Strange Sea Asymmetry: Strange Sea Asymmetry: Analysis MethodsAnalysis Methods

Laura Gilbert and Jeff Tseng, University of Oxford 16/08/07

OUTLINE

1) Background and motivation: quark asymmetries in the proton

2) Detecting a strange sea asymmetry3) Feynman diagrams4) Event generation5) Method 1: W+Jet 6) Method 2: W+D*7) Discussion of backgrounds8) Final thoughts

10-2

10-1

100

0.00

0.25

0.50

0.75

1.00

1.25

1.50

1.75

2.00

MRST2001

Q2 = 10 GeV2

ant

idow

n / a

ntiu

p

x

Motivation: Quark Asymmetries in the

Proton u, d distributions in the proton predicted to be

almost flavour symmetric within pQCD. MNC measured the flavour nonsinglet structure

function [Fp2(x,Q2) − Fn

2(x,Q2)]. → large (~30%) violation of Gottfried sum rule:

0))()((1

0

dxxuxd

d/u Confirmed by the NA51, E866 and HERMES.

Various theoretical models proposed. Meson Cloud model (MCM) seems physically intuitive as a way to explain observations.

Strange Sea Momentum Asymmetry

In the MCM the proton oscillates into virtual mesons/baryons

Sea q/q are in different environments thus carry different momenta.

Symmetric s/s distribution often assumed, but not established theoretically or experimentally.

MCM would imply a strange momentum fraction asymmetry too.

0))()((1

0

dxxsxsx

du

u qq

du

uoscillate

sq

du

u q

x(s(x) - s(x))

Ws at LHC sensitive to small x regime (<0.01). Difficult to

probe.

Phys.Lett. B590 (2004) 216-222: Ding & Ma

Calculations from Meson Cloud Model – 2-body wavefunctions [Gaussian (thick) and power-law (thin)]

Detecting a strange sea asymmetry in the proton

Feynman diagram sensitive to strange quark distribution needed. Use s+g→c+W, ie. NLO W production.

This mechanism is charge symmetric if the strange/anti-strange distributions are the same.

General W production at LHC already shows charge asymmetry in rapidity distributions of W.

Need to remove this bias and then look for limits on null hypothesis of signal channel.

Two suggestions: look for any charmed jet produced with W, or look for D* with W.

Using W→eν as it’s easy to work with but could look for muon too, in theory doubles rate although muon reconstruction efficiency significantly lower than electron.

NLO Feynman Diagrams: W NLO Feynman Diagrams: W productionproduction

LO DiagramLO Diagram No W transverse No W transverse

momentummomentum

NLO DiagramsNLO Diagrams W has transverse W has transverse

momentummomentums

c

W

s

c

W

g

s

g

W

c

cg

Ws

LO: 77%

NLO: 23%

NLO Gluon production:46% of NLO10% of total

Using MC@NLO

Event Generation MC@NLOMC@NLO ~3 million of each W~3 million of each W++→e→e++υυ, W, W--→e→e--υυ events, cross events, cross

sections 2.217nb and 1.640nb respectivelysections 2.217nb and 1.640nb respectively All Plots normalised to 1fbAll Plots normalised to 1fb-1-1.. Known issue: NLO diagrams show forward-Known issue: NLO diagrams show forward-

backward asymmetry in W (and also partner backward asymmetry in W (and also partner jets). Problem currently left with Jon jets). Problem currently left with Jon Butterworth. Butterworth.

W+Jet method : TheoryW+Jet method : Theory

W selection as usualW selection as usual Event has just one reconstructed Event has just one reconstructed

jet, displaced vertexjet, displaced vertex Few other mechanisms should Few other mechanisms should

provide large numbers of provide large numbers of displaced verticesdisplaced vertices

Very inclusive selectionVery inclusive selection

s

g

W

c

W+Jet method : Background W+Jet method : Background rejectionrejection

Background suppression:Background suppression: LO diagrams removed by jet requirementsLO diagrams removed by jet requirements 1) b jets: u suppressed by ~1) b jets: u suppressed by ~λλ33, c by ~, c by ~λλ22. t rare in proton. . t rare in proton. 2) c jets: d suppressed by ~2) c jets: d suppressed by ~λλ, b by ~, b by ~λλ22. . 3) t jets: produced mainly from bs in proton, rare. 3) t jets: produced mainly from bs in proton, rare.

Therefore mainly only charm jets produced Therefore mainly only charm jets produced from strange sea should remain (?)from strange sea should remain (?)

With symmetric input PDF the WWith symmetric input PDF the W++ and W and W-- passing all cuts should then show no charge passing all cuts should then show no charge asymmetry.asymmetry.

t, c, u

g

W

b

b, s, d

g

W

c

b, s, d

g

W

t

1) 2) 3)

W+Jet method : Background W+Jet method : Background rejectionrejection

Suspect this method won’t work due to gluon Suspect this method won’t work due to gluon splitting (10% of MC@NLO sample), not obviously splitting (10% of MC@NLO sample), not obviously removable?removable?

s

c

W

g bb

d

u

W-

g bb

u

d

W+

g bb

Signal: symmetric if s=s, c=c

Background: not symmetric

May also be very large uncertainties in strange sea May also be very large uncertainties in strange sea asymmetry measurements due to MI, pile-up, large x-asymmetry measurements due to MI, pile-up, large x-section QCD backgrounds such as cc etc. with this method.section QCD backgrounds such as cc etc. with this method.

W+Jet method : Event SelectionW+Jet method : Event Selection

W selection as usualW selection as usual Electron transverse momentum >25GeVElectron transverse momentum >25GeV Missing transverse energy > 25GeVMissing transverse energy > 25GeV Electron pseudorapidity < 2.4Electron pseudorapidity < 2.4

Event has just one reconstructed Event has just one reconstructed jetjet

Jet has with high impact Jet has with high impact parameter (B-tagging) and parameter (B-tagging) and ET>25GeVET>25GeV

W+Jet method: ATLFASTW+Jet method: ATLFAST Quick and dirty method: use ATLFAST built-Quick and dirty method: use ATLFAST built-

in b-tagging to check basic principlesin b-tagging to check basic principles B-tagging: ATLFASTBB-tagging: ATLFASTB

Provides jet energy and momentum calibrationProvides jet energy and momentum calibration Limited to inner tracker acceptance range of |Limited to inner tracker acceptance range of |ηη||

<2.5 so only jets in this range are accepted in <2.5 so only jets in this range are accepted in selection cuts.selection cuts.

Binary b-tagging efficiency (random) of 50% (60%) Binary b-tagging efficiency (random) of 50% (60%) high (low) luminosity.high (low) luminosity.

Rejection factors of Rc=10 for charm jets, Rj=100 Rejection factors of Rc=10 for charm jets, Rj=100 for light jets. Static (no for light jets. Static (no ηη, pT dependence)., pT dependence).

W+Jet method: ATLFAST plotsW+Jet method: ATLFAST plotsComplete NLO Sample all electrons:

Electrons

Positrons

After cuts:

True “Signal” only (s+g→W+ btagged jet):

It appears that this selection method is stillsubject to a dominating proton valence asymmetry.

Note NLO gen level f/b asymmetry is slightly

visible

W+Jet method: ATLFAST plotsW+Jet method: ATLFAST plotsComplete NLO Sample all electrons:

Electrons

Positrons

After cuts:

True “Signal” only (s+g→W+ btagged jet):

Equivalent asymmetry plots

W+D*W+D* Analysis Select W candidate (isolated

electron, |η|<2.4, pT>25GeV, ETmiss>25GeV)

Reconstruct D0→K-π+ (also D0→K-π +π0, D0→K-π +π-π +π0 etc)

D0 flight length: cτ=123μm so vertex displaced.

Add prompt (soft) pion. Consider 3 sign correlations: Consider 3 sign correlations:

(K(K-- with with ππ++, K, K-- with with ππBB++, , ππBB

+ + with ewith e--)) Plot reconstructed D*-D0 mass Plot reconstructed D*-D0 mass

difference = 145.4MeVdifference = 145.4MeV (small intrinsic (small intrinsic resolutions: D* width 96keV, D0 width resolutions: D* width 96keV, D0 width 1.6meV , small background)1.6meV , small background)

Consider backgrounds inc. cabibbo supressed wrong sign combinations, QCD, QED, MI, pile up etc.

Should find zero asymmetry in Monte-Carlo from accepted PDFs. Work out CL on limits of null hypothesis.

s

g

W

c

cg

Ws

Branching ratios: D*+→D0π+ 67.7%

D0 → K- π+ 3.8%c→D* 25.5%c→e 9.6%

W+D* AnalysisW+D* Analysis Preliminary Cuts:Preliminary Cuts:

1 electron with pT>25GeV, |1 electron with pT>25GeV, |ηη|<2.4|<2.4 MET>25GeVMET>25GeV Two oppositely signed tracks: assign one K, one Two oppositely signed tracks: assign one K, one ππ. . pT(K)>1.5GeV, pT(pT(K)>1.5GeV, pT(ππ)>1GeV)>1GeV Third track: assign bachelor Third track: assign bachelor ππBB, pT(, pT(ππBB)>0.5GeV)>0.5GeV ππB B charge opposite to e, opposite to Kcharge opposite to e, opposite to K Reconstructed D0 mass within 200MeV of true.Reconstructed D0 mass within 200MeV of true.

Further cuts indicated by sFurther cuts indicated by s22/(s+b) optimisation /(s+b) optimisation – compare efficiency of selecting “true” signal – compare efficiency of selecting “true” signal D*s with backgrounds of the same sign D*s with backgrounds of the same sign correlations.correlations.

W selection

W+D* AnalysisW+D* Analysis

- pT(e)>25GeV, |η(e)|<2.4- MET>25GeV- pT(K)>1.5GeV, - pT(π)>1GeV, - charge(K)*charge(π)<1- pT(πB)>0.5GeV

- charge(K)*charge(πB)<1,

- charge(e)*charge(πB)<1

- m(D0reco)- m(D0true)< 200MeV (loose)

Reconstructed D*-D0 mass difference: peaks at 145.4MeV.Reconstructed D*-D0 mass difference: peaks at 145.4MeV.

Reconstructed Unsmeared Real D*s

- m(D0reco)- m(D0true)< 40MeVW+D* SelectionW+D* Selection

D0 massD0 mass

Real D*s Full sample

Real D*s Full sample

- m(D0reco)- m(D0true)< 40MeV

- signed Lxy>0.35mm

W+D* Selection W+D* Selection LxyLxy

D0

D0 cτ=123μm K

πLxy

(Lxy –ve is tracks point towards vertex)

Reconstruct vertex: straight line approx

- m(D0reco)- m(D0true)< 40MeV

- signed Lxy>0.35mm

- d0/σ(d0)<3D* lifetime < 10-20s

Therefore batchelor π should be prompt: sanity cut at 3 σ

W+D* Selection W+D* Selection ππBB

d0/sigma(d0)d0/sigma(d0)

Real D*s Full sample

- m(D0reco)- m(D0true)< 40MeV

- signed Lxy>0.35mm

- d0/σ(d0)<3-

d0(K)*d0(π)<0mm2

Impact parameter is signed according to which side of the vertex it passes.

Therefore K, π have oppositely signed impact parameters.

W+D* Selection W+D* Selection ππBB

d0/sigma(d0)d0/sigma(d0)

Real D*s Full sample

- m(D0reco)- m(D0true)< 40MeV

- signed Lxy>0.35mm

- d0/σ(d0)<3-

d0(K)*d0(π)<0mm2

- d0(D0)<0.2mm

W+D* Selection W+D* Selection D0 impact parameterD0 impact parameter

D* lifetime < 10-20s

Therefore D0 should be prompt

Real D*s Full sample

This cut is not very effective – probably is reduntant due to d0(K)*d0(π) cut

Missing pTMissing pT At LO the W is produced with momentum along the direction

of the beampipe Electron and neutrino from W decay produced back-to-back in

transverse plane Resolve MpT along the direction of travel of the electron:

perpendicular to line of flight of electron we expect MpT perp = 0 at generator level.

Including detector smearing this results in a sharp Gaussian. At NLO W is produced at any angle so electron and neutrino

tend to be approximately back to back, but angle is no longer 180 degrees at gen level

the Gaussian will be much wider so this could be useful to select NLO diagrams.

Probable LO contribution

Probable NLO contribution

This cut is not useful for signal amplification

No improvement if calculated as the first cut, or if the MET >25GeV cut is entirely removed

Cut OptimisationCut OptimisationMissing pt perpendicular to Missing pt perpendicular to

electron ptelectron pt

Real D*s Full sample

Signal: ResultsSignal: Results

- pT(e)>25GeV, |η(e)|<2.4- MET>25GeV- pT(K)>1.5GeV, - pT(π)>1GeV, - charge(K)*charge(π)<1- pT(πB)>0.5GeV- charge(K)*charge(πB)<1, - charge(e)*charge(πB)<1- m(D0reco)- m(D0true)<

40MeV - signed Lxy>0.35mm- d0/σ(d0)<3- d0(K)*d0(π)<0mm2

- d0(D0)<0.2mm

Reconstructed Unsmeared Real D*s

NB different sample!

No. signal events = 119±27No “real” D*s in window = 102No. W- events = 56 ±18No “real” D*s = 49

No. W+ events = 62 ±19No “real” D*s = 53

More thoughts on cutsMore thoughts on cuts Sanity cut on pT, η of D* candidates due to track addition, consider η

of other backgrounds.

Will revisit missing ET considering MET parallel as well as perpendicular to lepton line of flight. In signal we expect W with relatively low pT (e, missing energy ~back to back) which may not be true in QCD backgrounds.

Parallel case is less well resolved in full simulation than perpendicular, also mean displaced from 0 since the electron calorimeter corrections are not perfectly tuned

Probable LO contribution

Probable NLO contribution

Plots from DC3 sample 005250 (MC@NLO), v 11.0.42

Reconstructed GEANT truth

Real D*s Full sample

Signal: Results and futher workSignal: Results and futher work Strange sea asymmetry: expect –ve Strange sea asymmetry: expect –ve

((s(x)s(x)>>s(x)s(x) at low x) at low x)

How many do we need in order to see How many do we need in order to see difference?difference?

Say 100 events at 1fbSay 100 events at 1fb-1-1. To exclude null . To exclude null hypothesis to 95% CL we need around 60% hypothesis to 95% CL we need around 60% asymmetry (80:20). Need a lot more data! asymmetry (80:20). Need a lot more data! 100 fb100 fb-1-1??

In this case we would plot D* asymmetry In this case we would plot D* asymmetry as a function of rapidity.as a function of rapidity.

cWgscWsg

cWgscWsg

NN

NNA

BackgroundsBackgrounds QCD heavy quark production (eg. cc, bb, tt) QCD heavy quark production (eg. cc, bb, tt)

cc (Pythia MSEL=4):cc (Pythia MSEL=4): x-sect 1.450x-sect 1.450μμb, cf. ~1nb for Ws.b, cf. ~1nb for Ws. ~8x10~8x107 7 events so farevents so far 13 events pass all cuts → ~250 events at 1fb13 events pass all cuts → ~250 events at 1fb-1 -1 lumi.lumi. More work needed on cuts to reduceMore work needed on cuts to reduce

D* backgrounds (wrong sign combinations, other D* backgrounds (wrong sign combinations, other kaon decay modes, D* correlated with fake Ws – kaon decay modes, D* correlated with fake Ws – eg. as seen in cc etc.)eg. as seen in cc etc.)

W backgrounds: ZW backgrounds: Z→ee; →ee; ZZ→→ττττ→l→lννννX;W→X;W→τντν;W→l;W→lνννν; ; WW; WZ; electrons from heavy quark decays, WW; WZ; electrons from heavy quark decays, dalitz decays or photon conversion; MI; pileup; dalitz decays or photon conversion; MI; pileup; missing jets. missing jets. W+extra jets: incl. W + cc (bb), one heavy quark lost: qq→Wg*→WQQ

Final ThoughtsFinal Thoughts Simple W+jet selection probably not effective on it’s Simple W+jet selection probably not effective on it’s

own – not clear how to remove gluon background.own – not clear how to remove gluon background. Could refine b-taggingCould refine b-tagging What happens with full sim (inc. MI etc)?What happens with full sim (inc. MI etc)?

Stick with D* analysis?Stick with D* analysis? Low stats but reasonably clear signalLow stats but reasonably clear signal Pleasing number of cross-checks available (eg. sign Pleasing number of cross-checks available (eg. sign

correlations)correlations) Need more data for convincing asymmetry measurementsNeed more data for convincing asymmetry measurements

Background statistics will be calculated in ATLFAST: Background statistics will be calculated in ATLFAST: much more work needed to reduce QCD backgroundsmuch more work needed to reduce QCD backgrounds

Need to consider how to do produce signal in full sim.Need to consider how to do produce signal in full sim.