W Mass From LEP W Mass From LEP Fermilab Wine and Cheese Seminar Fermilab Wine and Cheese Seminar 6th October, 2006 6th October, 2006 Ambreesh Gupta, University of Chicago Ambreesh Gupta, University of Chicago
Dec 17, 2015
W Mass From LEPW Mass From LEP
Fermilab Wine and Cheese SeminarFermilab Wine and Cheese Seminar
6th October, 20066th October, 2006
Ambreesh Gupta, University of ChicagoAmbreesh Gupta, University of Chicago
Ambreesh Gupta Fermilab Wine & Cheese 2
Outline
1. Introduction
- W Boson in the Standard Model of Particle Physics
2. W mass Measurement
- Identifying and reconstructing W’s.
- Mass extraction techniques used by LEP experiments
3. Systematic Uncertainties on W mass measurement
5. Summary
I will show results from all the four experiments with details on OPAL analyses.
Ambreesh Gupta Fermilab Wine & Cheese 3
Standard Model of Particle Physics Our picture of the fundamental constituents of nature
There are about 19 (+10) free
parameters in the theory to be
determined experimentally
Standard Model predicts
relationship between
these parameters.
Ambreesh Gupta Fermilab Wine & Cheese 4
Standard Model Relations• Standard Model predicts relation between the parameters; W boson mass(MW)
and Fermi constant(GF), fine structure constant(), Z boson mass (MZ)
: electron g-2 0.004 ppm
GF : muon life-time 9 ppm
MZ : LEP 1 lineshape 23 ppm
• Precision measurements require higher order terms in the theory and help
constraint the unknown pieces.
(running of )
weakrr Δ+Δ−≡Δ
f WHWtW W
Ambreesh Gupta Fermilab Wine & Cheese 5
Precision EW
top-quark mass “predicted” by electroweak corrections prior to direct discovery
The measured W mass precision is such that Top and Higgs loops required
for consistency in the Standard Model (SM)
This gives an indirect inference on the Higgs.
Better precision on W mass constraints the Higgs Indirect measurement of W mass
- W mass known to 20 MeV from indirect measurement (LEP1 + SLD +Tevatron).
- A direct measurement of W mass with similar precision is of great interest. Measurement of the width of W boson can also be carried out at LEP providing
further checks on consistency of the SM.
Ambreesh Gupta Fermilab Wine & Cheese 6
Large Electron Positron Collider (LEP)LEP I (1989-1993) : Z physics. 18 million Z bosons produced
LEP II (1996-2000) : W physics. 80,000 W’s produced. (Energies from 161 GeV – 209 GeV) W’s produced in pairs.
Ambreesh Gupta Fermilab Wine & Cheese 7
The Four LEP Experiments ALEPH
L3
Ambreesh Gupta Fermilab Wine & Cheese 8
WW Production and Decay at LEP
• W’s produced in pairs at LEP
- 700 pb-1/experiment; 40,000 WW BR ~ 44%
BR ~ 46%
BR ~ 10%WW ll
WW qql
WW qqqq
Efficiency Purity
l l 70% 90%
qql 85% 90%
qqqq 85% 80%
Backgrounds
Ambreesh Gupta Fermilab Wine & Cheese 9
Event Selection
Very good agreement between expected and observed.
Event selection primarily based on multivariate relative likelihood discriminants
OPAL
Ambreesh Gupta Fermilab Wine & Cheese 10
W Mass at LEP
• The WW cross section at s = 2Mw
sensitive to W mass
• LEP experiments collected
10 pb-1 data at s = 161 GeV
• Combined Result :
Mw = 80.40 0.21 GeV
Most of LEP 2 data at higher energies - use direct reconstruction There are two main steps in measuring W mass and width
1. Reconstruct event-by-event mass of W’s
2. Fit mass distribution Extract MW and W.
However, jet energies poorly measured ( /E ~ 12% ), neutrinos unobserved.
Kinematic fitting plays vital role
Ambreesh Gupta Fermilab Wine & Cheese 11
Kinematic Fitting Mass Reconstruction
- Identify lepton and jets (DURHAM)
-- Energy flow techniques
- Kinematic fitting
-- Use LEP beam energy as constraint
-- Total Energy = s; Total Momem. = 0;
-- Additionally, apply equal mass constraint
mw+ - mw- = 0;
Significantly improved mass resolution
Caveat - Photon radiation will change
s s’ (photon energy)
Need good WW4f theory model (~0.5% theory error )
Ambreesh Gupta Fermilab Wine & Cheese 12
Mass Reconstruction qql channel
- 1 or 2 constraint kinematics fit
- Golden channel
qqqq channel
- Well constrained events
- But, ambiguity in assigning jets to
W’s Combinatorial Background
- 5-jet event: 10 comb., 4-jet: 3-comb.
Ambreesh Gupta Fermilab Wine & Cheese 13
Fit Methods
Re-weighting
- Weight fully simulated events to create
sample with new W mass and width
parameter
- No external bias correction needed
- Need large event sample to derive stable
weights
80.33 81.33
LEP experiments used three likelihood methods to extract W mass and width
from the reconstructed mass spectrum.
1. Re-weighting
2. Breit-Wigner
3. Convolution
The primary difference between the methods is the amount of information
they try to use for the best measurement.
Ambreesh Gupta Fermilab Wine & Cheese 14
Fit Methods (continued)
Breit-Wigner
- Fit to W mass spectrum with Breit-Wigner function
- Width adjusted to account for resolution and ISR effects.
- Bias corrected by comparing to fully simulated MC.
Fitted Function (70-88) GeV mass
Convolution
- P(m1,m2|Mw,Gw)R(m1,m2)
- Build event-by-event Likelihood
- Maximize statistical sensitivity
- Need bias correction as in BW
Ambreesh Gupta Fermilab Wine & Cheese 15
Likelihood Variables - Likelihood built using three variables -- both in qqlv, qqqq channels
- ~ 400 events per bin for stable fit
- Fit for eight energy point, four channels, then combine => lots of MC needed.
5C fit mass error 5C fit mass Hadronic 4C mass
5C fit mass error 5C fit mass4C fit mass
difference
- OPAL variables
- ALEPH also
3-d fit
Ambreesh Gupta Fermilab Wine & Cheese 16
Performance of Likelihood Fucntion
Test the central value modeling with bias plot Test the uncertainty on central value with pull distribution.
OPAL
Check bias and pulls distributions. . .below a typical example
Ambreesh Gupta Fermilab Wine & Cheese 17
OPAL W mass Very good agreement between three methods in
channel and year Strong correlation between methods
=> Combining them had only small stat. gain CV, which has the smallest expected statistical
uncertainty is used as the main method. Use of momentum cut analysis makes
significant reduction in FSI uncertainty. Final W mass and total uncertainty from
the three methods on OPAL -
Mw ΔMw (Stat.+Syst.)
CV 80.416 0.053
RW 80.405 0.052
BW 80.390 0.058
Ambreesh Gupta Fermilab Wine & Cheese 18
LEP W Mass The combined preliminary LEP W mass
MW = 80.376 0.025 (stat) 0.022 (syst) GeV
Systematics on W mass
Source
Hadronisation
QED(ISR/FSR)
Detector
Colour Reconnection
Bose-Einstein Correlation
LEP Beam Energy
Other
Total Systematics
Statistical
Total
qql qqqq combined
14
7
10
9
2
10
4
19
5
8
35
79
11
13
8
10
0
0
9
3
21
30
36
44
40
59
22
25
33
Channel weights qqlv : 76% qqqq : 22% xs : 2%
(MeV)
Ambreesh Gupta Fermilab Wine & Cheese 19
LEP W Width The combined preliminary LEP W width
W = 2.196 0.063(stat) 0.055(syst) GeV
Systematics on W width
Source
Hadronisation
QED(ISR/FSR)
Detector
Colour Reconnection
Bose-Einstein Correlation
LEP Beam Energy
Other
Total Systematics
Statistical
Total
qql + qqqq
(MeV)
40
6
22
27
3
5
19
55
63
84
Ambreesh Gupta Fermilab Wine & Cheese 20
LEP Beam Energy• LEP beam energy used in event kinematics fit ΔMW/MW ΔELEP/ELEP
• Beam energy calibrated using
- Resonant De-Polarization (41- 60 GeV.)
- Extrapolated to LEP II energies NMR probes
- Main systematic error due to extrapolation
• Extrapolation checked with
1. Flux Loop
2. Spectrometer
3. Synchrotron Oscillation• Final results on LEP beam energy ( Eur. Phys. J., C 39 (2005), 253 ) - Reduction of beam energy uncertainty used in earlier W mass combination
- oldΔEbeam= 20-25 MeV ΔMW = 17 Mev
- new : ΔEbeam = 10-20 MeV ΔMW ~ 10 Mev -- OPAL Final 9 MeV
Ambreesh Gupta Fermilab Wine & Cheese 21
LEP Beam Energy Cross Check with Data
LEP beam energy can be estimated using radiavtive return events
- Z mass precisely known
- Measured mass in radiative events sensitive to beam energy Result consistent with zero within experimental errors
Ambreesh Gupta Fermilab Wine & Cheese 22
Detector
MC modeled to represent data;
Disagreements Systematic error
Systematics from MC Modeling
Main Sources- QED/EW radiative effects
- Detector Modeling
- Hadronisation Modeling
- Background Modeling
- Final State Interaction
Ambreesh Gupta Fermilab Wine & Cheese 23
• KoralW’s O(3) implementation adequate,
but misses
- WSR
- interference between ISR,WSR & FSR
• KandY includes
- O() corrections
- Screened Coulomb Correction
Error ~ 7 MeV
Photon Radiation
Ambreesh Gupta Fermilab Wine & Cheese 24
• Z0 calibration data recorded annually provides
a control sample of leptons and jets (~ 45 GeV).
• Data/Mc comparison used to estimate corrections for
- Jet/Lepton energy scale/resolution
- Jet/Lepton energy linearity
- Jet/Lepton angular resolution/biases
- Jet mass
• Error is assigned from the error on correction
qqlv qqqq Combined
10 MeV 8 MeV 10 MeV
Detector Systematics
• Raw Corrected
LEP Combined:
Jet energy scale
Jet energy resolution
Ambreesh Gupta Fermilab Wine & Cheese 25
Detector Systematics: Breakdown OPAL
Ambreesh Gupta Fermilab Wine & Cheese 26
• MC programs (JETSET,HERWIG,ARIADNE) model production of hadrons
but difference in particles and their distributions
• The difference interplays with detector response
- particle assignment to jets
- cuts applied to low momentum particles
- low resolution for neutral particles
- assumptions made on particle masses at reco.
• JETSET used by all LEP experiment with parameters tuned with
Z peak data
systematic shift estimated from shift with other hadronization models.
qqlv qqqq Combined
13 MeV 19 MeV 14 MeV
LEP Combined:
Hadronization Modeling
Ambreesh Gupta Fermilab Wine & Cheese 27
Final State Interactions
Two known sources that could potentially bias W mass
and width measurement
1. Color Reconnection
- color flow between W’s could bias their masses
- only phenomenological models exist.
- Most sensitive variable to CR is W mass itself
2. Bose-Einstein Correlation.
- coherently produced identical pions are closer in
phase space.
- BE correlation between decay products of same
W established
- Does the effect exist between W’s?
The Basic Problem: If products of hadronically decaying W’s (~0.1 fm)
interact before hadronization (~1.0 fm) Can create a mass bias.
Ambreesh Gupta Fermilab Wine & Cheese 28
Color reconnection
Only phenomenological models exist. - SK1 model produces largest shift CR strength parameter (ki)
LEP experiments estimate effect of color reconnection Measure particle flow in the inter-jet regions of the W’s - Extreme values of CR disfavored by data but it does not rule out CR - A 68% upper limit on ki is used to set a data driven uncertainty on W mass. - Combined LEP value of ki = 2.13 For this Reco. Prob., CR error ~ 120 MeV (OPAL)
Ambreesh Gupta Fermilab Wine & Cheese 29
A,D,L,O use varations of below
- OPAL Uses P-cut 2.5 GeV for
qqqq
- ~18% loss in statistics.
- Much reduced CR systematics
125 41 MeV (ki =2.3) OPAL
- A worthwhile tradeoff!
- ALEPH 28 MeV, L3 38 MeV
Cuts and Cones: Reducing CR effect CR affects mostly soft particles between
jets changes jet direction Re-calculate Jet direction
1. Within cone of radius R
2. Cut on particle momentum P
3. Weighted particle momentum |P|
Final CR error in qqqq 35 MeV
Ambreesh Gupta Fermilab Wine & Cheese 30
2.5 GeV P-cut to redefine jet direction also
reduces BEC W mass bias
- OPAL (default) 46 MeV (P-cut) 24 MeV. LEP experiments have measured BEC between
W’s
- Using “mixing method”
- A,D,L,O: only a fraction of Full BEC seen in
data (0.1713)
BEC in WW events
A 68% upper limit on BEC fraction seen in data
(OPAL), used to set W mass systematics
ΔMW = ( 0.33 + 044) ΔMW (Full BEC) = 19 MeV
L3 18 MeV, ALEPH 2 MeVFinal BEC error in qqqq 7 MeV
Ambreesh Gupta Fermilab Wine & Cheese 31
Results: qqqq and qqlv channels
QuickTime™ and aTIFF (LZW) decompressor
are needed to see this picture.
QuickTime™ and aTIFF (LZW) decompressor
are needed to see this picture.
Ambreesh Gupta Fermilab Wine & Cheese 32
Results: LEP W mass and WidthmW (LEP) = 80.376 ± 0.033 GeV W (LEP) = 2.196 ± 0.083 GeV
Ambreesh Gupta Fermilab Wine & Cheese 33
W’s as Calibration Sample at LHC“Yesterdays sensation is today’s calibration
and tomorrows background” - Telegdi
- W’s from top decay are foreseen to provide the
absolute jet scale.
- Fast simulation studies in the past showed feasibility
- Select samples with a four jets and lepton
(electron,muon) with two jets b-tagged.
- estimated 45K events from 10 fb-1
- Cross check with Z/ +Jet sample
- Sattistics not the issue but understanding
the physics of the events.
Ambreesh Gupta Fermilab Wine & Cheese 34
Summary
Final results from all the four LEP experiments Final LEP combination will use combined FSI error
Much reduced FSI error in final results A new preliminary LEP combination
Total LEP W mass uncertainty decreased to 33 MeV
It took about five years after LEP shut down to get final W mass
results from all the four experiment. Now it is up to Tevatron to
better the W mass precision before LHC turns on.