W mass and width at CDF

1
W mass and width at CDF

• Emily Nurse
• University of Manchester Seminar
• 30th March 2006

2
Outline

• EWK precision measurements
• Unstable particles mass and width
• The W mass and width
• Motivation
• Current status
• CDF width measurement
• CDF mass measurement (brief summary)
• Summary and projections

3
Introduction EWK measurements

• The Standard Model contains free parameters that
  must be found from experiment.
• Relationships between these parameters are given
  in the theory.
• Measuring the parameters to a high precision is a
  stringent test of the theory and deviations from
  expected values indicate physics Beyond The
  Standard Model.

4
Introduction unstable particles mass and width

• The properties of short lived particles can be
  measured by reconstructing the invariant mass of
  their decay products.
• This distribution peaks at the particle mass and
• has a finite intrinsic width.
• The width is due to the Heisenberg uncertainty
• principle ?E?t ?
• The shorter the lifetime the larger the width.

5

• Z bosons decaying to neutrinos cannot be
  detected.
• This decay mode will, however, contribute to the
  width.
• The LEP experiment measurements of the Z width
  gives the number of neutrino species.

6
W mass and width motivation
on-shell mass scheme
Measured to 0.014
Measured to 0.004
Measured to 0.0009
DrMt2
Drln MH
Plot from LEP EWK working group web page
7
W mass and width current status
direct
Plots from LEP EWK working group web page
indirect
cf/ Z-Boson mass 91.1876 0.0021 GeV
width 2.4952 0.0023 GeV
8
The Tevatron
The highest energy accelerator in the world,
collides protons with antiprotons at a centre of
mass energy of 1.96 TeV.
Each experiment currently has 1.2 fb-1 on
tape. Aim to have 8 fb-1 by 2009.
9
Collider Detector at Fermilab
10
W and Z production
P
e , ? , ? , q
q
W
q
?e , ?? ,?? , q
P
The large masses (100 GeV ) of W and Z bosons
means their decay products will have large
pT. The electron and muon channels are used to
measure W properties, due to their clean
experimental signature.
11
CDF W width analysis
current analysis based on 350 pb-1
note this analysis is work in progress!
12
Analysis strategy

• Ideally, the mass and width of the W would be
  found from the invariant mass distribution of its
  decay products (line-shape).
• We cannot reconstruct the W invariant mass
  distribution since the neutrino escapes
  detection.
• Instead we reconstruct the transverse mass
• MT 2 pTl pT? (1 - cos??)
• pT? found by summing total transverse energy in
  detector (? calorimeter towers) lepton to give
  the missing ET.

?l
??
l ?
13
Analysis strategy

• Width and mass found by fitting MT in data to
  that in Monte Carlo.
• The line-shape of the W is described by a
  Breit-Wigner distribution with a scale dependent
  width

x1x2s
14
Analysis strategy

• Simulate MT distributions using Monte Carlo
  event generator with various blinded input
  width values.
• Normalise to data in the low MT region.
• Fit to data in the high MT region (exactly
  where is a trade off between systematics and
  statistics).
• A fast simulation is required (rather than full
  detector simulation).
• Z events used to tune the detector response to
  particles.

Same principle for W mass, with fit region 50 -
100 GeV
15
The data

• Select 4 data samples
• Z??? and Z?ee (control samples).
• W??? and W?e?

Electrons identified in COT and EM calorimeter.
Muons identified in COT, calorimeters (MIP
signature) and muon detectors.
16
2 electrons, ET gt 25 GeV
2 muons, pT gt 25 GeV
?e lt 1
?? lt 1
1 electron, ET gt 25 GeV ETmiss gt 25 GeV
1 muon, pT gt 25 GeV ETmiss gt 25 GeV
17
The simulation
UA2 program

• Toygen event generator produces Z???, W???, Z?ee
  and W?e? samples.
• Purely electroweak (no gluon radiation, i.e.
  pT(boson) 0 GeV).
• pT(Z) is added from a functional form (fit
  to pT(Z) data). A theoretical calculation
  converts pT(Z)?pT(W).
• QED radiation - Toygen is interfaced with Berends
  and Kleiss program one photon FSR correction.

18

• Simulation of detector
• Simple tracking helix extrapolation through
  detector with calorimeter and muon geometry
  simulated.
• The ionisation energy loss of muon tracks in
  central trackers taken from the Bethe-Bloch
  equation.
• Electron bremsstrahlung and photon conversion
  within central trackers simulated.
• Clustering of electrons/photons in calorimeter
  simulated.

nIter 0
3 etc.
1
2
19

• Lepton id and trigger efficiencies input from
  data.
• COT momentum scale and resolution found by
  fitting to M?? in Z??? events.
• Calorimeter energy scale and resolution found by
  fitting to Mee in Z?ee events.
• Calorimeter energy scale and resolution also
  found from E/P in W?e? events.
• Recoil (? non-lepton calorimeter towers)
  distribution modeled by tuning to Z??? and Z?ee
  data.

E energy of electron measured in calorimeter. P
momentum of electron measured in central tracker.
20
Sources of systematic uncertainty(anything
affecting the MT distribution)

• PDFs and W pT distribution
• QED corrections
• Lepton energy scales and resolutions
• ETmiss distribution (recoil model)
• Backgrounds in W samples

21
Parton Distribution Functions

• PDFs are used as inputs to the MC to give the
  momentum distributions of partons within the
  incoming protons.
• The PDFs have an uncertainty associated with them
  due to the uncertainty in the many datasets used
  to fit them.
• This uncertainty gives an uncertainty on the MT
  distribution and hence the width determination.
• Use the CTEQ6 and MRST PDF error sets to obtain
  the uncertainty on the W width
• CTEQ6 25 MeV, -25 MeV
• MRST 14 MeV, -10 MeV

22
Z and W pT

• The Z pT is parameterised by an ad-hoc functional
  form
• The parameters are found by fitting
• to the Z data.
• A theoretical calculation
• converts this to a W pT distribution.
• The bosons (generated with 0 pT)
• obtain a pT from this functional form.

P1 0.564716 0.0235054 P2 5.04421
0.247411 P3 16.1793 1.00640 P4 0.921681
0.0616547
23
Lepton momentum/energy scales and resolutions

• Since we do not model the complete CDF
  detector and event reconstruction we must smear
  the true lepton momentum (as measured in the
  COT) and energy (as measured in calorimeter). The
  smearing parameters are found by tuning to the
  data.

24
Tuning method (e.g. track resolution)

• Get M?? distribution from Z-gt?? events in data
  in region
• 82 lt M?? lt 100 GeV.
• Simulate 100 million Z-gt?? events. Make many M??
  templates with different values for momentum
  resolution constant.
• Plot ?2 vs resolution. Fit to a polynomial and
  minimise!

?2 - ?2min
1
resolution
25
Track momentum
scale pTmeas 0.9986 ?
pT resolution ?(1/pT) 0.00055 ?
0.006 / pT
curvature term
multiple scattering term
26
Calorimeter energy
scale Emeas 1.0030 ?
E resolution ?(E) / E 13.5 / ET ?
2.2
stochastic term
constant term
scale 1.0030 ? 0.0006 kappa 2.22 ? 0.14
?2 18.5 / 18
27
Calorimeter energy E/P
e? mom in COT
e? energy in calorimeter
Calorimeter energy and track momentum resolutions
Energy leakage from electron calorimeter tower
(not yet simulated)
Momentum loss in trackers through bremsstrahlung
scale 1.0030 ? 0.0006 kappa (2.22 ? 0.14)

scales consistent resolutions have 4? different!
28
Recoil Model

• Recoil (U) comes from multiple interactions,
  underlying event and initial state gluon
  radiation.
• Defined as the sum of all non-lepton calorimeter
  towers.
• Measured and modeled in Z events and applied to W
  events.

29
Z?ee events

• U in Z events resolved into 2 directions
• U1 anti-parallel to the Z pT
• U2 perpendicular to Z pT
• U1 and U2 are gaussian in bins of Z pT.
• ltU1gt a bpT cpT2
• ltU2gt 0
• ?(U1) d epT f?(ET)g
• ?(U2) h?(ET)i

The recoil is then added in this way to the
simulation for both Z and W events. Note this
is ongoing work, the simulation currently uses a
Run I model.
30
Efficiencies
measured in the data and applied to MC as a
function of ?
31
Potential backgrounds to W events
We need to know MT distribution of backgrounds as
well as overall normalisation!!
electroweak
Z ?ll
W ???

• One lepton lost
• ETmiss due to
• missing lepton

• ? decays to e/?
• intrinsic ETmiss

Background suppression veto on additional high
pT, isolated track
Residual background run full event selections on
signal and background MC samples passed through
simulation of CDF detector to get fractional
background. muon channel Z??? 4.7
W ??? 1.9
32
di-jet
Background suppression total recoil in event lt
20 GeV

• Jet contains /fakes a lepton
• ETmiss from
• misconstruction

Residual background Found from difference in
isolation distributions in W and Z events. muon
channel 0.5, electron channel 1.4
cosmic muons
Background suppression Cosmic tagger takes a
seed track and searches for additional track hits
on the other side of the interaction point.
.
.
.
.
.
.
.
.
.
.
Residual background muon channel
negligible electron channel N/A
.
.
33
kaon decay
Background suppression Cut on track ?2 and d0
(impact parameter).

• Kaon decays to a ?? pair when passing through
  COT.
• kink in track can give a fake high pT and ETmiss.

Residual background Found by fitting the d0
distribution in W ??? events to that in Z??? W
??? kaon (with fraction of kaon varying).
muon channel 1.00 ? 0.19
electron channel N/A
Tracks with no silicon hits
34
Muon channel backgrounds
1 kaon background is probably too much for us!
Muon channel transverse mass distributions
35
Width measurement summary

• Fast simulation modeling the detector reasonably
  well.
• Work still required to ensure a good
  understanding of some effects
• electron energy resolution
• recoil model
• Need to determine systematic uncertainties for
  all the effects discussed.
• When this is done fits for the width will be
  un-blinded.

36
Width measurement estimated uncertainties

• Run I analysis achieved 130 MeV with 110 pb-1
  (one third of our luminosity).
• A naïve estimate of most systematics can be
  obtained by scaling Run I numbers by increased
  statistics (since they are controlled by Z
  statistics).
• Some systematics are not determined by Z
  statistics (e.g. QED corrections).
• Estimate 80 MeV with this dataset with dominant
  errors from statistics, recoil and backgrounds.
• In reality may be higher

37
W mass status

• Current analysis uses 200 pb-1.
• Very similar to the width analysis, but with 50 -
  100 GeV fit region.
• Some effects are more important and must be
  better understood (e.g. track momentum scale
  determined using J/? and ? decays to muons, with
  Z events as a cross check). Therefore their
  simulation is more sophisticated than ours.
• Use Resbos Monte Carlo event generator interfaced
  with Wgrad for QED corrections.

38
W mass status
Fits blinded with additive offsets
110 pb-1
current work
Total uncertainty 76 MeV (Run 1b 79 MeV)
39
Projections
W width to a similar level of precision
40
Outlook

• Hope to have these first round measurements out
  by the summer(?)
• After that the plan is join forces on the W mass
  1 fb-1 challenge.
• Watch this space.

41
Back-up slides
42
(No Transcript)