B-jet production cross section at CDF Monica D’Onofrio University of Geneva Wine&Cheese Seminar, September 9 th 2005

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th In the last 15 years…  Study of events with b-quarks has led to important Tevatron results Discovery and study of the top quark B physics in general (spectroscopy, lifetimes measurement, sin 2etc..) Measurement of quarkonium states and appreciation of color-octet-mediated production mechanisms  These results mostly obtained when a factor 3 discrepancy was reported between theory predictions and experimental data by both CDF and DØ in b-hadron cross sections

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th  « To claim that we need to understand b production in order to make new discoveries is therefore a bit exaggerated ….. Nevertheless, lack of confidence in the ability to describe properties of events containing b quarks, in addition to raising doubts over the general applicability of pQCD in hadronic collisions, does limit our potential for the discovery of possible subtle and unexpected new phenomena » (M.Mangano, HCP2004) Therefore, the study of b production properties should be one of the main priorities for RunII … …also considering high statistics.. E cm =1.96 TeV (bb) ~ 50 b  few kHz event rate!!

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Outline  Recent advances in the theory of b production cross sections in hadronic collisions and application to experimental results related to B-hadrons.  Exploring b-jets at CDF: Inclusive b-jet cross section  Event selection, experimental tools  Results and preliminary comparison with theoretical calculations at Next-to-Leading order (NLO) More exclusive b-jet cross sections  bb-jets correlations: to disentangle b production processes  Z+b-jet cross section: to probe b content of the proton  Conclusions

The theory and recent developments

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th B-quark production in hadron collisions Leading OrderNext to Leading Order Gluon splitting Flavor excitation Flavor creation g g g g Q Q other radiative corrections.. Experimental inputs are B-Hadrons or b-jets rather than b-quark Factorization theorem: factorize physical observable into a calculable part and a non-calculable but universal piece NLO QCD Observed Proton structure Fragmentation

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Proton structure: PDF Parton Distribution Functions (PDFs) are universal global fits to data on proton structure independent of the process  Momentum distributions of the partons inside proton Generally PDF uncertainties are estimated at ~ 15% Dominant contribution due to high x gluon distribution gluon u d x Uncertainty on gluon PDF (from CTEQ6) x

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Fragmentation functions D b  B Perturbative part: probability to find a hadron with fraction x’ of original parton momentum Hadronization: non perturbative QCD, need models

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Recent theory advances pQCD calculations: resummation of  s log(p T /m) terms  Fixed Order Next Leading Log ( FONLL) p T >> m  need large corrections Moment analysis to treat D meas for fragmentation (new approach: Cacciari et. Al. 2002) Release date of PDF  b NLO (|y|<1) (  b) < 1994 now Cross section very dependent on PDF evolution

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Testing FONLL: B hadron production J/  K+K+ B+B+   J/ decays of B-Hadrons used to measure the b production cross section  Find J/ inclusive cross section  Extract fraction of J/ from decay of long-lived b-hadrons  Find b-hadrons cross section for p T (B) down to 0 considering |y(J/)|<0.6 B  J/ X shape from MC templates  J/  from B  J/ X will be displaced Maximum likelihood fit on flight path to extract b fraction as function of p T (J/)

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th B hadron production cross section Total inclusive single b-hadron (H b ) cross section considering Br(H b  J/X) = 1.160.10% and Br(J/  ) = 5.880.10% Good agreement with theory prediction

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Summary on theory advances comparison with RunI data |y(H b )| < 1, RunII multiplied by B + fragmentation=0.4 (E cm rescaled) RunI RunII Reduction of discrepancy is due to four basic points: FONLL calculation brings 20% increase in intermediate p T region fragmentation step from perturbative b quark to B hadron at small p T was too strong: 20% increase in this p T region Peterson fragmentation function was too soft: use new LEP data =0.002 additional 20% increase PDF evolution (> 20% increase) Data moved ~ 20% down (still within errors) Many little changes combined together  big effect in Data/Theory comparison

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Why b-jets? b-jets include most of quark fragmentation remnants  small dependence on fragmentation wide P T spectrum In RunI studies performed to measure bottom and charm fraction in inclusive jet samples  b = 19 ± 2(stat) (syst) [nb] at P T > 35 GeV/c In RunI differential b-jet cross section using semi-leptonic decays of the b (muon tagger) P T (B) GeV/c (p T >p Tmin, |y|<1)(nb)

Inclusive b-jet cross section at CDF

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th The Tevatron in RunII Peak luminosity in 2005 above cm -2 s -1 CDF collected ~ 1 fb -1 on tape!! (but 1.2 fb -1 already delivered) Analysis shown here use ~ 300 pb -1

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th The CDF experiment Collider Detector Fermilab CENTRAL and PLUG Calorimeter  energy and direction  2 systems of passive layers-scintillators Silicon Detector With 750,000 channels, the largest Silicon detector in the world!  position  3 systems of single or double sided detector  down to 10  m spatial resolution (3D) COT  position  drift chamber  spatial resolution 100  m Muon Chamber (collision hall)  position and p T  4 systems of scintillators and proportional chambers  min scattering resolution [12/p;25/p] cm/p TOF  time  Scintillators  100 ns resolution Solenoid (1.4 T)

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th B-jet cross section  N i tagged is the number of tagged jets   i b-tag is the b-tagging efficiency  f i b is the fraction of b-jets among tagged jets  C i unfold are correction factors from Monte Carlo for acceptance and smearing effects  Y is the rapidity range  p i T is the size of bin in transverse momentum  ∫ L is the integrated luminosity

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Jet reconstruction Final state partons are revealed through collimated flows of hadrons called jets Beam remnants Hard scattering Multiple partons interaction Two main type of jet algorithms (in CDF): - Cone Algorithm  JETCLU and MIDPOINT - K T algorithm Jet  Seed towers Only iterate over towers above certain threshold (3 GeV at CDF) MidPoint adds extra seed in centre of each pair of seeds  Infrared safe  Ratcheting (JetClu only) All towers initially inside a cone must stay in a cone  Merging/Splitting f merge =0.75

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Event selection  MIDPOINT jets, R cone = 0.7, |Y jet |<0.7  P T range GeV/c  GeV/c for corrected P T jets use 5 samples with different E T jet threshold  Total luminosity used ~ 300 pb -1 Inclusive calorimetric triggers:  Level 1: selection based on E T of cal towers (EM+HAD)  Level 2: accept tower clusters with E T above a fixed threshold  Level 3: jets reconstructed (JETCLU, R cone =0.7, Z v =0) L1 L2 L3 Dataset Path > |Z primary vertex|<50 cm to assure good energy measurement, vertexing capability > Cut on missing E T Significance ( = E T /√∑E T ) implemented to reject to cosmic rays Event Selection Trigger efficiency

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Jet corrections: detector effects For each calorimeter jet in |Y|<0.7 look for the corresponding hadronic (particle) jet to remove dependence from detector effects Time 20 ÷ 10% b-quark-originated jets different from ordinary jets account for smearing effects for detector resolution  apply “unfolding” correction bin by bin for b-jet from Monte Carlo hadronic b-jet need inclusive correction that takes into account the bias due to the tagger  correction on tagged jets

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Jet Corrections: Pile up Effects in jet P T : about -1 GeV/c per each additional primary vertex Jet samples Average number of primary vertices as a function of instantaneous luminosity Small dependence on Lum. 6 different slices of Instantaneous luminosity UEM (P T ) = 0.932±0.002 GeV Main idea: measure P T in a random cone in Minimum Bias sample (central region) as a function of # primary vertices to define effect due to multiple interactions;

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th B tagging algorithm Looks for tracks associated with a jet the track selection is based on on measurement of impact parameter (d 0 ) with respect to primary vertex In general b-tagging procedures take advantage of the long life-time of B hadrons  c ~ 450 m Need ≥ two displaced tracks to reconstruct a secondary vertex (made in 2 steps) After secondary vertex reconstruction  require to be well separated from primary vertex in r- space by looking at L xy and its error: Jets passing those selections: tagged

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th B tagging efficiency (1)  Use Monte Carlo simulation to cover the wide P T spectrum [38-400] GeV/c  Measure efficiency scale factor to take into account simulation imperfections (tracking efficiency&resolution, B hadron decay models…)  For this purpose, use independent dataset Sample enhanced in b-jet content: dijet events  one e/jet + “away” jet tagged look e.g. at  b-jet for the muon jet  b jet = F tag bjet N tag+ jet F bjet N  jet b-jet content after tagging b-jet content before tagging  Measure b-tagging efficiency in Data and MC b-jet content extract using P T rel muon-jet

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th B tagging efficiency (2)  Data b-jet / MC b-jet  Scale Factor (SF) SF = 0.9090.06(stat+syst)  Systematic error due to hadronic VS semileptonic b-decay below 3%  Geometrical acceptance and energy dependence of the tagger  from simulation  Tagging rates parametrized as function of relevant variables to define systematic error on P T dependence  5%

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Extract fraction of b-tagged jets from data using shape of mass of secondary vertex as discriminating quantity  bin-by-bin as a function of jet p T  2 component fit:b and non-b templates (Monte Carlo PYTHIA) b-fraction tagged jets 82 < p T jet < 90 GeV/c

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Systematics errors Main sources: o jet reconstruction o Secondary vertex mass templates Uncertainty:  MC generator, fitting procedure  Heavy quark multiplicity in jets  Fragmentation 3% ES  Resolution  Smearing  effect on cross section ~ 20-40% Total Jet Reconstr. (%) JES only (%)

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Systematics on Secondary vertex Mass To estimate systematics:  Test fit stability depending on templates shape and statistic - also PYTHIA/HERWIG comparison  effects of fluctuation in relative composition of 2b/1b and 2c/1c estimate from NLO calculation  Check on templates variation due to fragmentation scheme: PYTHIA Lund model, = (default) VS Peterson model, = < p T < 202 GeV/c 1c/2c 1b/2b Total systematic from fraction: from 10% to 30% (but for last bin ~ 50%)

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th b-jet cross section: results Total systematic error ~ 25%  70% in the last bin Differential b-jet cross section at particle level (range p T GeV/c) Ratio Data/Pythia MC(CTEQ5L)

Preliminary comparison with NLO for inclusive b-jet cross section

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th NLO - NLO calculation for b-jets Mangano&Frixione - 2  3 process, so jets are very simple: 1 or 2 partons inside Shape very sensitive to bb content from gluon splitting (more likely to have 2 b inside same jet) gg qq qg total

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Scale dependence of NLO Fraction of double b-quark ending up in the same jet depend on gluon splitting, only appearing at LO  Strong scale dependence ~ 45% ~ 30% Rate bb-jets / All jets

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Hadronization and Underlying events  HadronizationUnderlying events ~ 20% correction for lowest bin None for b-jets above 130 GeV/c Corrections that need to be added to theory from PYTHIA Monte Carlo Before comparison with theoretical expectations  correct NLO b-jets for hadronization and underlying events

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Preliminary Data VS NLO b-jets NLO theoretical expectation for b-jet corrected at particle level in analogy to inclusive jet cross section measurements  m b =4.75 GeV/c 2  PDF Uncertainty: 7%  20%  Merging/splitting issue: R theory =R data *R sep, R sep =1.3  10% uncertainty  Include scale uncertainty ~ from 40%  20% (P T bjet >250 GeV/c)

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Data/NLO ratio Ratio up to 1.5 above 100 GeV/c jets  Poor agreement but still within systematics without considering scale uncertainty if considering scale uncertainty overlap region increase

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Comparison with Run I D  results  Use   for central theory value  data close to upper band of systematic (= 0 /2  PDF uncertainty)  direct comparison of data not possible (different center of mass energy s, different jet algorithm, different rapidity range …) For jets below 100 GeV/c, D0 data in RunI showed a similar pattern

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Issues on high P T b-jets Unknown impact of higher order contributions: reduced scale dependence event has more partons in the final state, thus closer to the real world better description of the transverse momentum of final state due to double radiation of initial states Logarithmic log(p T /m) enhancement of higher order contribution due to gluon splitting is not included in NLO calculations (neither in  at low P T effects are small (range of B-hadron cross section)  at high P T are very important and need to be considered Experimentally: study of bb-jets correlation  to disentangle different production mechanisms Z+b jets and +b jets  could help to constrain the b density in the proton

More exclusive bjet cross sections

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th bb jets cross section  dijet events JETCLU, R cone =0.7 jets E T 1 >30GeV, E T 2 >20GeV |jets| < 1.2  central jets: more likely to be sensitive to flavor creation  Jets corrected at particle level for b flavor jets  Tag 2 jets with b-tagging algorithm earlier described Flavor excitation g g g g Q Q Gluon splitting Flavor creation ( + radiative corrections) predominantly back-to-back Small data sample used  still preliminary  Analysis with larger sample in progress

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th bb jets cross section results b fraction from Mass secvtx fit (global fit all E T range) F bb =0.830.04 = 34.5  1.8 nb (stat.) (syst) Integrated cross section: A = trigger acceptance ~ 1 Pythia (CTEQ 5l) 38.71  0.62nb 28.49  0.58nb Monte Carlo prediction: = 35.7 ± 2.0 nb with preliminary UE tuning Need to add more statistics on with final Underlying Event tuning Data results

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th bb jets correlations  Predominantly back to back  Explains agreement in cross section with Pythia  LO MC deviates away at low (where statistics are still low) Differential cross section as function of  jets Compared to Linear scale Log scale

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Z+b jet production In QCD, Z+b can help constrain b density in the proton Probe the heavy flavor content of proton + With HERA F bb 2 data: CTEQ below MRST by down to 1/2 and below data  Z+b jets can help understand this picture Important background for new physics such as higgs search

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Analysis strategy  Leptonic decays for Z reconstruction: Z  e + e -,  +  -  Signal defined as Z 0,  l + l - events with 66<M ll <116 GeV/c 2   Z associated with jets - R cone jet = 0.7, || 20 GeV  Backgrounds: - fake electrons/same-sign muons (Data) - Real signatures e + e - / +  - + bjet (MC) Muon channel equivalent

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Mass of secondary vertex Fit  Look for tagged jets in Z events  same b-tagging algorithm as in previous analyses  extract fraction of b-tagged jets from secondary vertex Mass  Use negative tagged jets to better constrain light and c quarks  Make no assumption on the charm content  extract  b : N b Data /N b MC Main source of systematic uncertainty from dependence of mass templates on b/bb content in the jet

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Z+bjets results Z+b jet cross section corrected at particle (hadron) level:  b : N b Data /N b MC in Z events SF : scale factor for b-tag N Z+b MCHad : MC particle b-jets in Z evts N Z MCHad : total MC Z evts N Z MC : MC evts passing Z cuts N Z Data : Data evts passing Z cuts  meas (Z) : CDF Z cross section =M Z Uncertainty ~10% changing scale Central value for Z+b cross section also above theoretical expectations

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Summary and Conclusion  In the last year many advances in theory calculation for b production at intermediate and low P T  Good agreement with B-Hadron cross section  Big effort in study of b-jets production in CDF RunII  Inclusive b-jet cross section measurement done within a wide range in transverse momentum using ~300 pb -1 data  Similar behaviour to D RunI cross section w.r.t. theoretical expectations for jet P T below 100 GeV/c  Agreement with theory within uncertainties: NLO b-jet cross section calculation still shows big scale dependences  Study of more exclusive b-jet cross section could help:  bb correlations to disentangle production processes  Z+b to understand b content in the initial radiation

Back up

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Impact parameter resolution Typical d 0 resolution

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th B-jet cross section: Data/Herwig MC

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th bb jets differential cross section Differential cross section as function of dijet mass Differential cross section as function of leading jet E T Comparison Data VS + UE tuning Comparison Data VS + UE tuning

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Z      + b Z   +  - channel: same-sign muons events with a reconstructed jet Specific background in muon channel

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th NLO uncertainty for Z+b Scale dependence is small ( 10%) Big uncertainty on b density in the proton

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Sec. Vertex mass (GeV)  Et > 25 GeV (||<1.0) + jet with secondary vertex Determine b, c, uds contributions (fit secondary vertex mass) Subtract bkg, find cross- section as fn.  Et 25–29 GeV 29–34 GeV 34–42 GeV 42–60 GeV  +b/  +c production

Monica D'OnofrioWine&Cheese Seminar, Fermilab September 9th Results consistent with LO  (  +b)  (  +c)  +b/  +c production