Transcript The Top Mass Measurement in ATLAS at the LHC “A Test of Systematics”

The Top Mass Measurement in
ATLAS
at the LHC
“A Test of Systematics”
James Proudfoot
Argonne National Laboratory
Argonne, Illinois, USA
On behalf of the ATLAS Collaboration
J. Proudfoot (ANL)
1
Oct 29, 2004
Overview
• The LHC parameters
• A brief overview of the Atlas detector
– unique capabilities
– Construction and installation status
– Tile calorimeter commissioning
• LHC Status (from a talk by Lynn Evans in June 2004)
• The top mass measurement:
–
–
–
–
–
–
A Personal Bias
Trigger
Jet measurement and calibration
SYTEMATICS
Modeling
Comparison to the TeVatron
• Summary
J. Proudfoot (ANL)
2
Oct 29, 2004
The Large Hadron Collider
•  s = 14 TeV p on p collider
• Luminosity
– 1033 cm-2 s-1 to 1034 cm-2 s-1
• 25 ns bunch crossing (40 MHz)
• High luminosity and large inc
at LHC implies:
– ~ 23 minimum bias events per
BC
– ~ 70 charged tracks/event with
pT > 1 GeV/c for || < 2.5
• Mass reach up to  5 TeV
Production cross-sections and
dynamics are largely
controlled by QCD
J. Proudfoot (ANL)
3
Oct 29, 2004
The ATLAS Detector
• Tracking in 2T solenoid:
A General Purpose
Detector
– Si pixel + strips (3+4 layers)
– TRT  particle id.
–  / pT ~ 5 10-4 pT  0.01 in 2 Tesla field
• EM calorimeter: Pb – liquid Argon
presampler + segmented EM calo.
 / E ~ 10% / E(GeV)
• Had. Calorimeter:
– Fe –scintillator (barrel)
 / E ~ 50% / E(GeV)  0.03
– Cu/W – liquid Argon(endcaps/Forwd)
 / E ~ 60% / E(GeV)  0.03
• Muons: “Air”; instrumented large
toroid magnet
 / pT ~ 10 % at 1 TeV/c
• Luminosity Uncertainty < 5%
J. Proudfoot (ANL)
4
Oct 29, 2004
ATLAS has a Unique Capability in the
Muon Spectrometer
MH= 130 GeV
muon spectrometer
standalone
The muon spectrometer resolution
dominates for Pt > 100 GeV/c
Inner tracker
stand alone
Standalone measurement
better than Inner
Detector for
MH>180 GeV
J. Proudfoot (ANL)
Muon spectrum.
standalone
5
Oct 29, 2004
Next 18 Months on the Schedule
Barrel
installation will
be done in
October 04
J. Proudfoot (ANL)
•
•
•
•
•
BT construction and installation
Barrel calorimeter installation
Barrel services installation
Cabling campaigns
Muon barrel installation
6
First toroid will
be installed in
Oct 04
(Scheduled for
Oct 26th.)
Oct 29, 2004
The Atlas Cavern in June 2004
October 26th
October 28th=>
J. Proudfoot (ANL)
7
Oct 29, 2004
Concurrently: Install services and begin to
commission the detector electronics
J. Proudfoot (ANL)
8
Oct 29, 2004
Services Being Installed on
Barrel (in Cavern)
J. Proudfoot (ANL)
9
Oct 29, 2004
Tile Calorimeter Commissioning
(an example of what lies before us)
Readout part
Insert LVPS, and
close with Iron
Plates
PMT blocks
HV part
J. Proudfoot (ANL)
Drawer
10
Oct 29, 2004
Phase1: Tile Calorimeter Commissioning Work
Packages
Detector
Racks,
DCS
LV
system
HV
system
Interlocks
Cesium
system
Laser
system
FE
Electronics
on the truck
Racks,
DCS
Cooling,
Sniffers
Cables
Racks,
DCS
DAQ
DB
FE
electronics
Racks
TILES
DCS
Offline
Cosmics L1
receivers
TTC-LTP
system
BE
electronics
FE-BE
integration
L1
receivers
J. Proudfoot (ANL)
Laser
calibration
11
Drawers and LVPS are
inaccessible when the detector
is closed for data-taking
Oct 29, 2004
Status of the LHC Project
Lyn Evans
CERN, 24 June 2004
The most critical single issue is
the installation of the QRL
Installation planning is being
revised to maintain summer
2007 for first beams
J. Proudfoot (ANL)
12
Oct 29, 2004
The ATLAS Physics Capability – Some
Background
I. Fleck, Eur Phys J C 34, 201, s 185
Rates are for
an instantaneous
luminosity of
2x 1033 cm-2 s-1
30Hz of W→en
3 Hz of Z→ee
2Hz of ttbar
J. Proudfoot (ANL)
13
Oct 29, 2004
Atlas Trigger: Primarily an Inclusive
Trigger at Level 1 & 2
Level 1:
Uses coarse grained
calorimeter and muon
information
Level 2:
Uses full granularity and full
resolution but in Regions of
Interest
Keep trigger inclusive
since there might also be
physics that we have not
modeled !
Level 3:
On passing Level 2, the full
event is built and passed to the
Event Filter
J. Proudfoot (ANL)
14
Oct 29, 2004
The Top Mass Measurement
A Personal Bias
• Compositeness → di-photons, di-jets, Etmiss
→ Calorimeter Performance
• Higgs → Photons, Etmiss, Forward Jet Tagging
→ Calorimeter Performance
• SUSY → Jets, Etmiss, Leptons
→ Calorimeter Performance
• W, Z, W’, Z’ → Isolated Leptons
→ Calorimeter Performance
• Top → Isolated Leptons, Jet Et, Etmiss
→ Calorimeter Performance
J. Proudfoot (ANL)
15
Oct 29, 2004
Why Pick the the Top Mass?
• To establish the performance of the detector,
we need to MEASURE a process which has a
CLEAR signature
• A PRECISION measurement of the top mass is
an appropriate process:
– The production rate for t-tbar is
PREDICTED to be enormous
– The measurement of the top mass touches
the performance of the ENTIRE ATLAS
Detector
– The top mass is an important measurement in
its own right
J. Proudfoot (ANL)
16
Oct 29, 2004
The top mass: a precision SM
measurement to be made at the LHC
The top quark mass, when combined with
precision electroweak data, constrains
the mass of the elusive Higgs Boson
t
W
H
W
b
m  m
2
W
2
t
W
W
W
m  log mH
2
W
For an equal error on the indirect
determination of the Higgs mass, the
uncertainty on the W mass has to be
150 time smaller than that of the
Top quark Mass
J. Proudfoot (ANL)
17
mW2  (mt2 ,log m H )

(Chris Hill, VthRencontres du Vietnam)
Oct 29, 2004
Atlas Event Selection for Top:
Isolated lepton + Jets
Gold-plated channel : single lepton
• pT (lepton) > 20 GeV
• pTmiss > 20 GeV
• ≥ 4 jets with pT > 40 GeV
|| < 2.4
Cone DR = 0.4
Period
• ≥ 2 b-tagged jets
L
= 1033 cm-2s-1
J. Proudfoot (ANL)
evts
dMtop(stat)
1 year
3x105
0.1 GeV
1 month
7x104
0.2 GeV
1 week
2x103
0.4 GeV
18
Oct 29, 2004
Jet Reconstruction Algorithms
• Cone Algorithm
– Highest ET tower for jet seed + cone
– Iteration of cone direction, jet
overlap, energy sharing, merging
• KT Clustering Algorithm
– Pairs closest calorimeter towers,
merging all particles into jets
• Cone size influence on reconstructed jet
energy and resolution
Cone Algorithm
 / E  a / E b
a (%GeV1/2)
b (%)
Full
Calo
48.2 ± 0.9
1.8 ± 0.1
DR=0.7
52.3 ± 1.1
1.7 ± 1.1
△ DR=0.7
● DR=1.5
DR=0.4
62.4 ± 1.4
1.7 ± 0.2
B. Caron, LHC2003
J. Proudfoot (ANL)
19
▲ DR=0.4
Oct 29, 2004
Atlantis top events
J. Proudfoot (ANL)
20
Oct 29, 2004
Signal and Background Rates
Based on Pythia Generator, with Fast Simulation
Assumes b-jet efficiency = 60%
P. Roy, Thesis 2002, Aubière : Clermont-Ferrand 2. Lab.
C5: >=2 b-jets
Phys. Corpusc. Cosmol
C4: >= 1b-jet
J. Proudfoot (ANL)
21
Oct 29, 2004
Jet Reconstruction Issues
•
•
Physics effects
– Fragmentation
– Initial and final state radiation
– Underlying event
– Minimum bias events
Detector performance effects
– Non-linear response
– Magnetic field
– Dead material and cracks
between calorimeters
– Longitudinal leakage
– Lateral shower size and
granularity
– Finite cone size (out-of-cone
loss)
– Electronic noise
J. Proudfoot (ANL)
22
GOAL→
 


50%
 3%
E
Dead Material Map
Oct 29, 2004
Energy Calibration “Tools”
•
•
electronics calibration and/or response of the calorimeters to a
known calibration “source”
Testbeam data plus simulation to refine the energy scale
determination
– Di-jet balance to take out zeroth order non-uniformity (in )
– In situ jet calibration from Z/g + Jets (where jets may be
either tagged b-jets or non-tagged jets
– Z → e+ e- mass calibration to determine the
electromagnetic energy scale
– Z → m+ m- mass calibration to determine the momentum
scale
• Top specific jet calibration using constrained fit to W
mass
• Eventually calibrate b-jet energy scale using Z → b bbar
J. Proudfoot (ANL)
23
Oct 29, 2004
ATLAS Test Beam Results
(one example)
• Combined Test: EM LAr and
Hadronic Tile Calorimeter
–  Energy Resolution
 / E  a / E b c / E
a
(%GeV1/2)
b (%)
c (GeV)
Data
69.8 ± 0.2
3.3 ± 0.2
1.8± 0.1
G-CALOR
61.7 ± 0.1
2.9 ± 0.3
1.5 fixed
Expect to set jet scale to 5-10% from testbeam + simulation
J. Proudfoot (ANL)
24
Oct 29, 2004
Offline Jet Energy Determination
from Monte Carlo “Truth”
•
Sampling Method
E jet  EPS  EEM  gEHAD   EEM 3  EHAD
–
–
•
Weights applied to different calorimeter compartments
Enlarged cone size yields increased electronic noise
H1 Method
E jet  EPS   EM ( EM , j )   EM , j   HAD ( HAD, j )   HAD, j   C EC
j
–
–
j
Weights applied directly
Better resolution and
Parameter
||=0.3
to cell energies
residual nonlinearities
Sampling Method
H1 Method
DR=0.4
DR=0.7
DR=0.4
DR=0.7
a (%GeV1/2)
66.0 ± 1.5
61.2 ± 1.3
53.9 ± 1.3
51.5 ± 1.1
b (%)
1.2 ± 0.3
1.4 ± 0.2
1.3 ± 0.2
2.5 ± 0.2
2 prob. (%)
1.6
0.8
27.3
66.7
J. Proudfoot (ANL)
25
Resolution
for
Back-toBack dijet
events
Oct 29, 2004
Z + Jets: Pt Balance
• Use pT balance between highest
pT jet and leptonic Z decay
• Apply tight 2nd jet veto and
Df>3.06 to reach ±1% sensitivity
level
Full Simulation
Atlas
Reconstruction
code 8.4.0
GeV
J. Proudfoot (ANL)
26
Oct 29, 2004
W → Jet Jet in situ Calibration
1% mismeasurement of jet
energies induces a top mass
shift of 1.6GeV
Mismeasurement of cos(QbW)
induces a top mass shift of
1.2GeV
Residual effects:
FSR effects large for
Pt~50GeV due to out-offcone energy
Jet overlap for Pt>200GeV
J. Proudfoot (ANL)
Compare reconstructed Jet to
Monte Carlo “truth”
27
Oct 29, 2004
Before calibration
After calibration
Fit 1, 2 to minimize:

m
jj  mW 


2

2
W
2
 E
1
  E
j  E j1
1
2
1
2

2
j2
E j2
2

2
M (W) = 80.3 ± 0.4
 = 6.9 ± 0.5
2
Where W = 2.1GeV, and 1 and 2 are the jet
resolutions – taken from Monte Carlo
J. Proudfoot (ANL)
28
Oct 29, 2004
Mtop reconstruction
Reconstruction of hadronic part
W from jet pair with the closest
invariant mass to MW
cut on |mjj-mW| < 20 GeV
Associate W with a b-tagged-jet
Cut on |mjjb-<mjjb>| < 20 GeV
Kinematic fit
Using remaining l+b-jet, the leptonic part
is reconstructed
P. Grenier
|mlnb-<mjjb>| < 35 GeV
Kinematic fit to the ttbar hypothesis,
with Mtop and MW mass constraints
J. Proudfoot (ANL)
29
Predominately
combinatorial background
from real ttbar events
Oct 29, 2004
Systematics
• Light and B-Jet jet energy scales
– Apply miscalibration to reconstructed jet energy and
evaluate effect on reconstructed top mass
– For light jet energy scale, the in situ calibration greatly
reduces the impact of the scale uncertainty
• Initial State Radiation
– Compare difference in mass for ISR on/off. Use 20% of
the as systematic
• Final State Radiation
– Compare difference in mass for ISR on/off. Use 20% of
the as systematic
• B quark fragmentation
– Vary Person fragmentation parameter within 1 of it default
value
• Combinatorial background
– Estimate by varying the assumptions for the background
shape
J. Proudfoot (ANL)
30
Oct 29, 2004
Systematic error on Mtop (TDR performance, 10 fb-1)
Initial performance : uncertainty on b-jet scale expected to dominate
b-jet scale uncertainty
1%
10%
 Mtop
0.7 GeV
7.0 GeV
c.f.: 10% on q-jet scale  3 GeV on Mtop
J. Proudfoot (ANL)
31
Oct 29, 2004
Top Mass Systematics at the
TeVatron
CDF:- DLM
D0:- Template Analysis
(Inc. PDFs)
There are several systematics
at the ~0.5-1GeV level that are
missing in the Atlas enumeration
This is good as it means that
there is still work to do while
we wait for the accelerator !
J. Proudfoot (ANL)
32
Systematic
Uncertainties
DMtop(GeV/c2)
Jet Energy Scale
5.3
ISR
0.5
FSR
0.5
PDF
2.0
Generator
0.6
Spin correlation
0.4
NLO effect
0.4
Bkg fraction(±5%)
0.5
Bkg Modeling
0.5
MC Modeling(jet,UE)
0.5
Transfer function
2.0
Total
6.2
Oct 29, 2004
Comparison to present day
Measurements from TeVatron and LEP
(Chris Hill, VthRencontres du Vietnam)
• Recent, more precise D0
measurement
– mtop = 180.1 ± 5.4 GeV/c2
– Nature 429, 638-642 (2004)
• New world average
– mtop = 178.0 ± 4.3 GeV/c2
– Changes Higgs mass value favored
by electroweak fits
• mH  113 GeV/c2
In first year of data-taking, Atlas can
reduce the uncertainty on the top mass
by a factor of ~2
J. Proudfoot (ANL)
33
Oct 29, 2004
Other Modeling Issues
(I Don’t Know but I’ve Been Told…….)
• Top Pt spectrum
• Monte Carlo simulation
of Z + jets:
– Jet Pt spectrum
– Jet multiplicity
– s
• Full Detector
simulation
– Signal AND Background
• Improve understanding
of b-jet energy scale
J. Proudfoot (ANL)
Comparison to analytic calculation
(Weber & Frixione) suggests that
Pythia is not a good
representation - another area
for further study
34
Oct 29, 2004
Some Other Analysis Options
Possibilities
• Rate has its advantages !
– Different jet reconstruction algorithms
(e.g. continuous jet algorithm)
– Use only large transverse momentum
ttbar events (has different
systematics)
– Measure top mass in events with a J/y
in the final state
J. Proudfoot (ANL)
35
Oct 29, 2004
Backgrounds ?
One Instrumental Effect in the Tile Calorimeter
The Atlas Trigger and
event selection will
preferentially select
background events
from events with
poorly measured
Etmiss
J. Proudfoot (ANL)
36
Oct 29, 2004
More Ways to Generate Fake
Etmiss & Thereby Backgrounds
• Gap between barrel and endcap
calorimeters
• e/h = 1
• Mis-calibration
• Barrel EM calorimeter insensitive region at
~0
• Detector “average center” vs beam center
• +…
J. Proudfoot (ANL)
37
Oct 29, 2004
The Impact of Staging
• Staged detector components include:
– The middle layer of the pixel detector
– A large part of the high level trigger and
data acquisition hardware
• The impact include:
– A 30% degradation in the b-jet identification
efficiency
– Reduced capability for triggering on high rate
QCD processes, which may be needed to
understand backgrounds and detector
efficiencies
J. Proudfoot (ANL)
38
Oct 29, 2004
Summary
Many members of the Atlas collaboration have been intensively
studying detector performance – I have shown some of these
results and want to acknowledge in particular input I received
from the Atlas Top Conveners Dominique Pallin and Marina Cobal
I hope I have convinced you that these studies give confidence
that Atlas will make a precision measurement of the top quark
mass and that in combination with the W mass and (hoped for)
Higgs mass we can explore the limits of the Standard Model
Along the way, it is clear that there is much yet to be learned
about the detector performance and basic W & Z production ..
Both for energy scale calibration and as a “calibration” of the
QCD predictions for jet spectra and multiplicity distributions
and Etmiss
J. Proudfoot (ANL)
39
Oct 29, 2004