1 / 68

Precision Measurement of the Top Quark Mass at CDF

Precision Measurement of the Top Quark Mass at CDF. Università di Siena Dottorato in Fisica Siena, 25 Giugno 2007. Candidato: Michele Giunta Relatori: Prof. Giorgio Bellettini Dr. Georgui Velev Relatore interno: Prof. Angelo Scribano. Outline.

mauli
Download Presentation

Precision Measurement of the Top Quark Mass at CDF

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Precision Measurement of the Top Quark Mass at CDF Università di SienaDottorato in FisicaSiena, 25 Giugno 2007 Candidato: Michele Giunta Relatori: Prof. Giorgio Bellettini Dr. Georgui VelevRelatore interno: Prof. Angelo Scribano

  2. Outline Top Quark: introduction Decay Channels Combinatorics Template Method Improvement to TM: 3Best BLUE Sanity Checks Mass Measurement Results Michele Giunta, 25/06/07

  3. The Top Quark In the Standard Model, there are 3 generations of elementary matter fermions. Four bosons (not including gravity) carry interactions among them. Fermions and bosons get masses through their interaction with the yet-to-be discovered Higgs boson. Each generation comprises a lepton and a quark doublet. Members differ in one unit of electric charge. The third generation of quarks was discovered in 1977 when the b-quark was observed in a bound b-antib state and soon understood to be a down quark type. After that discovery it was expected its doublet to be completed by a “top” quark. The enormous mass of the top quark delayed discovery by about 20 years. Michele Giunta, 25/06/07

  4. Why Top Mass It is important to measure the top quark properties, primarily its mass, to proof the validity of the SM. • In the SM the top and the W masses include a contribution by Higgs virtual loops. These contributions can be computed as a function of the MH and by correlating the Mtop and MW masses one gets indirect but important information on MH. • The top lifetime is so short the it decays before hadronizing. This unique feature makes it possible to measure its mass as the invariant mass of its decay products, as currently done for unstable non-elementary particles. Michele Giunta, 25/06/07

  5. Top Discovery Evidence for the top quark was announced in 1994 by the CDF experiment at the Fermilab Tevatron by observing an excess of top-like events in two decay channels and reconstructing 6 events (19.3 pb-1 data), showing a mass of about 175 GeV. Full discovery was reached with more statistics and announced in 1995 by the CDF and the other Tevatron experiment, DØ. Tevatron Michele Giunta, 25/06/07

  6. The Tevatron The last particle accelerator of the Fermilab chain, the Tevatron, is a proton syncrotron accelerating the proton and antiproton beams in opposite direction from 150 to 980 GeV to provide about 2 TeV CM energy. Top luminosity reached: 286·1030cm-2s-1 Integrated luminosity: up to today ~2.5 fb-1 on tape 2 more years of data-taking scheduled (6 fb-1) CDF CDF Michele Giunta, 25/06/07

  7. The CDF Detector • Muon chambers: • CMP, CMU • CMX • Tracking system: • L00, SVX, ISL • COT • Calorimeters: • Electromagnetic • Hadronic Michele Giunta, 25/06/07

  8. Top Production ~85% ~15% The top quark is expected single or in ttbar pairs. The expected production cross section of single top is about 3 times smaller, detection efficiency is lower and separation from BKG is more difficult. Even if some evidence has been shown, single top production has not been demonstrated yet. Our analysis addresses t-tbar production. Strong production Michele Giunta, 25/06/07

  9. Warning! SINGLE TOP TTBAR SYSTEM • From now on we shall consider only: • ttbar pair production Michele Giunta, 25/06/07

  10. Top Decay Channels Weak decay The top decades ~100% in Wb. The W decay is shared among the allowed channels: en, mn, tn, ud(x3), cs(x3). Depending on the decay of the W, there are the following 3 t-tbar decay channels (“lepton” = e or m): DILEPTON : (2/9)·(2/9) = 4.9% SEMILEPTONIC : 2·(2/9)·(6/9) = 29.6% ALL HADRONIC : (6/9)·(6/9) = 44.4% Michele Giunta, 25/06/07

  11. Warning! DILEP, ALLHAD SEMILEPTONIC • From now on we shall consider only: • ttbar pair production • semileptonic decay channel Michele Giunta, 25/06/07

  12. L+J: Combinatorics The detector measures energy and direction of four jets. With no information on jet flavour (b-tagging) we cannot assign uniquely jets to partons. A number of t-tbar reconstructions are possible for each event. Michele Giunta, 25/06/07

  13. Combinations Count • 4 objects: 4! = 24 combinations (jet-to-parton associations) • The two light jets (u,d,s) given to the W can be swapped: • 12 combs. • The 2nd degree equation for the neutrino longitudinal momentum has in general 2 different solutions: 24 combs • 1 b-tag: 3!=6 combinations for jets. • W, pz(n) 6 combs again. Ambiguity on the b-side: hadronic or leptonic? : 12 combs • 2 b-tag: 2 combinations for W jets. W, pz (n): again 2 combs. Ambiguity on the b-side: hadronic or leptonic? : 4 combs Michele Giunta, 25/06/07

  14. Warning! • From now on we shall consider only: • ttbar pair production • semileptonic decay channel • pretag data sample 2TAG, 1TAG, 0TAG PRETAG SAMPLE Michele Giunta, 25/06/07

  15. Analysis Methods Template (TM): the shape of reconstructed masses from the data sample is compared to MC predictions (templates) for discrete input top masses. Parametric are functions used to smooth discrete MC samples. Simple method, low CPU needed. Matrix element (ME): To each observed event and possible reconstruction of it, a probability of having been produced in that specific kinematic configuration is associated. A transfer function describes how a produced signal/BG would be seen by the detector. Difficult estimate of systematics, smaller data samples, heavy CPU usage, but better use of the information. Michele Giunta, 25/06/07

  16. Warning! • From now on we shall consider only: • ttbar pair production • semileptonic decay channel • pretag data sample • Template method MATRIX ELEMENT TEMPLATE Michele Giunta, 25/06/07

  17. TM: Analysis Procedure Selection Establish cuts to select MC signals and BKG events and real data. Compute relative/absolute contaminations of BKG in the data sample. BKG estimation: ratios, #events Kinematic Fit Associate to each reconstructed event a top mass and a reconstruction quality factor (c2). Obtain mass shapes for the BKG and for a number of MC generated top masses: [155;195] GeV stepping by 2.5 GeV. Parametrize them. Mass Templates And BKG Samples Final (LH) Fit Perform a LH fit to data. Selection Michele Giunta, 25/06/07

  18. Event Selection The same selection cuts are applied to MC signal, MC BKG, data. The criteria for an event to be accepted are as standard in CDF top studies. • trigger • electron Et>20 GeV or muon pT>20 GeV/c • dilepton veto • MET > 20 GeV • ≥4 jets with ET>15 GeV • cosmic veto • lepton solation > 0.1 • |z|<60cm • track reconstruction quality • … BKGs Michele Giunta, 25/06/07

  19. Backgrounds We must evaluate the amount of background (BKG). BKG process BKG fractions BKG templates • W+light jets • W+H.F. • non-W (QCD) • diboson • Single Top (<1 evt) We estimate the relative contribution of each process to the total BKG We get the BKG templates by reconstructing events as t-tbar Michele Giunta, 25/06/07

  20. BKG Estimation BKG % W+L.F. 63.3 W+H.F. 13.9 Non-W 14.6 Diboson 8.2 c2<9 1 fb-1 Any c2 Total 700 645 Signal 353 339 W+L.F. 223 194 W+H.F. 49 43 Non-W 46 45 Diboson 27 25 RATIOS: The BKG fractions had been computed for tagged events in the course of the cross-section measurement with 695 pb-1. We correct by MC to pretag events and normalize to our 1030 pb-1 sample. NUMBER OF EVENTS: By applying the selection to data, we find 700 events. A kinematical c2 cut accepts 645 events. We can run a constrained fit Michele Giunta, 25/06/07

  21. Combination Sorting In the classical TM, all the possible jet-to-parton assignments are reconstructed. The reconstructionc2gauges how well a combination can fit the t-tbar assumption. The reconstructions are sorted by increasingc2 and the mass associated to the minimum c2 is entered the mass spectrum. Michele Giunta, 25/06/07

  22. Template Final Fit Mtop, sstat A maximum likelihood technique fits the data mass distribution in terms of a BKG and of a mass-dependent signal template. The Mtop for which the LH is maximum indicates the top mass measure. Michele Giunta, 25/06/07

  23. Our Analysis We developed an improved Template Method and analyzed the 1030 pb-1 data sample accordingly Michele Giunta, 25/06/07

  24. c2 Rank Does the best c2 indicate the correct jets-to-parton assignement? In the subsample of events where the 4 t-tbar decay quarks generate the 4 leading jets, only about 50% of times the best c2 is found to correspond to the correct jet/parton association. The rest of times, it is distributed on the other rank bins as in the figure. To recover part of the neglected information we exploit not only the best, but also the second and the thirdbest reconstructions (3Bestc2 Method) 3Bestc2s Michele Giunta, 25/06/07

  25. 3Best c2s Yes, First No: skip Yes, Second Yes, Third Michele Giunta, 25/06/07

  26. Multiple c2 Analysis Comb 1 Comb 2 Comb 3 TMT TMT TMT M1, s1 M2, s2 M3, s3 Combine results Mcomb, scomb Three Mass template sets and three BKG templates are generated, one for each combination. We perform three independent Top mass measurements. Results are combined after taking correlations into account Michele Giunta, 25/06/07

  27. Template Parameterization I II III Michele Giunta, 25/06/07

  28. BKG Parameterization I II III BLUE Michele Giunta, 25/06/07

  29. BLUE Best Linear Unbiassed Estimate [L. Lyons D. Gibaut, How to combine correlated estimates of a single physical quantity, NIMA A270 (1988), pp 110, 117.] The combined mass is a linear combination of the input masses. The weights are constrained. The a1,2,3 are obtained by imposing the combination to have the smallest s2 The correlation factors are related to the covariances in the error matrix. PE Michele Giunta, 25/06/07

  30. Pseudo-Experiments • Pseudo-Experiments (PE) are data-similar sets of MC events. • Each PE: 645 total events out of which 306 ± s (s=44) BKG. • The MC events are fished out from the MC samples. • Example: • take the Mtop=170 GeV MC • extract 337 events (fluctuation!) from it • join with 308 events piked from the BKG samples (using the correct ratios) • fit as they were the data set • repeat. • Correlations (cov1,2, cov1,3, cov2,3) were computed from 2000 PEs. Michele Giunta, 25/06/07

  31. How to Apply BLUE Mtop TEMPLATES: • A set of 2K PEs is created for each Mtop. The distributions of the 3 best combinations are fitted preserving their correlations and the mass correlation factors r12, r13, r23 are determined. • For the n-th event, a1(n), a2(n), a3(n) are determined and the average values a1,a2,a3 are calculated. • The BLUE mass and error mB,n and sB,n are computed for each PE. • The mB and sB are determined by fitting to Gaussians their distributions.. SYSTEMATIC ERRORS and DATA: The correlation factorsr12(175), r13 (175), r23 (175) from the mass template Mtop=175 GeV are used (see next slide). Michele Giunta, 25/06/07

  32. Correlation Factors To measure the DATA w/ BLUE, we cannot study the correlation factors (data is one single experiment). We must compute the rs from a MC sample. Systematic uncertainties must be measured in the same conditions as data. We must use the same rs as for data. How could we choose the best r triplet? We picked a value in the middle of the mass range and showed that the measure is insensitive to other choices. Michele Giunta, 25/06/07

  33. Alpha Weights The values of a1,2,3 as a function of Mtop are the mean values of their distributions. Michele Giunta, 25/06/07

  34. BLUE mass/error Distributions Michele Giunta, 25/06/07 Michele Giunta 11-30-06 prebless

  35. BLUE Masses Michele Giunta, 25/06/07

  36. 2Best c2s 2Best Template 3Best How much does the third combination contribute? To check this, we turn the third combination off and measure the different precision on MC simulations Michele Giunta, 25/06/07

  37. 2Best c2s The 3rd combination itself improves the overall stat error by about 1 more % (about 13% of the BLUE improvement). It would not be worth studying the fourth combination. Sanity Checks Michele Giunta, 25/06/07

  38. Sanity Checks Before applying the method to data, “sanity checks” are made on simulated events in order to assess its ability to correctly reconstruct a top mass.   Sanity checks include: • Minput vs Moutput plots • Pull = (Mfit-Mtrue)/sfit distributions • Blind masses study Michele Giunta, 25/06/07

  39. Min vs Mout Michele Giunta, 25/06/07

  40. Pull Distribution for Mtop=175 GeV Michele Giunta, 25/06/07

  41. Average Pulls vs Mtop Michele Giunta, 25/06/07

  42. BLUE Pulls Michele Giunta, 25/06/07

  43. Blind Masses This test is preliminary to accessing the data. We do not know the real mass for these 5 samples. The test is passed if the residuals between our measure and the nominal value is consistent with zero. The Pythia samples are shifted up by the generator systematic uncertainty. Random order blind samples. Michele Giunta, 25/06/07

  44. Systematic Uncertainties The correlation factors used are: r12(175) = 0.3728 r13(175) = 0.2979 r23(175) = 0.3314 Data! Michele Giunta, 25/06/07

  45. Data Measurement MBLUE = 168.9 ± 2.2stat ± 4.2syst GeV/c2 This result is obtained with no use of b-tag information. Kolmogorov-Smirnov test on the data fit: KS(first) = 0.904 KS(second) = 0.998 KS(third) = 0.953 Michele Giunta, 25/06/07

  46. BLUE Improvement 23% 77% How likely is our BLUE improvement by 5% on the statistical error? From the PEs for Mtop=175 GeV we found that we could have been luckier but that the result was not unlikely at all. Data Michele Giunta, 25/06/07

  47. Among Other Results Our result fits well into the full picture of CDF top mass measurements. Michele Giunta, 25/06/07

  48. Aknowledgements • Questa tesi è stata possibile grazie all’aiuto ed alla collaborazione di molte persone: • dei miei relatori: G. Bellettini, A. Scribano, G. Velev (FNAL) • del prof. L. Lyons (Oxford) • del mio tutor a Fermilab G. Chlachidze (FNAL) • del mio collega bielorusso F. Prokoshin (Dubna) • delle discussioni con G. Punzi, F. Bedeschi, G. Latino, A. Annovi, G. Compostella. • delle enormi forze dispiegate dai fisici e dagli ingegneri che negli anni hanno immaginato, studiato, creato e mantenuto sistemi complessi come CDF ed il Tevatron • della posta elettronica (CERN) Michele Giunta, 25/06/07

  49. Backup slides

  50. “Quark” In 1963 Gell Mann named “quarks” the three constituents hidden in the nucleon. He took the name from “The Finnegan’s wake” by James Joyce. "Three quarks for Muster Mark! Sure he hasn't got much of a bark And sure any he has it's all beside the mark." The word “quark” is probably an onomatopeia for the albatros cry. Michele Giunta, 25/06/07

More Related