
Poissonian function:
PU= # Vertexes –1
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Results - Cross section
For each of the PU distribution we obtain a value of the cross
section and then these 9 values are averaged
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Proton Dissociation
Diffractive events in which the proton,
after the Pomeron exchange, splits into
a leading baryon and into a system of
particles (Y)
It is interesting to calculate
the Ratio
Dissociative/Diffractive
~ 1/2.5
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011
Migration Studies: Other Results
Requiring 2 bins, migration map,
efficiency and purity are within the
limits
27/34
Matteo Marone –Ph.D. Final Dissertation
Torino- June 20th 2011