Top quark pair cross section measurements at the Tevatron experiments and ATLAS. Flera Rizatdinova (Oklahoma State University)

Transcription

1 Top quark pair cross section measurements at the Tevatron experiments and ATLAS (Oklahoma State University)

2 Outline Introduction: Large Hadron Collider (LHC) and major experiments on it ATLAS detector description Top quark physics at ATLAS Top quark production cross section at the Tevatron Conclusions

3 The Standard Model success and incompleteness Accurately predicts the electromagnetic, weak, and strong interactions between quarks and leptons. Why do we observe matter and almost no antimatter if we believe there is a symmetry between the two in the universe? What is this "dark matter" that we can't see that has visible gravitational effects in the cosmos? Why can't the Standard Model predict a particle's mass? Are quarks and leptons actually fundamental, or made up of even more fundamental particles? Why are there exactly three generations of quarks and leptons? How does gravity fit into all of this?

4 Possible solutions of the problems in the SM SUSY as an extension of the SM provides a natural solution to the hierarchy problem; a dark matter candidate; and GUT-scale unification. Extra dimensions predict 3 generations of particles Include the gravitational interaction Include a viable dark matter candidate

5 The Large Hadron Collider LHC is located at CERN CERN is located near Geneva Part of CERN is in France

6 Large Hadron Collider: Why Hadron? Particle physics needs parton-parton center of mass energy in the 1 TeV region where the Standard Model without Higgs Sector is expected to fail. Need for an exploratory machine (= high discovery potential) to search for Higgs boson and the spectrum of new particles and phenomena. Hadron colliders are exploratory machines giving access to a diversity of physics processes Parton-parton center of mass energy Not fixed = covers wide energy range ( broad band ) Lower than the pp center of mass energy (~1/6) Need pp center of mass energy of O(10 TeV) to reach 1 TeV parton-parton center of mass energy.

7 LHC: Why Collider? Why Large? The useful energy of a collision that can be converted into mass For two colliding particles with 4-vectors p 1, p 2 : s (p 1 p 2 ) 2 Proton beam on fixed target : s = (2pm p ) ~ p Two protons of equal energy colliding: s = 2p ~ p Colliders use magnetic fields in order to keep particles on a circular orbit with radius R R [m] = 3.33 p [GeV] / B [T] B is limited by technology, cost, power consumption to < 10 T For p = GeV, R > 3 km (circumference C > 20 km)

8 LHC: Experiments

9 ATLAS ATLAS will learn about the basic forces that have shaped our Universe since the beginning of time and that will determine its fate. Among the possible unknowns are the origin of mass, extra dimensions of space, unification of fundamental forces, and evidence for dark matter candidates in the Universe. CMS Twin of the ATLAS experiment in terms of physics goals

10 Heavy Ion Physics The ALICE Experiment is going in search of answers to fundamental questions: What happens to matter when it is heated to 100,000 times the temperature at the center of the Sun? Why do protons and neutrons weigh 100 times more than the quarks they are made of? Can the quarks inside the protons and neutrons be freed? Just one second after the Big Bang, antimatter had all but disappeared, leaving matter to form everything that we see around us from the stars and galaxies, to the Earth and all life that it supports. LHCb is an experiment set up to explore what happened after the Big Bang that allowed matter to survive and build the Universe we inhabit today. LHCb

11 A Toroidal LHC Apparatus (ATLAS) 7000 Tons 15 years to build 500M$ in materials 3300 Scientists 172 universities and laboratories from 37 countries 800 graduate students

12 ATLAS detector observes collisions But this collision is so small you cannot see it so you measure all the particles that come out Measure: Direction, Momentum, Energy of every particle

13 The ATLAS collaboration

14 Early physics at LHC: SM physics Typical Standard Model processes Process σ (nb) (Ldt = 100 pb -1 ) Min bias 10 8 ~10 13 bb ~10 12 Inclusive jets 100 ~ 10 7 p T > 200 GeV W eν, 15 ~ 10 6 Z ee, 1.5 ~ 10 5 tt 0.8 ~ 10 4 W Z LHC is a W, Z, b, top factory Small statistical errors in precision measurements can search for rare processes large samples for studies of systematic effects

15 Motivation for accurate measurements of ttbar production cross section 15 Top quark, the heaviest known quark, was discovered in 1995 at the Tevatron by CDF and D0 experiments Test of perturbative QCD at high Q 2 Sensitive to new physics Measure in different channels Non-SM top decays (th + or stop) modify contributions to different channels di-lepton/l+jets ratio probes non-w top decays Measure with different techniques b-tagging method Kinematic fit methods Gives input for searches for which top quark pair production is a dominant background.

16 Top quark pair production at the LHC Top quark pair production at LHC is significantly different from the one at the Tevatron LHC 10% qq 90% gg 85% qq 15% gg Tevatron LHC s=14 TeV 833 ± 100 pb LHC s=7 TeV 165 ± 14 pb Tevatron: 7.62 ± 0.85 pb x m s / 2 s 1.96 TeV x 0.18 s 14 TeV x t

17 How do we observe top quark pair decays? Based on the expected event signature and kinematics of the decay products Require to have: one and only one high-pt lepton (from W decay) Missing Transverse Energy MET> 20 GeV (neutrino from W decay) At least 3 jets with large transverse momentum and large angle with respect to the beam axis But there are other processes that have the same characteristics this is what we call background

18 Top quark production and decay Top quark pair production via strong interaction. For m t =172.5 GeV, tt tt tt pb pb pb S.Moch and P.Uwer N.Kidonakis and R.Vogt Cacciari et al. Theoretical tt tt ~ 9% Assuming unitarity of 3-generation CKM matrix B(tWb) ~ 100% Distinguish top quark pair events by the consequent decays of the W bosons: Both Wl - dilepton channels One W l and other Wqq lepton+jets channels Both W qq all jets events Seminar at Wichita State Universty

19 Ttbar cross section measurements in ATLAS I will describe four analyses performed without applying b- tagging: Multivariate analysis, two 1- dimensional fits (lepton and (lepton,jet)), and cut-and-count analysis Multivariate analysis is done in 3 exclusive and 4 inclusive jet multiplicity bins in muon and electron channels, then results are combined; 1D analysis also uses 3 jet events. 1D lepton and cut-and-count analyses are using 4 th inclusive jet multiplicity bin for the ttbar x-section measurement 19

20 Selections E+jets 1 tight electron with p T >20 GeV, veto on the second lepton E T 35 GeV, M T (W ) 25 GeV At least 3 jets with p T >25 GeV for Multivariate A and 1D, 4 jets for 1D and cut-andcount 20 Mu+jets 1 muon with p T >20 GeV, veto on the second lepton E T 20 GeV, E T M T (W ) 60 GeV At least 3 jets with p T >25 GeV for Multivariate A and 1D, 4 jets for 1D and cut-andcount Major backgrounds: how they were evaluated by analyses W+jets Multivariate A and1d fit analyses Cut and count Normalized from fit From W+jets/Z+jets ratio QCD All analyses Data driven methods Small BG All analyses Use theoretical cross sections

21 W mass in selected sample Comparison of data and Monte Carlo simulations in the low jet multiplicity bin, where no ttbar is expected. 21

22 Idea of the multivariate analysis A This analysis exploits the difference in kinematic distributions of signal and background events. To separate signal from background, use projective likelihood method from TMVA package For each event i, we calculate the likelihood discriminant D i : D i L i,s, where L L i,s L i,s(b ) p i,s(b ) (x j ) i,b Assume that the variables are uncorrelated Fit likelihood discriminant distribution in data by sum of two templates, signal and background, and get fraction of the tt events. k j 1 22

23 Comparison of data and MC for chosen variables in lepton + 4 jets Observe good agreement between data and MC for all variables. e+4 jets µ+4 jets 23

24 Signal and background templates e+3 jets (left) E+4 jets (right) µ+3 jets (left) µ+4 jets (right) 24

25 Fit to the data e+3 jets (left) and e+4 jets (right) µ+3 jets (left) and µ+4 jets (right) 25

26 Final results Measured stat. error ±10.1% Combined (measured stat+syst.) error is ( )% Cross section: 19 tt (stat.) (syst.) 5(lumi.) pb 18 26

27 1D Kinematic Fit to lepton η in 4j Combined fit in muon and electron channels W/Z+jets and ttbar templates from MC Templates have been cross-checked charge subtraction and double b-tag samples QCD template from W+jets MC, studied data driven templates as systematic uncertainty 40 tt (stat.) (syst.) 7(lumi.) pb 38 Plots to go to the CONF note 27

28 1D Fit to Δη max (lepton,jet) (first 3 jets) Used a genetic algorithm to determine event selection with lowest error including only expected stat + JES uncertainties No significant improvement found Some useful information about possible ranges of cuts for future rounds (i.e. how high we can go without loss of sensitivity) Binned modified chi2 (mclimit) fit Float freely ttbar and W+jets Other sources (QCD, Z+jets, dibosons, single top) constrained as prescribed No QCD fit for muons or combination Fit separately e and µ (in 3 and 4 jets selection) Fit combination in single template ( 4 jets only as best error) 28

29 1 2 1) Count selected events 2) Estimate background QCD: data-driven (from note M3) W+jets: data-driven, W/Z ratio method Single top, dibosons, Z+jets: from MC 3) Compute acceptance for signal 3 W+jets from W/Z ratio method C MC from MC, uncertainty from alpgen samples with different parameters, alpgen vs. sherpa, and PDFs Dominant uncertainty from statistics of Z+4jet sample 29

30 Top Mass plots pre-tag Mjjj is invariant mass of the 3-jet (>20 GeV jets) combination with the highest pt 30

31 Summary on all pretag analyses Black dashed line shows NNLO ttbar cross section; yellow band theoretical uncertainties 31

32 Conclusions on ATLAS measurements Four analyses performed without using b-tagging technique are presented All cross sections measured with different methods are in a good agreement with each other and with the SM prediction Current accuracy of the ATLAS measurements is ~15% 32

33 Kinematic methods for ttbar cross section measurements at the Tevatron Event selection at D0: One high p T (>20 GeV) e or Large missing E T (>20 GeV) At least two high p T jets (>20GeV) Leading jet p T >40 GeV Event selection at CDF: One high p T (>20 GeV) e or Large missing E T (>35 GeV) At least three high p T jets (>20GeV) Leading jet p T >35 GeV Use difference in kinematics of signal events (ttbar) and major background (W+jets) Use variables that provide the maximal separation between signal and background and are well described by Monte Carlo simulation D0 uses Boosted Decision Trees (BDT) technique to distinguish between signal and background CDF uses Artificial Neural Network for the same purpose 33

34 D0 results obtained with kinematic fit Used 7 variables as input for BDT Performed analysis in three jet multiplicity bins, 2 jets, 3 jets and 4 jets Data set 4.3 fb 1 ; used tt = 7.7 pb to draw the expected signal Largest sources of the final uncertainly Source +, pb, pb Statistical Electron ID Signal modeling Luminosity tt stat. syst. lumi pb

35 CDF results with ANN Used NN technique with 7 variables. To reduce the luminosity error, replace it with variation of theor using meas Z relation: tt theor meas Z Z Leading sources of syst. uncertainties Source +, %, % Jet ET scale W+jets Q 2 scale Signal modeling Z theory tt (stat.) 0.35(syst.) 0.15(theor.) pb; 7% 35

36 B-tagging for ttbar cross section measurements Dominant backgrounds have small b-jet fractions Requirement of b-jets presence significantly improves signal-tobackground ratio Assume that Br(tbW) = 1 Both experiments employ NN b-tagging algorithms, incorporating information from track impact parameters, a possible secondary vertex, semileptonic decays and others. Typical efficiency per b-jet is 50-60% at 1% mistag rate 36

37 ttbar cross section measurement in l+jets channel using b-tagging D0: Split sample into 0-tags, 1-tag and 2 or more tags CDF: additional selection cut on H T >230 GeV D0 tt CDF tt stat. syst. lumi pb (stat.) 0.59(syst.) 0.14(Z theor.) pb 37

38 CDF and D0 combined results on ttbar cross section measurements Results of several analyses performed on fb 1 CDF tt (stat.) 0.34(syst.) 0.15(lumi.) pb tt tt 6.5% 38

39 Conclusions At this moment, Tevatron results are twice more accurate than ATLAS (and probably CMS) results Combined result would reduce the total error, but both collaborations did not release their measurements officially yet Nevertheless, ttbar cross section measurements at the LHC still have crucial importance for experiments Verified theoretical calculations at the energy frontier Important for new physics searches Demonstrated a good understanding of the detector and good agreement between simulation and data Will achieve Tevatron accuracy in less than a year 39

40 Background slides ttbar x-section calculations Data/MC agreement for variables used in NN, BDT Systematic errors 40 Universty Seminar at Wichita State

41 Ttbar x-section calculation in BDT analysis 41 D0 kinematic Fit: Perform a maximum likelihood fit to the BDT discriminants distributions to get a number of ttbar pairs in the sample tt W MC QCD o o i (here p is Poisson pdf for n observed events in i-th bin of the discriminant, m is expected n of events in this bin, and N W, N MC (diboson, single top, Z+jets), N QCD are numbers of corresponding types of events. The negative of the log-likelihood function is minimized tt t The extraction of is performed using nuisance parameter method Each individual source of systematic uncertainty is allowed to float All results are provided for m t =172.5 GeV L N, N, N, N pn, pn, N W t t t MC t t QCD t t i i i lt lt o o ni log i i Nlt Nlt Nlt log L( N, N, N, N ) log

42 Data/MC agreement in CDF ANN analysis CDF variables used in ANN analysis 42

43 Data/MC agreement in D0 BDT analysis D0 variables used in BDT analysis 43

44 CDF b-tag analysis: data/mc comparison CDF analysis with secvtx tagger Shown are distributions of some variables in 4-jet multiplicity bin 44

45 Systematic errors on kinematic ttbar cross section measurements ANN (CDF analysis) BDT (D0 analysis) 45

46 Systematics on b-tagging analysis CDF analysis D0 Analysis Seminar at Wichita State Universty

47 Systematics on dilepton analyses CDF Pretag analysis D0 B-tag analysis