New Results from D0 on the W width, charge asymmetry and on gauge couplings

Wine & Cheese, FNAL 1

New Results from D0 on the W width, charge asymmetry and on gauge couplings

Sarah Eno (U. Maryland)for the

D0 Collaboration

4 Sep 2009


Outline

W discovered in 1983. 26 years of W physics!How well do we know the W?

• Updated measurement of the muon charge asymmetry from W decays (4.9 fb-1) (How is the W made?)

• New limits on trilinear gauge boson couplings (0.7-1 fb-1) (How well does it play with its siblings?)

• W width (1 fb-1) (What’s its life expectancy?)

4 Sep 2009Run IIa Run IIb

D0 Detector

3Junjie Zhu

Silicon Microstrip Tracker (SMT) |η|<3 Central Fiber Tracker (CFT) |η|<2 2 T magnetic field

Liquid-argon sampling calorimetersCentral (CC) |η| < 1 and Endcap (EC)Coverage: |η| < 4.2

Muon systemDrift chambers and scintillator counters1.8 T toroids|η|<2


Data Samples

4 Sep 2009

Many, many thanks to the accelerator division!!


W

4 Sep 2009

How are W’s made?


Muon asymmetry

4 Sep 2009

u quark momentum distribution (in proton) harder than d. W+ tends to go along proton direction, the W- in the antiproton direction.

A(yW ) =dσ(W +) / dyW −dσ(W −) / dyWdσ(W +) / dyW + dσ(W −) / dyW

Y =ln(

σ⋅xaMW

) xaxb =M2 / σ Y =

12ln(

E + PLE −PL

)⇒ Y=12ln(

xaxb)

At Tevatron energies, for W’s 0.0017<x<1 at LO.

W rapidity

W a

sym

met

ryarXiv: 0901.0002

x


Muon asymmetry

4 Sep 2009

Since can not reconstruct pZ of the neutrino, measure muon charge asymmetry instead.*

A(η)=dσ(μ+) / dη−dσ(μ−) / dηdσ(μ+) / dη + dσ(μ−) / dη

Muon asymmetry influenced by W rapidity asymmetry but also by polarization and left-handed couplings of the W and the V-A structure of the W decay.

Cartoon stolen from CDF web site

* For a method of indirectly reconstruction the rapidity, see CDF, Phys. Rev. Lett. 102, 181801 (2009)


Muon asymmetry

4 Sep 2009

Muon asymmetry is more similar to W asymmetry for high PT muons from low PT W’s.


Muon charge asymmetry

4 Sep 2009

New D0 result, using 4.9 fb-1, updates previous results. Most recent published results are:

• D0: Measurement of the Muon Charge Asymmetry from W Boson Decays, Phys. Rev. D77, 011106 (2008), 0.3 fb-1

• D0: Measurement of the Electron Charge Asymmetry, Phys. Rev. Lett 101, 211801 (2008), 0.7 fb-1

• CDF: W boson charge asymetry vs W rapidity, Phys. Rev. Lett. 102, 181801 (2009), 1 fb-1

• CDF: lepton charge asymmetry, Phys. Rev. D71, 051104, 0.17 fb-1


Selection

4 Sep 2009

ηdet < 1.6 ⇒ 594203 W → μν

1.6 < η det < 2.0 ⇒ 5950 W → μν

• an appropriate muon trigger (limited to η < 1.6 for Run IIb)

• a muon with PT > 20 GeV with η < 1.6 (Run IIb) or η < 2.0 (Run IIa)

• muon should be isolated in the tracker PTtracks < 2.5 GeV

ΔR<0.5∑

• muon should be isolated in the calorimeter ET < 2.5 GeV0.1<ΔR<0.5

∑• ET > 20 GeV• transverse mass (MT ) > 40 GeV• timing cuts to reject cosmics• remove Z → μμ by vetoing events with second muon


Challenges

4 Sep 2009

• As long as they are charge independent, efficiencies, acceptances, and luminosity cancel in the ratio (frequent reversal of solenoid and toroid polarities helps)• Backgrounds can dilute asymmetry• Charge mis-identification is a potential problem that can dilute asymmetry• Need to correct for momentum smearing since asymmetry depends on muon PT


Backgrounds

4 Sep 2009

Background Run IIa Run IIb

multijet 2±0.1% 2.4±0.1%

W->τν 3±0.02% 3±0.03%

Z->μμ 2.7±0.1% 3.1±0.01%

Z->ττ 0.16±0.001% 0.18±0.002%

• Multijet background estimated using matrix method based on isolation. Background efficiency (εB(η,PT)) estimated using events with low MET (<10 GeV) and a jet with PT>10 GeV. Systematics on εB estimated by varying cuts used to reduce W contamination.• W/Z backgrounds estimated from MC (PYTHIA) normalized to NNLO cross section. Monte Carlo statistics dominant source of uncertainty.• Background fractions binned in eta and muon PT. Only 1 bin in eta for eta>1.6


Charge misidentification

4 Sep 2009

Run IIa, 3 same-sign Z’s out of 48452Run IIb, 14 same-sign Z’s out of 120417

negligibleMuon

Jet

Muon

Track


Momentum smearing

4 Sep 2009

Because of finite momentum resolution, bins in reconstructed PT contain events from other PT bins. Since the shape of this asymmetry depends on PT, correction is needed.Uncertainty determined by varying the momentum resolution within uncertainties.


Results

4 Sep 2009

Data sample is large enough that errors on asymmetry are smaller than spread from PDF uncertainties. Will be useful for global PDF fits and reduce PDF uncertainty on W mass and width.


Results

4 Sep 2009

For muon PT> 35 GeV.

Systematic uncertainties completely dominated by muon momentum resolution correction.


W

4 Sep 2009

How well does the W play with its siblings?


Charged triple gauge couplings

Charged Triple Gauge Couplings (TGC)Probed by WW, WZ, and Wγ production

4 Sep 2009

a(s) =a0

1+σL2

⎛⎝⎜

⎞⎠⎟2

•SM

• couplings that respect CP, SU(2)LxU(1)Y and EM gauge invariance

• assume equal couplings for ZWW and γWW respecting CP

all 0 except g1V =κ V =1

14 parameters at LO

L =2 TeV


New from D0

4 Sep 2009

• using the PT spectrum of the di-jet system from recent evidence forWW/ZW → jjlν

• combining several different previous measurements of chargedanomalous TGC (aTGC)

WW / WZ → ln jj WW → lnlnWγ → lnγWZ→ lnll

WZ final state currently only accessible at the Tevatron.


Evidence for WW or WZ to lνjj*

4 Sep 2009

• muon ( η < 2) or electron ( η < 1.1) with PT > 20 GeV• ET > 20 GeV• 2 jets with PT > 20 GeV and η < 2.5• at least 1 jet with PT > 30 GeV• MT > 35 GeV• Random Forest classifier with 13 kinematic variables• (dominant) W+jets background and signal fit from data simultaneously

σmeasured = 20.2 ± 2.5(stat) ± 3.6(sys) ± 1.2(lumi) pbσ SM = 16.1 ± 0.9 pb (Phys. Rev. D60, 113006 (1999))

Phys. Rev. Lett. 102, 161801 (2009)


Anomalous couplings

4 Sep 2009

• affect total cross section• affect kinematic distributions. The PT of the dijet system is particular sensitive.

L =2 TeVSU(2)LxU(1) conserving aTGC

WZ

cros

s sct

aTG

C/SM

WW

cro

ss sc

t aTG

C/SM

hadronic W PT

hadronic W rapidity

ΔΔ

λ

Δκ

Δκ


aTGC

4 Sep 2009

Simulation: SM events produced with PYTHIA reweighted according to generator-level PT and ΔR using MC@NLO-based weights. Anomalous couplings distributions are generated by reweighting SM predictions using fit to ratio of the (LO) HZW* generator with and without aTGC.

X is the PT of the dijet system

* Phys. Rev. D 41, 2113 (1990)


Data: WW or WZ to lνjj

4 Sep 2009


Limits

4 Sep 2009

equal couplings (γWW=ZWW)

SU(2)LxU(1)Y

L =2 TeV

SU(2)LxU(1)Y

SU(2)LxU(1)Y

See arXiv:0907.4398


aTCG combining channels

4 Sep 2009

Combine limits from 4 measurements by minimizing χ 2 , for diσtributionσ for σenσitive variableσ (σηown) for eaχη μ eaσureμ ent σiμ ultaneouσly, inχludinγ σyσteμ atiχσ aσ nuiσanχe paraμ eterσ

arXiv: 0907.4398: WZ and WW → lnjjarXiv: 0904.0673: WW → lnlnPηyσ. Rev. D 76, 111104(R) (2007): WZ→ lnllPηyσ. Rev. Lett. 100, 241805 (2008): Wγ

Wine & Cheese, FNAL

aTGC Limits combining channels

26

WZ→lvll

Wγ→lvγ

WW lvjj

WW→lvlv

4 Sep 2009


Uncertainties

4 Sep 2009

Systematics handled using methodology/code from W. Fisher, FERMILAB-TM-2386-E

Type I: affects only normalizationType II: can change shapes of kinematic distributions as well

Most important systematics are background cross sections and luminosity. Incorporating the systematic uncertainties degrades the limits by 30%.

aTGC Limits

28

Can be interpreted as measurements of the magnetic dipole and quadrupole moments.

4 Sep 2009 Wine & Cheese, FNAL

x2-x3 less sensitive to combined LEP results.Comparable sensitivity to an individual LEP exp.

arXiv:0907.4952

L =2 TeV


W

4 Sep 2009

What’s the W’s life expectancy?


W Width

4 Sep 2009

ΓW ≈ (3 + 2 fQCD )GF MW

3

6 2π(1+ δ SM ) = 2.089 ± 0.002 GeV

fQCD = 3(1+ α s (MW ) / π + 1.409(α s (MW ) / π )2 + ...

(Rosner, Phys. Rev. D49, 1363 (1994) Renton: arXiv : 0804.4779(2008), Denner: Fortsch. Phys. 41, 307 (1993))

Due to insensitivity to “Oblique” corrections, expected to agree with SM prediction almost regardless of new physics.

Current world average is : 2.050± 0.058 GeV (2.8% measurement)

Although one of the best-predicted, one of the least well-measured properties of the W.

Rosner et al.


New from D0

4 Sep 2009

New measurement with 1 fb-1 of data

• previous highest luminosity measurement from CDF (Phys. Rev. Lett. 100, 071801 (2008), 0.35 fb-1

•Using same data samples, MC simulations, and much of the same methodology as recent D0 W mass measurement, arXiv:0908.0766, submitted to Phys. Rev. Lett.

4 Sep 2009 Wine & Cheese, FNAL 32

W transverse mass

MT = 2ETeET

n (1−χoσDf )

Mass

width

Width, to LO, is proportional to the fraction of events at high MT


W basics

4 Sep 2009

Two objects are measured in the detector: Lepton & Hadronic recoil (u). Neutrino is vector sum of lepton and recoil.

MT = 2ETeET

n (1−χoσDf )

≈2ETe +u||

Event Selection


• an inclusive electron trigger• an electron with PT > 25 GeV • matching track with PT > 10 GeV and an SMT hit • shower shape consistent with electron

isolation (ER<0.4total / ER<0.2

EM ) − 1 < 0.15

EEM / E total > 0.9 • η < 1.05 and not close to module boundaries• ET > 25 GeV• uT < 15 GeV (avoid regions where production model has large uncertainties,reduce affect of uT and reduce backgrounds)• 50< MT < 200 GeV• vertex with z < 60 cm

499,830 W's (5272 with 100 < MT < 200 ΓeV )

Outline


Need Monte Carlo simulation to predict shapes of these observables for given width hypothesis

NLO event generator : DØ uses ResBos [Balazs, Yuan; Phys ReV D56, 5558] + Photos [Barberio, Was; Comp Phys Com 79, 291] for W/Z production and decay: O(108) events

+Parameterized detector model

Detector calibrationdata

Reweighted using relativistic BW to produce W transverse mass templates

+backgrounds

binned likelihood fit • data and template normalized to same area for MT<100 GeV• fit to high MT region (MT>100 Gev) gives width

Model tested via a detailed “MC closure” test. As with mass, “blind” analysis.

For more details, see talk by Jan Stark, FNAL Wine & Cheese, Mar. 20, 2009


Sensitivity

4 Sep 2009

1

2

3

3


Electron energy scale

4 Sep 2009

4 X0 of dead material in front of the calorimeter made understanding the scale challenging.

depth in radiation lengths (X0)

EM1

EM2

EM3

EM4

FH1

DEAD

dE/d

X 0 (arb

itrar

y un

its)

EM1

EM2

EM3

EM4

DEAD

eta = 0(normal incidence)

eta = 1

Same as for W mass measurement

For more details, see talk by Jan Stark, FNAL Wine & Cheese, Mar. 20, 2009


Tuning dead material on longitudinal shower shape

4 Sep 2009

EM1 EM2

EM3 EM4

Fractional energydeposits, electronswith |η| < 0.2

Before tuning of material model:

EM1 EM2

EM3 EM4

Fractional energydeposits, electronswith |η| < 0.2

After tuning of material model:


Final Scale with Z’s

4 Sep 2009

Emeasured = α x Etrue + β

Use energy spread of electrons in Z decay to constrain α and β.

α = 1.0111 ± 0.0043β = -0.404 ± 0.209 GeVcorrelation: -0.997

After having corrected for the effects of the uninstrumented material:final energy response calibration, using Z e e, the known Z mass value from LEP:

Result:

Uncertainty dominated by Z statistics

40

Selection EfficiencyNeed to carefully model any dependence of the electron identification efficiency on the PT of the electron that might sculpt the shape of the MT distribution (instead of just changing the normalization). Because of the kinematics, the PT can correlate with other kinematic quantities that affect the identification efficiency.

Electron identification efficiencies affected by:• geometry (z of primary vertex, distance from module boundary in phi and eta)• electron PT

• photon final state radiation• hadronic activity in the event

• correlated with electron PT through W decay kinematics

• scales with component parallel to electron direction (u||)

• also scales with the magnitude of the overall activity in the event (Scalar ET)


Electron ID efficiencies

4 Sep 2009

Track match efficiency versus eta and primary vertex z

Overall identification efficiency versus u|| (data versus MC)


Dependence on electron PT

4 Sep 2009

Check of dependence of shower shape efficiency on electron PT. Black data, red: fast MC. The shape of the dependence is consistent with being the same.

electron PTelectron PT


Uncertainties on efficiency

4 Sep 2009

Most important is dependence on electron PT. Determined by comparing efficiency versus PT from Z->ee events to that from fast simulation.

Compare data and MC for Zee for efficiency versus SET and electron PT and for efficiency versus eta and electron PT and look for evidence of slope. Compatible with no slope.

D0 Preliminary

Events at high transverse mass are from off-shell W’s and from high PT W’s with recoil underestimated.


Modeling the Recoil

Transverse mass spectra from • the generator level (Red histogram), • electron energy response/resolution Included (Blue histogram),• recoil response (scale of 0.6) and resolution included (Green histogram), • MET resolution due to zerobias events included (Light blue histogram), • hadronic scale is set to 1 and also met resolution due to zerobias events included (Black histogram).

No uT cut

Modeling the recoil


Theorist view

Cartoon version

As seen in the detector

Measured “recoil” includes ISR, Underlying event, pileup, detector noise.


Two Recoil Methods

Overlay recoils taken from the Z data on MC W’sarXiv: 0907.3713

The recoil library method


Recoil Library Method

PTtrue =0.4 ΓeV

PTtrue =10 ΓeV

PTtrue =29 ΓeV

D0 MC

D0 MC D0 MC



Map also includes total hadronic activity to use with electron ID dicing



PTmeasured =

7 ΓeV

Note this method has no tunable parameters

Use Bayesian method of unfolding* to produce weights that can be used when assigning a measured recoil to a bin in true boson PT

* G. D’Agostini, NIM A362, 487 (1995)

Recoil Library


Using library without unfolding After unfolding

Recoil library uncertainties


• Same as uncertainty from parameterized method, and much easier to implement• Uncertainty for both methods dominated by Z statistics

MC closure test


PDF Uncertainties

4 Sep 2009

20 pairs of CTEQ6.1M PDF’s“data” generated with central PDF (+pythia)MC templates created reweighting events according the 40 error PDF’s

Use method suggested by CTEQ:

DΓW =1

2x1.6(Γ i

+ − Γ i− )2

i=1

20

∑ where Γ is the

width determined using the ith uncertainty PDF set

Converts from 90% cl to 68% cl


Width uncertainties

4 Sep 2009

Width


ΓW = 2.028 ± 0.038(stat) ± 0.061(sys) GeV (3.5% uncertainty)( ± 0.072 GeV)

Systematics dominated, but systematics are dominated by Z statistics.


Fit Results

4 Sep 2009

Comparison between data and MC at best fit mass for MT, electron PT, and MET.

From electron PT spectrum: 2.012 ±0.046 ΓeVFroμ /ETσpeχtruμ : 2.058 ±0.036 ΓeV

D0 Preliminary D0 Preliminary

D0 Preliminary


Tail events

4 Sep 2009

Comparison between data and MC at best fit width for distributions of various kinematic variables for events with 100<MT<200 GeV


D0 Preliminary


Comparison to previous measurements

4 Sep 2009

D0 Preliminary


Conclusions

4 Sep 2009

• New limits on anomalous triple gauge couplings from WZ/ZZ to lνjj events.• New limits combining channels on anomalous triple gauge couplings (best limits from Tevatron). • When the full 5 fb-1 is analyzed and combined, may improve on combined LEP sensitivity on aTGC.• New muon asymmetry results have errors smaller than spread due to current PDF uncertainties. When incorporated into PDF fits, should decreased error on W mass and width.• new W width measurement (3.5% precision)


Backup Slides

4 Sep 2009

Wine & Cheese, FNAL 604 Sep 2009

Interactionpoint

First active layer ofliquid argon

about3.7 X0 in between !

0.9 X0

0.3 X0 plus 1 X0 of lead

cryo walls: 1.1 X0

inner detector: 0.1 X0

4X0


Other uncertainties

4 Sep 2009

• assume uT cut makes us only sensitive to uncertainties on the non-perturbative part of the prediction, so uncertainties on W PT distribution: vary g2 within uncertainty from global fit (0.68±0.02 GeV2) (F. Landry, R. Brock, P. Nadosky, C.P. Yuan, Phys. Rev. D 67, 073016 (2003)• photon FSR: compare PHOTOS to WGRAD and ZGRAD (which includes interference at 1 photon level). Fit data with an effect to templates without.

Pierre Petroff/ Dzero CERN seminar, May 05, 2009 62

Background to W e ν


Stat Error MET, MT

4 Sep 2009


Modeling the efficiency

4 Sep 2009

• get “intrinsic” dependence on electron PT, dependency on overall event activity, and dependency on FSR from detailed GEANT-based MC, (both single electrons and Z->ee and W->eν events with overlay of real zero-bias-triggered data)• get average efficiency versus eta and vertex z, and u|| dependence from Z data (tag and probe)• correct and test with Z data


SET and electron PT correlations

4 Sep 2009

Ratio of efficiency versus electron PT to average efficiency as function of scalar ET for various bins in electron PT. The complex shape of these functions comes from kinematic correlations between electron PT and scalar ET through the recoil. High PT electrons tend to come from high PT W’s that decay so the electron is along the direction of the boost.. While the SET can be large for these events, it is not near the electron and therefore does not cause inefficiencies.


Width consistency tests

4 Sep 2009

D0 Preliminary D0 Preliminary D0 Preliminary

D0 Preliminary D0 Preliminary D0 Preliminary


Stability of MT fits

4 Sep 2009

D0 PreliminaryD0 Preliminary


Stability of electron PT and MET fits

4 Sep 2009

D0 PreliminaryD0 Preliminary



Width from simple ratio

4 Sep 2009


muon PT requirement and x

4 Sep 2009


Comparison to electron

4 Sep 2009

New Results from D0 on the W width, charge asymmetry and on gauge couplings

Documents

wine cheese

w play

w width4

w decays

w production

fnal sm couplings

years of w physics

fnalequal couplings