bJets production at Tevatron Sandro De Cecco INFN
b-Jets production at Tevatron Sandro De Cecco INFN Roma European Physical Society meeting Strong Interactions session Manchester, 20 -07 -2007 1 Sandro De Cecco - EPS, Manchester 20 -07 -07
Outline • B-jets production introduction – B-Jets Energy Scale and Z bb at Tevatron – Inclusive B-Bbar Jets production at Tevatron • B-Jets associate production with gauge bosons and its impact on Tevatron and LHC physics – Heavy Flavor Jet + g at Tevatron – W / Z + Heavy Flavour production at Tevatron • Recent CDF developments on Jet Flavor taggers – first application to Z + Heavy Flavors 2 Sandro De Cecco - EPS, Manchester 20 -07 -07
bb pairs production in hadronic collisions Leading Order and Next to Leading Order Q g g g Flavor excitation and … radiative corrections Flavor creation Gluon splitting … but: b/c-jets or B/D hadrons are the observables rather than b/c-quark Proton structure Observed NLO QCD Fragmentation Factorizes into a calculable part and a non-calculable but universal piece 3 Sandro De Cecco - EPS, Manchester 20 -07 -07
Inclusive b cross section - low PT • Run I: b cross-section ~ 3 x old NLO QCD • Theoretical approaches: Next-to-Leadinglog resummations, non perturbative fragmentation function from LEP, new factorization schemes… • Experimentally: unbinned maximum likelihood fit to the J/y decay time in the R- plane to extract the b fraction Run II: Bottom Quark Production cross-section: PRD 71, 032001 (2005) Good agreement 4 Sandro De Cecco - EPS, Manchester 20 -07 -07
High PT Jets and b tagging Secondary Vertex b-tag Algorithm (CDF Sec. Vtx) • Select tracks within the Jet Cone • Reconstruct 2 or 3 tracks vertex • Cut on vertex displacement significance to reduce light Jets rate • Secondary vertex mass has some separating power between light, charm and b Flavors mass 5 Sandro De Cecco - EPS, Manchester 20 -07 -07
Inclusive bb di-jets production • Specific Trigger based on L 2 SVT (Secondary Vertex Trigger) used. • Sensitive to the different production mechanisms – Flavor creation at high – Flavor excitation or gluon splitting at low Purity ~ 85 % : extracted from data using shape of secondary vertex mass 6 Sandro De Cecco - EPS, Manchester 20 -07 -07
b-Jets Energy Scale and Z bb • Generic Jet Energy Calibration of the Detector needs specific correction for b-Jets • Reduction of uncertainty in b-JES important for Top mass measurement • Test algorithms to improve b-jet energy resolution is crucial for low mass Higgs search. • Use of L 2 SVT based b-tag di-jet trigger Extract a signal, measure data/MC b-JES Result on 600 pb-1 b-jet Data/MC Energy Scale Factor = 0. 974 ± 0. 020 Z bb signal templates for 0. 9<SF<1. 09 Background template N(Z bb)~5600 7 Sandro De Cecco - EPS, Manchester 20 -07 -07
W / Z /g bosons + HF Jets production The study of W / Z / + Jets production is very relevant: • to QCD: Ø High q 2 ( ~Mboson) interactions, perturbative theory Ø Less leading diagrams ( wrt pure Jets production) Ø Test Monte Carlo generators (ALPGEN…) Ø Probe for protons PDF (in particular W/Z + H. F. ) • for many high p. T analysis where W/Z+HF is a main background: ü Top quark cross section and mass ü Single Top cross section ü Search for low mass Higgs boson ü Several SUSY searches … and also crucial process at LHC ! Sandro De Cecco - EPS, Manchester 20 -07 -07 8
W+HF backgrounds in WH l bb = 1 jet tagged jet sample signal region Mjj Expectations based on: • MC predictions for EW and top backgrounds • Fake tags (mistag matrix) applied to light jets • Heavy Flavor background: admixture of MC and data driven procedure 9 Sandro De Cecco - EPS, Manchester 20 -07 -07
Z + HF x-section from Tevatron to LHC from hep-ph/0312024 - Tevatron : p. T 15 Ge. V, │h│< 2, DRjj > 0. 7 Tevatron LHC Production ( pb ) 10. 4 0. 169 2. 19 0. 631 - - ~1000 3. 32 1. 92 - 1. 59 - - ~50 16. 5 0. 130 3. 22 0. 49 - - ~1400 5. 66 6. 45 - 1. 70 - - ~90 - - 870 137 ~16000 10 Sandro De Cecco - EPS, Manchester 20 -07 -07
LHC example: b quark PDF from Z+b bb->Z @ LHC is ~5% of entire Z production Spread of existing pdf’s gives a 10% spread in the prediction of the SM Higgs cross section Even bigger impact on MSSM Higgs production (sensitive to bb coupling) 11 Sandro De Cecco - EPS, Manchester 20 -07 -07
W + bb production results In secondary-vertex-tagged sample, fit for light, c, b contributions CDF measure: 0. 90± 0. 32 pb for Wbb (ET(j)>20 Ge. V, | |<2. 0 LO calculation (ALPGEN): 0. 74± 0. 18 pb Sandro De Cecco - EPS, Manchester 20 -07 -07 12
g + b Jet production results • Probes heavy-quark PDFs • Background for SUSY (light stop) • Experimentally difficult because of large background from 0 decays SVT+ dataset Inclusive dataset 13 Sandro De Cecco - EPS, Manchester 20 -07 -07
Z+b production (old) results b-tagged jet Campbell, et al. (hep-ph/0510362) ( <1. 5): (Z+b)/(Z+j) = 0. 0237 ± 0. 0085 ( <2. 5): (Z+b)/(Z+j) = 0. 024 ± 0. 007 ( assuming b/c ratio from MC ) Sandro De Cecco - EPS, Manchester 20 -07 -07 14
Z+b jets new CDF analysis • Luminosity from 330 pb-1 1. 5 fb-1 • Improved selection, higher acceptance from 7. 7% 9. 6% Mℓℓ Z+≥ 1 b tag • Improved systematics ET(jet) Inclusive Jets, Cone R=0. 7 Ask for the Jet to have a Secondary Vertex Tag and …exploit light, charm to b separation from sec. vtx. Mass 15 Sandro De Cecco - EPS, Manchester 20 -07 -07
Z+b jets new CDF result Sec. Vtx mass fit in events with at least 1 tagged Jet: Systematic table MSec. Vtx • Overall precision improved by almost a factor of 2. • s(Z+b) is 2 s away from MCFM NLO preliminary calculation by Campbell Sandro De Cecco - EPS, Manchester 20 -07 -07 16
A new CDF H. F. Neural Network Tagger • The underlying idea is improving the ability to boost performance exploiting as much Roma NN b-tag M. C. information as possible: – number of vertices (not just A SINGLE secondary vertex a’ la Sec. Vtx) and their masses; uds c – displaced but un-vertexed tracks (tracks not belonging neither to secondary vertex nor to the primary); – Soft leptons, Jet. Prob, global jet variables • Best to combine all info together in a single discriminant: use (a series of) Neural Networks (NN) • NN training performed on generic sample to be used in different analysis (top, single top, low mass Higgs, …) b • A Continuous tagging variable allow weighting events to get the most of it in terms of search sensitivity • GOOD b-c-uds statistical separation allows the fraction of c jet to be studied as well, HIGH Jet taggability: ~ 75% • Increased per-jet efficiency at same background rate than Sec. Vtx: 17 Sandro De Cecco - EPS, Manchester 20 -07 -07
Z+HF fractions with Roma NN b-tagging Roma NN b-tag output fit in Z+jets events, L= 1. 6 fb-1 CDF Run II Preliminary D AT -B A lin F d- it MC Toy MC estimate on b, c fractions uncertainty L = 1. 6 fb-1 Roma NN output – Could allow for the first time a simultaneous measurement of Z+c and Z+b cross sections. – Application of the Roma NN tagger to other samples: one could fit top/higgs cross section together with the W/Z+HF background as a function of jet multiplicity. Sandro De Cecco - EPS, Manchester 20 -07 -07 18
Summary: • W/Z + HF processes not yet fully studied, now starting to characterize at the Tevatron. • Many Tevatron results on bb and W/g/Z + b-jet productions • New Z+b-jet measurement from CDF • New generation NN Jet Flavor Tagger being developed at CDF and being tested in Z+b, Z+c measurement • … lot of knowledge progresses on HF production processes will be an important Tevatron legacy to the LHC 19 Sandro De Cecco - EPS, Manchester 20 -07 -07
Backup 20 Sandro De Cecco - EPS, Manchester 20 -07 -07
Monte Carlo issues Naïve: W+n p (ME)+(PS) ~ W + ≥ N jet W+(n+1) p (ME)+(PS) ~ W + ≥ N+1 jet. . . Hard scattering ME Soft radiation PS shower jet parton ~ n ME N-2 N-1 Jets outside acceptance N N+1 N+2 N+3 N+4 Small contribution Higher order in αs 2 topologies ~ O(α) Same jet multiplicity double counting Problems: how to generate the whole n jet spectrum avoiding double counting ? how be sure that the selected jet is coming from the ME and not PS ? Goal: have a prescription to safely merge different MC multiplicity samples Sandro De Cecco - EPS, Manchester 20 -07 -07 21
NN tagger structure Tracks info PV info Jet. Prob info SLTm info Sec. Vtx info New Vertexing Vertices NN multiple verticies reconstruction (based on CTVMFT) Neuro. Bayes. TM TMulti. Layer. Perceptron Tracks NN b-q c-q b-c 3 flavour NN bb-b cc-c 5 flavour NN outputs Sandro De Cecco - EPS, Manchester 20 -07 -07 22
New CDF Jet Flavor Tagger - Outlook There is some separation power between single and double b Jet CDF Monte Carlo Roma NN b-tag • Further possibility: building block are there to try to tackle also the b/bbar (c/ccbar) jet content and more closely investigate the b production mechanisms 23 Sandro De Cecco - EPS, Manchester 20 -07 -07
- Slides: 23