The Measurement of the W The Measurement of the W mass mass at LEP at LEP XXXIX Recontres de Moriond, April 2004 Ann Moutoussi, CERN
The Measurement of the W The Measurement of the W massmassat LEPat LEP
XXXIX Recontres de Moriond,
April 2004
Ann Moutoussi, CERN
OutlineOutline
Introduction: the Standard Model and MW
Measurement of MW: Direct reconstruction
Systematic errors
QCD related errors
Results and conclusions
Mw within the Standard ModelMw within the Standard Model
Mw can be computed
at Born level from , Mz , GF
Higher order radiative corrections involve Mt, MH:
e.g
O(, sMz , GF , Mtop, Mhiggs)
Precision measurements of Mw check the prediction
If consistent > SM still OK,
Use measurements to predict Mhiggs
If not consistent > Hints for New Physics?
tW W
HW W
The LEP goal for MThe LEP goal for MWW
PP Colliders 80.454 ±0.059GeV/c2
EW fits (LEP/SLD) 80.373±0.033GeV/c2
EW Fits (LEP/SLD)
with Mtop 80.378±0.023GeV/c2
LEP Goal:precision of ~40 MeV …Very difficult task..
First phase optimise statistical power of analysis
Last years fight known and new systematics!
No update since last winter
ee++ee- - WW++WW--
W decay modes: Leptonic: W l(32%) Hadronic: W qq
(67%)
q
q
q
q
q q
Semileptonic (qqSemileptonic (qqll)) Hadronic (4q) Hadronic (4q) LeptonicLeptonic
44% 46%Low Mw sensitivity
~40K events in total
MW measurementMW measurement
Identify and best reconstruct leptons (e,m,tau) Best Cluster jets and measure energy and direction
Event-by-event reconstruction of the invariant masses of W decay products
Statistical sensitivity limited by resolution of jet/lepton energies and momenta
Can improve resolutions using the knowledge of ECM and
Energy-Momentum conservation
optionally: equal W mass constraint
WW
4
1
4
1
,0 ,
mm
psEi
ii
i
Mreco
Mw Reconstruction(1) qqqqMw Reconstruction(1) qqqq
Minus :− Particle to Jet association
mixing between jets from different Ws smearing of Mw distribution
Statistics: Optimise clustering algorithms
− Jet-Jet association to a W Wrong pairLoss of all Mw information
Statistics: Optimise pairing algorithms(~85% correct pairing)
Plus:
No unmeasured particles,
Fully constrained system
Minus :− Neutrino
3 3 unknowns only 2 constrains fit
Plus:
Only two jets
no loss of information due to particle mixing or combinatorial bkg
Considered golden channel
Mw Reconstruction(2) qq Mw Reconstruction(2) qq l
Reconstructed MwReconstructed Mw
Mreco still far from underlying Mw distribution
After the kinematic Fit: Mreco
True Mw
Mjet
Mreco
qq l
W Mass extractionW Mass extraction
Assume MC events are identical to data, except from Mw!
Discrepancies between data and MC are sources of systematic errors
In practice, only one MC sample is generated, at a reference value
MWref. Predictions at other values of MW are obtained by re-weighting
the events
To relate Mreco to Mw use Monte Carlo events•Fit Mreco with analytical function(eg BW) and then correct it using MC
or
•Compare Mreco distribution to MC predictions at different Mw values
Systematics
Systematics(largest)Systematics(largest)
Source Currently/MeV LEP Energy determination 17
Detector Simulation Jet & Leptons energy/direction 15
QCD simulation
Jet Fragmentation 18
Jet-Jet interactions(4q) 93
(Expected final statistical error for LEP 25 MeV)
Unacceptable!!
parton shower (large Q2, pQCD)
Fragmentation (quarks hadrons):
Simulation of a MC event(1)Simulation of a MC event(1)
hadronisation (phenomenological)
e+
e-
W+
W-
q
_qq
_q
Available models: Jetset, Herwig, Ariadne.
All models: need to be tuned to data (generally Z qq, LEP1). Simulate Data ~as well/bad!
Jetset globaly better
used as Reference MC from all LEP experiments
parton shower (large Q2, pQCD)
Fragmentation (quarks hadrons):
Simulation of a MC event(2)Simulation of a MC event(2)
Hard process: e+e-4q
e+
e-
W+
W-
q
_qq
_q
Interconnection effects
Bose-Einstein correlations: momenta of identical bosons tend to be correlated.
d~0.1 fm
Colour reconnection: hadronic interaction between W decays
•d(W+,W-) < 1 fm
hadronisation (phenomenological)
Not included in reference MC
Bose-Einstein Correlations Bose-Einstein Correlations (BEC)(BEC)
Intra-W :BEI not relevant for Mreco
W1
W2
Bettwen-W’s:BEB: could cause wrong particle-dijet association
Mw shifts ~ 35 MeV(LUBOEI)
Main Observable: distance in momentum space
between pairs of charged pions: Q2=(pi-pj)2
Any evidence for such effects?
Look for BE in data
Observation BEC in WObservation BEC in W++WW-- eventsevents
Inter W, BEI confirmed
Between W’s, BEB, disfavoured
MW down from ~35 to ~15 MeV
Final
BEBBEI
eg
CR modelsCR models
Based on the JETSET string model: SK1:
• it has a free parameter I controlling
the reconnection probability P
Based on Ariadne, AR2: Based on HERWIG (Herwig-CR)
P=1 MW ~400MeV
P=0.5 MW ~115MeV
P=0.3 MW ~ 50MeV
MW ~ 40MeV
MW ~ 70MeV
MW Far too large!
Any evidence for such effects/models?
Look for CR effects in data
The particle flow analysisThe particle flow analysis Most CR models predict a
modified particle flow in W+W- events:
CR:
No CR:
W-
W+
W-
W+
The ratio of particle flow between the inter and intra-W regions is built:
(A + B) / (C + D)
A
B
C
D
•Data
-SK1(extreme parameter)-Jetset
Measurement sensitive only to extreme scenarios, i.e SK1 with high CR probability and not so to Herwig, Ariadne
LEP results from particle flowLEP results from particle flow
preferred value: kI=1.18,
P~0.5 kI value excluded at 1
value used for CR studies and MW evaluation= ~100 MeV!
Do something to make analyses more robust!
Fit LEP measurement
for free parameter k (CR P)
(CR P)
Towards a less CR sensitive analysis:
The logic The logic Interconnection effects mainly occur in the
inter-W region and between soft particles
Proposed solution: modify clustering algorithm to dismiss information from those particles. “purer” information
loss of statistical precision
Many variations of jet algorithms (cones, pcuts) have been considered
aiming for the best combination of
Robustness against reconnection effects with minimal information loss
Reduction of Reduction of MMW W
MW (MeV)
Model Standard R=0.5rad
SK1, kI~2 ~115 ~50
Herwig ~40 ~15
Good reduction factors
for all available models!
e.g for R=0.5, 2.3-2.6 smaller MMWW
with ~25% increase of stat. error:
Algorithms simple and intuitive
measurement less sensitive to CR
independent of specific model implementation
A by-product: Measure CR?A by-product: Measure CR?
The difference between MW measured with cone/pcut
and standard analyses (MC-S) is sensitive to CR effects:
DELPHI, Cone algorithm R=0.5
e.g
DELPHI preliminary:
Exclude extreme scenarios.
Minimum at ~1.3, P~0.5
Results
ResultsResults
80.411±0.032(stat) ±0.030(syst)GeV/c2
80.420±0.035(stat) ±0.101(syst)GeV/c2
(Weight of qqqq in combination: 0.09%)
qq qq l
qqqqqqqq
Mw=80.412±0.042 GeV/c2LEP Combination
Mw GeV
±
±
±
±
±
±
80.426±0.034
80.378±0.023
Mw and Mtop, MHiggsMw and Mtop, MHiggs
mH
MwMtop GeV
Mw wants a low Higgs Mass...
After all this work….After all this work….
Ongoing LEP efforts to find optimal jet clustering and make qqqq measurement robust against CR
If all experiments use them Total error in hadronic channel: ~110 ~60 MeV. Total error from ~42 to ~39 MeV Weight of hadronic channel in combination: 0.09% 0.29%.*Learn something about Final State Interactions too...*
Detector Systematics still an issue after all these years..
Final values for Summer?!?!
ResultsResults
Mw=80.420±0.035(stat) ±0.101(syst)GeV/c2
Mw=80.411±0.032(stat) ±0.030(syst)GeV/c2
Combined: Mw=80.412±0.042 GeV/c2
qq qq l qqqq qqqq
Weight of qqqq in combination: 0.05%
Detector SimulationDetector Simulation
As measurement is calibrated using MC Systematic errors related to the detector arise from discrepancies in the detector simulation.
Most effort devoted to Jet Energy, Mass and Direction:
Jet Energy (mass, multiplicity,etc) calibrated, checked
and MC tuned using Z qq events:Clean enviroment, Ebeam~Ejet, Jets back to back well separated
e.g Compare Ejet/Ebeam
as a function of polar angle , for Data and MC (ratio) for total energy, Ejet
And for individual types of particles (Echarged, Ephotons, etc)
Jet Energy Simulation Jet Energy Simulation
(Ejet/Ebeam)Data/MC
cos
2000 publication Preliminary results Towards final results
Better: simulation of Calorimeter endcaps, photon energy calibration,
treatment of small calorimeter measurements ,etc etc
Small changes on Mw ~ size of calorimeter systematic
(Ejet/Ebeam)Data/MC
cos
Data-MC=
-0.024±0.007
rad
Data
MC
qq qq e
Jet Direction simulation Jet Direction simulation Test done with W events:
Compare Data and MC = neutral-Chargerd , being the dijet angle Jet1-Jet2
Charged1
•Collecting the full statistics allowed
relevant sensitivity
•qq qq e bad surprise
Data different from MC
by 24mrad
neutChar
Jet1
Jet2Charged2
The electron channel: qq The electron channel: qq e
Angle to lepton/degrees
# Data
MC
Particles associated to a jet
qq qq e
What could make neutral dijet Q be more open in Data than in MC?
Look near the electron…..
•EM shower of v.energetic electrons
not well simulated.
•Existing algorithm to collect electrons cloud not adequate.
New electron reconstruction
Mw from qq qq e moved by ~100 MeV…
WW production at LEPWW production at LEP
1. Theoretical precision ~0.5% Thanks to 2000 calculations RACOONWW, YFSWW with improved O(a)
corrections
2. LEP measurement precision ~1%
Very good agreement
Example: Jet Mass and Baryon Example: Jet Mass and Baryon ## Jet Mass
Jetset Herwig Data
Identical W 2q events, Hadronised with Jetset/Herwig
Study Mjet1, Mjet12, Mreco Vs (No of neutrons)
Jet Mass enters into Dijet-Mass (Mjet12) and also
shows some discrepancy between Data and MC
0 2 4 6 8 10
Mjet(12)/GeV
neutron
8
4
0
0 2 4 6 8 10
Effect much smaller
but ~20MeV
neutron
Mreco/GeV
0 2 4 6 8 10
8
4
0
Mjet1/GeV
neutron
5
2
0
LEP EnergyLEP Energy
At LEP2:Error mainly from extrapolation.
Ebeam~20MeV (E/E~10-4!)
mW~17MeVmW~17MeV
Ebeam measured from total bending fieldtotal bending field
Calibrated with resonant depolarizationresonant depolarization:spin precession freq Ebeam
intrinsic resolution ~ 200keV !!
only works up to 60GeV extrapolation
Kinematic fit the absolute energy/momentum scale is calibrated by the LEP beam energy measurement
…and will stay ~there
ALEPH: Energy resolutionALEPH: Energy resolution Energy resolution for a
calorimeter object adding ECAL + HCAL is:
E/GeV0.6E)(
E/GeV2.1)E( = 6 GeV
Peak=90.5GeV
Total Visible Energy (GeV)
@ 91GeV ecm Z qq
Take into account particle ID to:•use momentum measurement of
tracks pointing to calorimeter objects
•avoid double counting of energy.
•apply specific calibrations.
build new objects with:
ALEPH: Jet DirectionALEPH: Jet Direction
Jet and resolution
=18mrad
=19mrad
Jet direction information is based on tracks,
addition of neutral objects improves resolution by 15%
Jet Direction simulation(1)Jet Direction simulation(1)
Still at the Z pole:
Difficult, as no refference (like Ebeam)
Tests rely on “correct” position of tracks and check calorimeter objects
by comparing neutral toChargerd as a function of jet
neut
Charged
Z axis
cos
mra
d
Charged -neutral ) Data/MC
No significant effect,
small systematic error
But..
The electron channel: qq The electron channel: qq e
Particles near an electron
Angle to lepton/degrees
ee++ee- - ee++ee--#
What is all this stuff there? Look near the electron…..at bhabha events..
QCD models at LEPQCD models at LEP
Model Parton Shower Hadronisation
JETSET abcString
ARIADNE CDM
HERWIG abc Cluster
All models: need to be tuned to data
(generally Z qq, LEP1). Simulate Data ~as well/bad!
Available Models:
Jetset somewhat better
used as Reference MC from all LEP experiments
Specific systematics for Specific systematics for cones?cones? Cone and standard analysis can have different
sensitivity to fragmentation:cone could be more sensitive to angular distribution of
particles inside jet
Use Z qq events from LEP1.
Data and MC Energy and multiplicity distributions were compared as a function of angle to jet axis
No indications of new sources of systematics
Data/JETSETHERWIG/JETSET
HERWIG
DataJETSET
Angular distributionsAngular distributions
Jet Energy: Velocity:
HERWIG
DataJETSET
Data/JETSETHERWIG/JETSET
Inter-jet angle in WInter-jet angle in W++WW-- events events
M212 ~ 2E1E2(1 - cos )
Z qq events too different semileptonic W+W- events used. independent sample free from CR effects
DataJETSET
Variable checked:
SC
No indications of new sources of systematics
S - C
For Data and Jetset
Conclusions(2)Conclusions(2)
Statistical errors exceeded all expectations
(analyses really pushed to the limit!)
Systematic errors dominantA lot of effort invested to fight against the larger known
(eg Colour Recconection) lead to more understanding of the causes and the design of promisingly more robust analyses
Detector Systematics. The precision required from Mw exceeds this of all previous analyses. Jets and the simulation (especially of neutral part) cannot rely on LEP1, more detail needed (10MeV!)
Effort put on guessing those unexpected systematics!
FragmentationFragmentation
“Traditionally”: Compare different models: (various X)pass them through full analysis : Max Mw ~20 MeV (Jetset-Herwig)
Latest work:
But.. Mw is due to X between Data & reference MC(Jetset)
1. Identify fragmentation variable, X, with significant dMw/dx
2. Estimate X(Data-MC) at some control sample, eg Z events
3. Propagate dX(Data-MC) in refference MC Mass
distribution Mw
Method for MW
measurement
Introduction
The Standard Model and
Mw
QCD effects on MW
FragmentationFragmentation If all particles are detected and associated to Ws perfectly,
discrepancies in fragmentation do not bias MW measurement.
Biases come from interplays:
E, p spectra
Baryon rates (e.g
n,p)
Thresholds
charged ->
mneutrals ->
m
Angular size of jetsAcceptance
Jet algorithms
Discrepancies
MC-reality on fragmentation x
Detector
fD (X)
Reconstructio
n fA (XE)
e.g
W W event selectionW W event selection Semileptonic channel (qql
2 jets 1 isolated lepton, 1 neutrino: missing E&P
Efficiency ~70% Purity ~90-95% main bkg We, qq(
Hadronic channel (qqqq) ( 4 jets
• large multiplicity• spherical topology
low missing E&P Efficiency ~80% Purity ~85% main bkg qq(
Statistics: Use multivariable analyses (e.g neural networks, even for qql events!)