Search for Standard Model Higgs Boson Production in Association with a W Boson at CDF
aa r X i v : . [ h e p - e x ] M a r FERMILAB-PUB-08-070-E
Search for Standard Model Higgs Boson Production inAssociation with a W Boson at CDF
T. Aaltonen, J. Adelman, T. Akimoto, M.G. Albrow, B. ´Alvarez Gonz´alez, S. Amerio u , D. Amidei, A. Anastassov, A. Annovi, J. Antos, G. Apollinari, A. Apresyan, T. Arisawa, A. Artikov, W. Ashmanskas, A. Attal, A. Aurisano, F. Azfar, P. Azzurri s , W. Badgett, A. Barbaro-Galtieri, V.E. Barnes, B.A. Barnett, V. Bartsch, G. Bauer, P.-H. Beauchemin, F. Bedeschi, P. Bednar, D. Beecher, S. Behari, G. Bellettini q , J. Bellinger, D. Benjamin, A. Beretvas, J. Beringer, A. Bhatti, M. Binkley, D. Bisello u , I. Bizjak, R.E. Blair, C. Blocker, B. Blumenfeld, A. Bocci, A. Bodek, V. Boisvert, G. Bolla, D. Bortoletto, J. Boudreau, A. Boveia, B. Brau, A. Bridgeman, L. Brigliadori, C. Bromberg, E. Brubaker, J. Budagov, H.S. Budd, S. Budd, K. Burkett, G. Busetto u , P. Bussey, A. Buzatu, K. L. Byrum, S. Cabrera p , C. Calancha, M. Campanelli, M. Campbell, F. Canelli, A. Canepa, D. Carlsmith, R. Carosi, S. Carrillo j , S. Carron, B. Casal, M. Casarsa, A. Castro t , P. Catastini r , D. Cauz w , V. Cavaliere r , M. Cavalli-Sforza, A. Cerri, L. Cerrito n , S.H. Chang, Y.C. Chen, M. Chertok, G. Chiarelli, G. Chlachidze, F. Chlebana, K. Cho, D. Chokheli, J.P. Chou, G. Choudalakis, S.H. Chuang, K. Chung, W.H. Chung, Y.S. Chung, C.I. Ciobanu, M.A. Ciocci r , A. Clark, D. Clark, G. Compostella, M.E. Convery, J. Conway, K. Copic, M. Cordelli, G. Cortiana u , D.J. Cox, F. Crescioli q , C. Cuenca Almenar p , J. Cuevas m , R. Culbertson, J.C. Cully, M. Datta, T. Davies, P. de Barbaro, S. De Cecco, A. Deisher, G. De Lorenzo, M. Dell’Orso q , C. Deluca, L. Demortier, J. Deng, M. Deninno t , P.F. Derwent, G.P. di Giovanni, C. Dionisi v , B. Di Ruzza w , J.R. Dittmann, M. D’Onofrio, S. Donati q , P. Dong, J. Donini, T. Dorigo, S. Dube, J. Efron, A. Elagin, R. Erbacher, D. Errede, S. Errede, R. Eusebi, H.C. Fang, S. Farrington, W.T. Fedorko, R.G. Feild, M. Feindt, J.P. Fernandez, C. Ferrazza s , R. Field, G. Flanagan, R. Forrest, M. Franklin, J.C. Freeman, I. Furic, M. Gallinaro, J. Galyardt, F. Garberson, J.E. Garcia, A.F. Garfinkel, K. Genser, H. Gerberich, D. Gerdes, A. Gessler, S. Giagu v , V. Giakoumopoulou, P. Giannetti, K. Gibson, C.M. Ginsburg, N. Giokaris, M. Giordani w , P. Giromini, M. Giunta q , G. Giurgiu, V. Glagolev, D. Glenzinski, M. Gold, N. Goldschmidt, A. Golossanov, G. Gomez, G. Gomez-Ceballos, M. Goncharov, O. Gonz´alez, I. Gorelov, A.T. Goshaw, K. Goulianos, A. Gresele u , S. Grinstein, C. Grosso-Pilcher, R.C. Group, U. Grundler, J. Guimaraes da Costa, Z. Gunay-Unalan, C. Haber, K. Hahn, S.R. Hahn, E. Halkiadakis, B.-Y. Han, J.Y. Han, R. Handler, F. Happacher, K. Hara, D. Hare, M. Hare, S. Harper, R.F. Harr, R.M. Harris, M. Hartz, K. Hatakeyama, J. Hauser, C. Hays, M. Heck, A. Heijboer, B. Heinemann, J. Heinrich, C. Henderson, M. Herndon, J. Heuser, S. Hewamanage, D. Hidas, C.S. Hill c , D. Hirschbuehl, A. Hocker, S. Hou, M. Houlden, S.-C. Hsu, B.T. Huffman, R.E. Hughes, U. Husemann, J. Huston, J. Incandela, G. Introzzi, M. Iori v , A. Ivanov, E. James, B. Jayatilaka, E.J. Jeon, S. Jindariani, W. Johnson, M. Jones, K.K. Joo, S.Y. Jun, J.E. Jung, T.R. Junk, T. Kamon, D. Kar, P.E. Karchin, Y. Kato, R. Kephart, J. Keung, V. Khotilovich, B. Kilminster, D.H. Kim, H.S. Kim, J.E. Kim, M.J. Kim, S.B. Kim, S.H. Kim, Y.K. Kim, N. Kimura, L. Kirsch, S. Klimenko, B. Knuteson, B.R. Ko, S.A. Koay, K. Kondo, D.J. Kong, J. Konigsberg, A. Korytov, A.V. Kotwal, M. Kreps, J. Kroll, N. Krumnack, M. Kruse, V. Krutelyov, T. Kubo, T. Kuhr, N.P. Kulkarni, M. Kurata, Y. Kusakabe, S. Kwang, A.T. Laasanen, S. Lami, S. Lammel, M. Lancaster, R.L. Lander, K. Lannon, A. Lath, G. Latino r , I. Lazzizzera u , T. LeCompte, E. Lee, J. Lee, Y.J. Lee, S.W. Lee o , S. Leone, S. Levy, J.D. Lewis, C.S. Lin, J. Linacre, M. Lindgren, E. Lipeles, A. Lister, D.O. Litvintsev, C. Liu, T. Liu, N.S. Lockyer, A. Loginov, M. Loreti u , L. Lovas, R.-S. Lu, D. Lucchesi u , J. Lueck, C. Luci v , P. Lujan, P. Lukens, G. Lungu, L. Lyons, J. Lys, R. Lysak, E. Lytken, P. Mack, D. MacQueen, R. Madrak, K. Maeshima, K. Makhoul, T. Maki, P. Maksimovic, S. Malde, S. Malik, G. Manca, A. Manousakis-Katsikakis, F. Margaroli, C. Marino, C.P. Marino, A. Martin, V. Martin i , M. Mart´ınez, R. Mart´ınez-Ballar´ın, T. Maruyama, P. Mastrandrea v , T. Masubuchi, M.E. Mattson, P. Mazzanti, K.S. McFarland, P. McIntyre,
2. McNulty h , A. Mehta, P. Mehtala, A. Menzione, P. Merkel, C. Mesropian, T. Miao, N. Miladinovic, R. Miller, C. Mills, M. Milnik, A. Mitra, G. Mitselmakher, H. Miyake, N. Moggi, C.S. Moon, R. Moore, M.J. Morello q , J. Morlok, P. Movilla Fernandez, J. M¨ulmenst¨adt, A. Mukherjee, Th. Muller, R. Mumford, P. Murat, M. Mussini t , J. Nachtman, Y. Nagai, A. Nagano, J. Naganoma, K. Nakamura, I. Nakano, A. Napier, V. Necula, C. Neu, M.S. Neubauer, J. Nielsen e , L. Nodulman, M. Norman, O. Norniella, E. Nurse, L. Oakes, S.H. Oh, Y.D. Oh, I. Oksuzian, T. Okusawa, R. Orava, K. Osterberg, S. Pagan Griso u , C. Pagliarone, E. Palencia, V. Papadimitriou, A. Papaikonomou, A.A. Paramonov, B. Parks, S. Pashapour, J. Patrick, G. Pauletta w , M. Paulini, C. Paus, D.E. Pellett, A. Penzo, T.J. Phillips, G. Piacentino, E. Pianori, L. Pinera, K. Pitts, C. Plager, L. Pondrom, O. Poukhov, N. Pounder, F. Prakoshyn, A. Pronko, J. Proudfoot, F. Ptohos g , E. Pueschel, G. Punzi q , J. Pursley, J. Rademacker c , A. Rahaman, V. Ramakrishnan, N. Ranjan, I. Redondo, B. Reisert, V. Rekovic, P. Renton, M. Rescigno, S. Richter, F. Rimondi t , L. Ristori, A. Robson, T. Rodrigo, T. Rodriguez, E. Rogers, S. Rolli, R. Roser, M. Rossi, R. Rossin, P. Roy, A. Ruiz, J. Russ, V. Rusu, H. Saarikko, A. Safonov, W.K. Sakumoto, O. Salt´o, L. Santi w , S. Sarkar v , L. Sartori, K. Sato, A. Savoy-Navarro, T. Scheidle, P. Schlabach, A. Schmidt, E.E. Schmidt, M.A. Schmidt, M.P. Schmidt, M. Schmitt, T. Schwarz, L. Scodellaro, A.L. Scott, A. Scribano r , F. Scuri, A. Sedov, S. Seidel, Y. Seiya, A. Semenov, L. Sexton-Kennedy, A. Sfyrla, S.Z. Shalhout, T. Shears, P.F. Shepard, D. Sherman, M. Shimojima l , M. Shochet, Y. Shon, I. Shreyber, A. Sidoti, P. Sinervo, A. Sisakyan, A.J. Slaughter, J. Slaunwhite, K. Sliwa, J.R. Smith, F.D. Snider, R. Snihur, A. Soha, S. Somalwar, V. Sorin, J. Spalding, T. Spreitzer, P. Squillacioti r , M. Stanitzki, R. St. Denis, B. Stelzer, O. Stelzer-Chilton, D. Stentz, J. Strologas, D. Stuart, J.S. Suh, A. Sukhanov, I. Suslov, T. Suzuki, A. Taffard d , R. Takashima, Y. Takeuchi, R. Tanaka, M. Tecchio, P.K. Teng, K. Terashi, J. Thom f , A.S. Thompson, G.A. Thompson, E. Thomson, P. Tipton, V. Tiwari, S. Tkaczyk, D. Toback,
3. Tokar, K. Tollefson, T. Tomura, D. Tonelli, S. Torre, D. Torretta, P. Totaro w , S. Tourneur, Y. Tu, N. Turini r , F. Ukegawa, S. Vallecorsa, N. van Remortel a , A. Varganov, E. Vataga s , F. V´azquez j , G. Velev, C. Vellidis, V. Veszpremi, M. Vidal, R. Vidal, I. Vila, R. Vilar, T. Vine, M. Vogel, I. Volobouev o , G. Volpi q , F. W¨urthwein, P. Wagner, R.G. Wagner, R.L. Wagner, J. Wagner-Kuhr, W. Wagner, T. Wakisaka, R. Wallny, S.M. Wang, A. Warburton, D. Waters, M. Weinberger, W.C. Wester III, B. Whitehouse, D. Whiteson d , A.B. Wicklund, E. Wicklund, G. Williams, H.H. Williams, P. Wilson, B.L. Winer, P. Wittich f , S. Wolbers, C. Wolfe, T. Wright, X. Wu, S.M. Wynne, A. Yagil, K. Yamamoto, J. Yamaoka, T. Yamashita, U.K. Yang k , Y.C. Yang, W.M. Yao, G.P. Yeh, J. Yoh, K. Yorita, T. Yoshida, G.B. Yu, I. Yu, S.S. Yu, J.C. Yun, L. Zanello v , A. Zanetti, I. Zaw, X. Zhang, Y. Zheng b , and S. Zucchelli t (CDF Collaboration ∗ ) Institute of Physics, Academia Sinica,Taipei, Taiwan 11529, Republic of China Argonne National Laboratory, Argonne, Illinois 60439 University of Athens, 157 71 Athens, Greece Institut de Fisica d’Altes Energies,Universitat Autonoma de Barcelona,E-08193, Bellaterra (Barcelona), Spain Baylor University, Waco, Texas 76798 Istituto Nazionale di Fisica Nucleare Bologna, t University of Bologna, I-40127 Bologna, Italy ∗ With visitors from a Universiteit Antwerpen, B-2610 Antwerp, Belgium, b Chinese Academy of Sciences,Beijing 100864, China, c University of Bristol, Bristol BS8 1TL, United Kingdom, d University of CaliforniaIrvine, Irvine, CA 92697, e University of California Santa Cruz, Santa Cruz, CA 95064, f Cornell University,Ithaca, NY 14853, g University of Cyprus, Nicosia CY-1678, Cyprus, h University College Dublin, Dublin4, Ireland, i University of Edinburgh, Edinburgh EH9 3JZ, United Kingdom, j Universidad Iberoameri-cana, Mexico D.F., Mexico, k University of Manchester, Manchester M13 9PL, England, l Nagasaki In-stitute of Applied Science, Nagasaki, Japan, m University de Oviedo, E-33007 Oviedo, Spain, n QueenMary, University of London, London, E1 4NS, England, o Texas Tech University, Lubbock, TX 79409, p IFIC(CSIC-Universitat de Valencia), 46071 Valencia, Spain, Brandeis University, Waltham, Massachusetts 02254 University of California, Davis, Davis, California 95616 University of California, Los Angeles, Los Angeles, California 90024 University of California, San Diego, La Jolla, California 92093 University of California, Santa Barbara, Santa Barbara, California 93106 Instituto de Fisica de Cantabria, CSIC-University of Cantabria, 39005 Santander, Spain Carnegie Mellon University, Pittsburgh, PA 15213 Enrico Fermi Institute, University of Chicago, Chicago, Illinois 60637 Comenius University, 842 48 Bratislava,Slovakia; Institute of Experimental Physics, 040 01 Kosice, Slovakia Joint Institute for Nuclear Research, RU-141980 Dubna, Russia Duke University, Durham, North Carolina 27708 Fermi National Accelerator Laboratory, Batavia, Illinois 60510 University of Florida, Gainesville, Florida 32611 Laboratori Nazionali di Frascati, Istituto Nazionaledi Fisica Nucleare, I-00044 Frascati, Italy University of Geneva, CH-1211 Geneva 4, Switzerland Glasgow University, Glasgow G12 8QQ, United Kingdom Harvard University, Cambridge, Massachusetts 02138 Division of High Energy Physics, Department of Physics,University of Helsinki and Helsinki Institute of Physics, FIN-00014, Helsinki, Finland University of Illinois, Urbana, Illinois 61801 The Johns Hopkins University, Baltimore, Maryland 21218 Institut f¨ur Experimentelle Kernphysik,Universit¨at Karlsruhe, 76128 Karlsruhe, Germany Center for High Energy Physics: Kyungpook National University,Daegu 702-701, Korea; Seoul National University, Seoul 151-742,Korea; Sungkyunkwan University, Suwon 440-746,Korea; Korea Institute of Science and Technology Information, Daejeon,305-806, Korea; Chonnam National University, Gwangju, 500-757, Korea Ernest Orlando Lawrence Berkeley National Laboratory, Berkeley, California 94720 University of Liverpool, Liverpool L69 7ZE, United Kingdom University College London, London WC1E 6BT, United Kingdom Centro de Investigaciones EnergeticasMedioambientales y Tecnologicas, E-28040 Madrid, Spain Massachusetts Institute of Technology, Cambridge, Massachusetts 02139 Institute of Particle Physics: McGill University, Montr´eal,Canada H3A 2T8; and University of Toronto, Toronto, Canada M5S 1A7 University of Michigan, Ann Arbor, Michigan 48109 Michigan State University, East Lansing, Michigan 48824 University of New Mexico, Albuquerque, New Mexico 87131 Northwestern University, Evanston, Illinois 60208 The Ohio State University, Columbus, Ohio 43210 Okayama University, Okayama 700-8530, Japan Osaka City University, Osaka 588, Japan University of Oxford, Oxford OX1 3RH, United Kingdom Istituto Nazionale di Fisica Nucleare, Sezione di Padova-Trento, u University of Padova, I-35131 Padova, Italy LPNHE, Universite Pierre et MarieCurie/IN2P3-CNRS, UMR7585, Paris, F-75252 France University of Pennsylvania, Philadelphia, Pennsylvania 19104 Istituto Nazionale di Fisica Nucleare Pisa, q University of Pisa, r University of Siena and s Scuola Normale Superiore, I-56127 Pisa, Italy University of Pittsburgh, Pittsburgh, Pennsylvania 15260 Purdue University, West Lafayette, Indiana 47907 University of Rochester, Rochester, New York 14627 The Rockefeller University, New York, New York 10021 Istituto Nazionale di Fisica Nucleare, Sezione di Roma 1, v Sapienza Universit`a di Roma, I-00185 Roma, Italy Rutgers University, Piscataway, New Jersey 08855 Texas A&M University, College Station, Texas 77843 Istituto Nazionale di Fisica Nucleare Trieste/ Udine, w University of Trieste/ Udine, Italy University of Tsukuba, Tsukuba, Ibaraki 305, Japan Tufts University, Medford, Massachusetts 02155 Waseda University, Tokyo 169, Japan Wayne State University, Detroit, Michigan 48201 University of Wisconsin, Madison, Wisconsin 53706 Yale University, New Haven, Connecticut 06520 (Dated: October 27, 2018)
Abstract
We present a search for standard model Higgs boson production in association with a W bosonin proton-antiproton collisions ( p ¯ p → W ± H → ℓνb ¯ b ) at a center of mass energy of 1.96 TeV.The search employs data collected with the CDF II detector which correspond to an integratedluminosity of approximately 1 fb − . We select events consistent with a signature of a singlelepton ( e ± /µ ± ), missing transverse energy, and two jets. Jets corresponding to bottom quarksare identified with a secondary vertex tagging method and a neural network filter technique. Theobserved number of events and the dijet mass distributions are consistent with the standard modelbackground expectations, and we set 95% confidence level upper limits on the production crosssection times branching ratio ranging from 3.9 to 1.3 pb for Higgs boson masses from 110 to150 GeV /c , respectively. PACS numbers: 13.85.Rm, 14.80.Bn . INTRODUCTION Standard electroweak theory predicts a single fundamental scalar particle, the Higgsboson, which arises as a result of spontaneous electroweak symmetry breaking [1]; however,the Higgs boson has not been direct observed experimentally. The current constraint on theHiggs boson mass, m H > . /c at 95% confidence level (C.L.), comes from directsearches at LEP2 experiments [2]. Global fits to electroweak measurements exclude massesabove 144 GeV /c at 95% CL [3].At the Tevatron p ¯ p collider at Fermilab, the next-to-leading-order (NLO) Higgs bosonproduction cross section by gluon fusion is about ten times larger than for W H associatedproduction, and the cross section for
W H is about twice that of ZH [4]. The Higgs bosondecay branching ratio is dominated by H → b ¯ b for m H <
135 GeV /c and by H → W + W − for m H >
135 GeV /c [5]. Background QCD b ¯ b production processes in the same invariantmass range have cross sections at least four orders of magnitude greater than that of Higgsboson production [6], and this renders searches in the gg → H → b ¯ b channel extremelydifficult. However, requiring the leptonic decay of the associated weak boson reduces thehuge QCD background rate. As a result, W H → ℓνb ¯ b is considered to be one of the mostsensitive processes for low mass Higgs boson searches .Searches for W H → ℓνb ¯ b at √ s = 1 .
96 TeV have been most recently reported by CDF(using data corresponding to an integrated luminosity of 319 pb − )[7] and D0 (440 pb − )[8].The CDF analysis used a secondary vertex b -tagging algorithm ( secvtx ) to distinguish b -quark jets from light flavor or gluon jets [9]. Upper limits on the Higgs boson production rate,defined as the cross section times branching ratio ( σ · B ), were derived for mass hypothesesranging from 110 to 150 GeV /c . The rate was constrained to be less than 10 pb at 95%C.L. for m H = 110 and less than 2.8 pb for 150 GeV /c . In that analysis, about 50% of thejets tagged by the secvtx tagging algorithm were actually falsely b-tagged jets originatingfrom light flavor, gluon, or charm quarks. This effect is due to the finite resolution of trackmeasurements and the long lifetime of D mesons. Even the small fraction of mistaggedevents in the dominant W q ¯ q process is significant compared to true W b ¯ b production. Toreduce this contamination and enhance the b -jet purity of our sample, we introduce a b - In this paper, lepton ( ℓ ) denotes electron ( e ± ) or muon ( µ ± ), and neutrino ( ν ) denotes electron neutrino( e ν ) or muon neutrino ( µ ν ). W H → ℓνb ¯ b production at CDF using about 1 fb − of data. Section II describes the CDF II detector. The event selection criteria are explainedin Sec. III. In Sec. IV, the b -tagging algorithm with secvtx and neural network (NN) arediscussed in detail. Contributions from the standard model (SM) background are calculatedin Sec. V for various sources. In Sec. VI, signal acceptance and systematic uncertainties areestimated. The search optimization and statistical interpretation of the results are presentedin Secs. VII and VIII, respectively. Finally, our conclusions are presented in Sec. IX. II. CDF II DETECTOR
The CDF II detector geometry is described using a cylindrical coordinate system [10].The z -axis follows the proton direction, and the polar angle θ is usually expressed throughthe pseudorapidity η = − ln(tan( θ/ η andin the azimuthal angle φ .Charged particles are tracked by a system of silicon microstrip detectors and a large opencell drift chamber in the region | η | ≤ . | η | ≤ .
0, respectively. The tracking detectorsare immersed in a 1 . p T = p sin θ is measured to be δp T /p T ≈ . · p T (GeV) for the combined tracking system. The resolution on the track impactparameter ( d ), or distance from the beamline axis to the track at the track’s closest approachin the transverse plane, is σ ( d ) ≈ µ m, about 30 µ m of which is due to the transverse sizeof the Tevatron interaction region.Outside of the tracking systems and the solenoid, segmented calorimeters with projectivetower geometry are used to reconstruct electromagnetic showers and hadronic jets [11, 12, 13]over the pseudo-rapidity range | η | < .
6. A transverse energy E T = E sin θ is measured ineach calorimeter tower where the polar angle ( θ ) is calculated using the measured z positionof the event vertex and the tower location.Small contiguous groups of calorimeter towers with signals are identified and summedtogether into an energy cluster. Electron candidates are identified in the central electromag-9etic calorimeter (CEM) as isolated, mostly electromagnetic clusters which match a track inthe pseudorapidity range | η | < .
1. The electron transverse energy is reconstructed from theelectromagnetic cluster with a resolution σ ( E T ) /E T = 13 . / q E T / (GeV) ⊕
2% [11]. Jetsare identified as a group of electromagnetic (EM) and hadronic (HAD) calorimeter clusterswhich fall within a cone of radius ∆ R = √ ∆ φ + ∆ η ≤ . E T seedcluster [14]. Jet energies are corrected for calorimeter non-linearity, losses in the gaps be-tween towers, multiple primary interactions, out-of-cone losses, and inflow from underlyingevent [15].For this analysis, muons are detected in three separate subdetectors. After at least fiveinteraction lengths in the calorimeter, the muons first encounter four layers of planar driftchambers (CMU), capable of detecting muons with p T > . / c [16]. Four additionallayers of planar drift chambers (CMP) behind another 60 cm of steel detect muons with p T > . | η | ≤ .
6. Muons which exit the calorimeters at 0 . ≤ | η | ≤ . p T > E T > E HAD /E EM < . p T > c (CMUP) or 8 GeV/ c (CMX)pointing to a muon stub. A complete lepton reconstruction is performed online in the finaltrigger stage, where we require E T >
18 GeV /c for electrons and p T >
18 GeV /c for muons. III. EVENT SELECTION
The observable final state from the
W H → ℓνb ¯ b signal consists of two jets plus a leptonand missing transverse energy. The leptonic W decay requirement in W H events yields thehigh- p T lepton and large missing transverse energy due to the neutrino.The results presented here use data collected between February 2002 and February 2006.The data collected using the CEM and CMUP triggers correspond to 955 ±
57 pb − , while10he data from the CMX trigger corresponds to 941 ±
56 pb − .The missing transverse energy ( E T ) is a reconstructed quantity that is defined as theopposite of the vector sum of all calorimeter tower energy depositions projected on thetransverse plane. It is often used as a measure of the sum of the transverse momenta of theparticles that escape detection, most notably neutrinos. To be more readily interpretable assuch, the raw E T vector is adjusted for corrected jet energies, for the transverse momentumof the muons, and for the energy deposition of any minimum ionizing high- p T muons.Events are considered as W H candidates only if they have exactly one high- p T isolatedlepton [19], with E T >
20 GeV for electrons or p T >
20 GeV /c for muons. The isolationcone of ∆ R = 0 . z must be less than 5 cm to ensure the lepton and the jets comefrom the same hard interaction. Some leptonic Z decays would mimic the single-leptonsignature if a lepton is unidentified. Events are therefore rejected if a second track with p T >
10 GeV /c forms an invariant mass with the lepton which falls in the Z -boson masswindow (76 < m ℓX <
106 GeV /c ). The selected events are required to have E T greaterthan 20 GeV.The W H signal includes two jets originating from H → b ¯ b decays; these jets are expectedto have large transverse energy. The jets are required to be in the pseudorapidity rangecovered by the silicon detector so that secondary vertices from b decays can be reconstructed.Specifically, we require the jets satisfy E T >
15 GeV and | η | < .
0. The search for
W H → ℓνb ¯ b is performed in the sample of events with W + exactly 2 jets; however, samples of eventswith W +1,3, ≥ W +2-jet events, at least one jet must be b -taggedby the secvtx algorithm. If only one of the jets is b -tagged, the jet must also pass theNN b -tagging filter. If there are two or more secvtx b -tagged jets, the NN is not applied.With a secvtx mistag rate of 1%, it is rare that two or more jets in the same events aremistagged by secvtx . 11 V. SECONDARY VERTEX b -TAGGING Multijet final states have dominant contributions from QCD light flavor jet production,but the standard model Higgs boson decays predominantly to bottom quark pairs. Correctlyidentifying the b quark jets helps to remove most of the QCD background. An algorithm hasbeen developed and used to tag displaced secondary vertices from b quark decays; however,the sample tagged by the secvtx algorithm still has significant contamination from falsely-tagged light-flavor or gluon jets and the misidentification of c quarks as b -jets [20]. Thissearch introduces a multivariate NN technique intended to improve the secvtx taggingpurity.The b -quark has a relatively long lifetime, and B hadrons formed during the hadroniza-tion of the initial b quark can travel a significant distance on the order of millimeters beforedecaying into a collection of lighter hadrons. The decay vertex can be reconstructed by iden-tifying tracks which form a secondary vertex significantly displaced from the p ¯ p interactionpoint (primary vertex).The secvtx b -tagging algorithm is applied to each jet in the event, using only the trackswhich are within η - φ distance of ∆ R = 0 . secvtx reconstruction and are distinguished by a large impact parametersignificance ( | d /σ d | ) where d and σ d are the impact parameter and the total uncertaintyfrom tracking and beam position measurements. Secondary vertices are reconstructed witha two-pass approach which tests for high-quality vertices in the first pass and allows lower-quality vertices in the second pass. In pass 1, at least three tracks are required to passloose selection criteria ( p T > . /c , | d /σ d | > . p T > . /c . If pass 1 fails, then a vertex is sought in pass 2 from at least two trackssatisfying tight selection criteria ( p T > . /c , | d /σ d | > . p T > . /c ). If either pass is successful, the transverse distance ( L xy ) fromthe primary vertex of the event is calculated along with the associated uncertainty. Thisuncertainty σ L xy includes the uncertainty on the primary vertex position. Finally jets aretagged positively or negatively depending on the L xy significance ( L xy /σ L xy ): L xy /σ L xy ≥ . L xy /σ L xy ≤ − . b -jet samplesfrom decays of top quarks. The energy spectrum for those jets is similar to the spectrumfor b jets from decays of Higgs bosons.The sign of L xy indicates the position of the secondary vertex with respect to the primaryvertex along the direction of the jet. If the angle between the jet axis and the vector pointingfrom the primary vertex to the secondary vertex is less than π/ L xy is positively defined;otherwise, it is negative. If L xy is positive, the secondary vertex points towards the directionof the jet, as in true B hadron decays. For negative L xy the secondary vertex points awayfrom the jet; this may happen as a result of mismeasured tracks, so jets tagged with anegative L xy are labeled mistagged jets. In order to reject secondary vertices due to materialinteraction, the algorithm vetoes two-track vertices found between 1.2 and 1.5 cm from thecenter of the silicon detector (the inner radius of the beampipe and the outer radius of theinnermost silicon layer being within this range). All vertices more than 2.5 cm from thecenter are rejected.The negative tags are useful for evaluating the rate of false positive tags, which are defined“mistags” in the background estimates. Mismeasurements are expected to occur randomly;therefore the L xy distribution of fake tags is expected to be symmetric with respect to zero.Simulated events are used to correct a small asymmetry due to true long-lived particles inlight flavor jets.The efficiency for identifying a secondary vertex is found to be different in the simulatedand observed datasets. We measure an efficiency scale factor, which is defined as the ratioof the observed to the simulated efficiencies, to be 0 . ± .
06 in a sample of high- E T jetsenriched in b jets by requiring a soft lepton ( p T > /c ) from semileptonic heavy quarkdecays [9].Secondary vertex secvtx b -tagging exploits the long lifetime of B hadrons. D hadronsoriginating from c -quarks also have fairly long lifetime, and secondary vertices in c -jets arefrequently tagged. Therefore jets tagged by secvtx are contaminated not only by falselytagged light flavor ( uds or gluon) jets, but also by long-lived charmed hadrons in c -jets. Aneural network has been developed to filter the b -tagging results in order to improve the b -tagging purity.The neural network used in this article employs the jetnet [21] package. The tagger isdesigned with two networks in series. The b − l network is trained to separate b -jets from13ight-quark jets ( l -jets), and the b − c network is trained to separate b -jets from c -jets. Jetswhich pass a cut on both of the NN outputs are accepted by the tagger. These neuralnetworks are trained and applied only to events which are already tagged by the secvtx algorithm. The current NN b -tagging is tuned to increase the purity of the secvtx b -taggedjets, not to increase the tagging efficiency.The neural networks take as input the 16 variables listed in Table I. These variablesare chosen primarily because the b -quark jets have higher track multiplicity, larger invariantmass, longer lifetime and a harder fragmentation function than c - and l -quarks jets. Thetrack parameters and L xy significance are good discriminators for b -jets. The vertex p V T XT and invariant mass M V T X are useful variables for identifying l -jets; however c -jets have p T spectra similar to b -jets. Pseudo- cτ ( L xy × M V T X /p V T XT ), the vertex fit χ , and the track-based probability of a jet to come from the primary vertex are the best discriminators. Theoutputs of the two neural networks are shown in Fig. 1.The NN b -tagger is validated by comparing the performance on data and Monte Carloevents. The NN output from b − l network on a sample of secvtx tagged heavy-flavorjets from events with an electron candidate with E T > b − l network on tagged light-flavor jets from data and Monte Carlo . Figure 2 shows the goodagreement in NN b -tagger performance between data and Monte Carlo.We tune the cut value for 90% b efficiency (after the secvtx efficiency), correspondingto a value of NN b − l = 0 .
182 and NN b − c = 0 . . ± .
02. Note that this is an additional scale factorwith respect to the secvtx efficiency scale factor because all of the jets under considerationhave already been tagged by secvtx . At these cut values, the NN filter rejects 65% oflight-flavor jets and about 50% of the c jets while keeping 90% of b -jets after being taggedby secvtx . A small but purified b -jet sample is obtained by requiring a soft lepton in the jet. ecvtx variable secvtx -independent variableNumber of tracks in fitted vertex Number of good tracksVertex fit χ Jet Probability [22]Transverse decay length ( L xy ) Reconstructed mass of pass 1 tracks L xy significance ( L xy /σ L xy ) Reconstructed mass of pass 2 tracksVertex Mass ( M vtx = p ( P | p vtx | ) − ( P p vtx ) ) Number of pass 1 tracksPseudo- cτ ( L xy × M vtx /p vtx T ) Number of pass 2 tracks p vtx T / ( P good tracks p T ) P Pass1 track p T /p jetT Vertex pass number (pass 1 or 2) P Pass2 track p T /p jetT TABLE I: Input variables used in the NN b -tagging filter. The variables in the first column areproperties of the identified secondary vertex, while variables in the second column are jet propertiesindependent of any identified vertex. b-l Trained Network Output0 0.2 0.4 0.6 0.8 100.020.040.060.080.10.120.14 bottom jetscharm jetslight jets b-c Trained Network Output0 0.2 0.4 0.6 0.8 100.020.040.060.080.10.120.14 bottom jetscharm jetslight jets FIG. 1: Neural network outputs obtained from trainings of b vs. l jets (left) and b vs. c jets (right).Output distributions for b , c and l jets are shown in solid, dashed, and dotted lines, respectively. V. BACKGROUND
The final state signature from
W H → ℓνb ¯ b production can also be reached by other pro-duction processes. The main background processes are W +jets production, t ¯ t production,and non- W QCD multijet production. Several electroweak production processes also con-tribute with smaller background rates. In the following subsections the contribution from15 -l Trained Network Output0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 j e t f r ac t i on / . un i t s DataSimulation b-l Trained Network Output0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 j e t f r ac t i on / . un i t s DataSimulation
FIG. 2: Comparisons of NN b -tag output in data (solid line), and Monte Carlo (dashed line) for secvtx -tagged heavy-flavor-enriched jets (left) and tagged light-flavor jets (right). each background source is calculated in detail. A. Non- W QCD Multijet
Events from QCD multijet production sometimes mimic the W -boson signature with fakeleptons or fake E T . Non- W leptons are reconstructed when a jet passes the lepton selectioncriteria or a heavy-flavor jet produces leptons via semileptonic decay. Non- W E T can beobserved via mismeasurements of energy or semileptonic decays of heavy-flavor quarks. Itis difficult to model and produce the former class of events in detector simulation sincethe reasons for mismeasurement are not known quantitatively. Instead, we estimate thecontribution of non- W events directly from the data sample before b -tagging is applied.Generally, the bulk of non- W events are characterized by a non-isolated lepton and small E T . Lepton isolation I is defined as the ratio of calorimeter energy inside a cone of ∆ R = 0 . I is small if the lepton is well-isolated from the rest of the event, as typified by a true leptonic W decay. This feature isused to extrapolate the expected non- W contribution into our signal region, namely, small I and large E T . The following 4 sideband sectors are used for the extrapolation: I > . E T <
15 GeV (region A),
I < . E T <
15 GeV (region B),
I > . E T > I < . E T >
20 GeV (region D). Here, region D corresponds tothe signal region. In extracting the non- W background contribution from data, we makethe following two assumptions: lepton isolation and E T are uncorrelated in non- W events,and the b -tagging rate is not dependent on E T in non- W events. The level at which theseassumptions are justified determines the assigned uncertainty.With the first assumption, the number of non- W events ( N non − WD ) and their relativefraction in the signal region before requiring b -tagging ( f non − W ) obey the following relations: N non − WD = N B × N C N A , (3) f non − W = N non − WD N D = N B × N C N A × N D , (4)where N i ( i = A, B, C, D ) are the number of pretag events in each sideband region. Thenumber of pretag events has been corrected for known sources of prompt leptons. By in-voking the second assumption, the secvtx b -tagging efficiency obtained in region B can beapplied to the signal region D . Here we define an event tagging efficiency per taggable jet(one with at least two good secvtx tracks) as follows: r B = N (tagged event) B N (taggable jet) B , (5)where N (tagged event) B and N (taggable jet) B are the number of tagged events and taggable jets inregion B , respectively. Then the number of non- W background in region D after secvtx b -tagging( N +non − WD ) is obtained by using the “Tag Rate” relation: N +non − WD = f non − W × r B × N (taggable jets) D . (6)It is also possible to estimate non- W contribution solely from the secvtx -tagged sampleas: N ′ +non − WD = N + B × N + C N + A , (7)where N + X ( X = A, B, C, D ) in the “Tagged Method” are the number of events with positivetags. These methods are data-based techniques, so the estimates could also contain otherbackground processes. The contributions from t ¯ t and W +jets events to each sideband regionare subtracted according to the calculated cross sections for those processes, including theappropriate tagging efficiencies. 17o validate the four-sector method and estimate their systematic uncertainties, we varythe boundaries of the four regions and divide the I and E T sidebands into two E (0 .
20 GeV) and F (
I < . < E T <
20 GeV) sidebands. The observeddeviations imply a 25% systematic uncertainty in the non- W background yield, assignedconservatively for both the pretag and tagged estimates.The independent estimates from the tag rate method (Eq. 6) and the tagged method(Eq. 7) are combined using a weighted average. The result from the tagged method gives aslightly higher estimate than the tag rate method, but the two results are consistent withinthe 25% uncertainty.A non- W rejection factor associated with the NN b -tagging filter is measured from datain region C . Region C has event kinematics similar to non- W events in the signal region D because lepton isolation is the only difference between the two regions. The non- W estimatecalculated before applying NN b -tagging is scaled by this NN rejection factor; this assumesthe NN filter is uncorrelated with the isolation.The non- W estimate for events with at least two secvtx tags is obtained by measuringthe ratio of the number of events with at least one b -tag to the number with at least two b -tags in sideband regions and applying the ratio to the estimate of tagged non- W events inthe signal region D. B. Mistagged Jets
The rate at which secvtx falsely tags light-flavor jets is derived from generic jet samplesin varying bins of η , φ , jet E T , track multiplicity, and total event E T scalar sum. Tag rateprobabilities are summed for all of the taggable jets in the event, jets with at least two trackswell measured in the silicon detector. Since the double-mistag rate is small, this sum is agood approximation of the single-tag event rate. Negative mistags – tags with unphysicalnegative decay length due to finite tracking resolution – are assumed to be a good estimateof falsely tagged jets, independent to first order of heavy flavor content in the generic jetsample. The systematic uncertainty on the rate is largely due to self-consistency in theparameterization as applied to the generic jet sample. The positive mistag rate is enhancedrelative to the negative tag rate by light-flavor secondary vertices and material interactionsin the silicon detectors. As a result, the positive mistag rate is corrected by multiplying18he negative mistag rate by a factor of 1 . ± .
15. This factor is measured in a controlsample by fitting the asymmetry in the vertex mass distribution of positive tags over negativetags [23]. An additional correction factor of 1 . ± .
03 is applied for data collected afterDecember 2004, when the Tevatron beam position changed slightly. The mistag rate per jetis applied to events in the W +jets sample. The total estimate is corrected for the non- W QCD fraction and also the top quark contributions to the pretag sample. To estimate themistag contribution in NN-tagged events, we apply the light flavor rejection power of theNN filter 0 . ± .
05 as measured using light-flavor jets from various data and simulatedsamples. C. W +Heavy Flavor The
W b ¯ b , W c ¯ c , and W c states are major background sources of secondary vertex tags.Large theoretical uncertainties exist for the overall normalization in part because currentMonte Carlo programs generate W +heavy-flavor events only to leading order. Consequently,rates for these processes are normalized to data. The contribution from true heavy-flavorproduction in W +jet events is determined from measurements of the heavy-flavor eventfraction in W +jet events and the b -tagging efficiency for those events, as explained below.The fraction of W +jets events produced with heavy-flavor jets has been studied exten-sively using an alpgen + herwig combination of Monte Carlo programs [24, 25]. Calcula-tions of the heavy-flavor fraction in alpgen have been calibrated using a jet data sample,and measurements indicate a scaling factor of 1 . ± . W +heavy flavor background estimate, the heavy-flavor fractions andtagging rates given in Tables II and III are multiplied by the number of pretag W +jetscandidate events in data, after correction for the contribution of non- W and t ¯ t events to thepretag sample.The previous CDF analysis using 319 pb − of data provided some evidence that the19 et Multiplicity 1 jet 2 jets 3 jets ≥ W b ¯ b (1B) (%) 1.0 ± ± ± ± W b ¯ b (2B) (%) - 1.4 ± ± ± W c ¯ c (1C) (%) 1.6 ± ± ± ± W c ¯ c (2C) (%) - 1.8 ± ± ± W c (%) 4.3 ± ± ± ± W + jets sample. The results from alpgen Monte Carlo have been scaled by the data-derived calibration factor of 1 . ± .
4. (
W c fractions have not been rescaled.) disagreement between the predicted and observed numbers of W +1 jet and W +2 jet events isdue to the heavy-flavor fraction [7]. In this analysis, an updated correction factor of 1 . ± . W +1 jet events only, is applied to the heavy-flavor fraction. The W + heavy flavor background contribution is obtained by the following relation: N W +HF = f HF · ǫ tag · [ N pretag · (1 − f non − W ) − N TOP − N EWK ] , (8)where f HF is the heavy-flavor fraction, ǫ tag is the tagging efficiency, N TOP is the expectednumber of t ¯ t and single top events, and N EWK is the expected number of
W W , W Z , and Z boson events. D. Top and Electroweak Backgrounds
Production of both single top quark and top-quark pairs contribute to the tagged lep-ton+jets sample. Several electroweak boson production processes also contribute.
W W pairs can decay to a lepton, neutrino as missing energy, and two jets, one of which may becharm.
W Z events can decay to the signal
W b ¯ b or W c ¯ c final state. Finally, Z → τ + τ − events can have one leptonic τ decay and one hadronic decay. The leptonic τ decay givesrise to a lepton + missing transverse energy, while the hadronic decay yields a narrow jet ofhadrons with a non-zero lifetime.The normalization of the diboson and single top backgrounds are based on the theoreticalcross sections listed in Table IV, the luminosity, and the acceptance and b -tagging efficiency20 et Multiplicity 1 jet 2 jets 3 jets ≥ ≥ secvtx b -tag (%) W b ¯ b (1B) 33.2 ± ± ± ± W b ¯ b (2B) - 51.3 ± ± ± W c ¯ c (1C) 6.2 ± ± ± ± W c ¯ c (2C) - 14.4 ± ± ± W c ± ± ± ± ≥ secvtx and NN b -tag (%) W b ¯ b (1B) 29.9 ± ± ± ± W b ¯ b (2B) - 47.2 ± ± ± W c ¯ c (1C) 3.8 ± ± ± ± W c ¯ c (2C) - 9.9 ± ± ± W c ± ± ± ± ≥ secvtx b -tag (%) W b ¯ b (2B) - 9.7 ± ± ± W c ¯ c (2C) - 1.2 ± ± ± b -tagging efficiencies in percent for various b -tagging strategies on individual W +heavy-flavor processes. Categories 1 B , 2 B refer to number of taggable b -jets in the events,with similar categories for charm jets. Those numbers include the effect of the data-to-MonteCarlo scale factors algorithm and the neural network filter. derived from Monte Carlo events [19, 26, 27, 28]. The acceptance is corrected for leptonidentification, trigger efficiencies, and the z vertex cut. The tagging efficiency is alwayscorrected by the b -tagging scale factor. E. Summary of Background Estimate
We have described the contributions of individual background sources to the final back-ground estimate. The background estimates for the condition of exactly one b-tagged jetafter applying the NN filter and at least two secvtx b -tagged jets are summarized in Ta-21 heoretical Cross Sections W W ± W Z ± ZZ ± s -channel 0.88 ± t -channel 1.98 ± Z → τ + τ − ± t ¯ t +0 . − . pbTABLE IV: Theoretical cross sections and uncertainties for the electroweak and single top back-grounds, along with the theoretical cross section for t ¯ t at m t = 175 GeV /c . The cross sectionof Z → τ + τ − is obtained in the dilepton mass range m ττ >
30 GeV /c together with a k -factor(NLO/LO) of 1.4. bles V and VI. The estimates are plotted in Figs. 3 and 4 for the case of exactly one b -tagbefore and after applying the NN b -tag filter. The observed number of events in the dataand the SM background expectations are consistent both before and after NN b -tagging isapplied. The same is true for the number of events with at least two b -tagged jets. (SeeTable VI and Fig. 4.) VI. HIGGS BOSON SIGNAL ACCEPTANCE
The kinematics of the SM
W H → ℓνb ¯ b process are well defined, and events can besimulated accurately by Monte Carlo programs. The pythia program was used to generatethe signal samples [29]. Only Higgs boson masses between 110 and 150 GeV /c are consideredbecause this is the mass region for which the decay H → b ¯ b dominates. The number ofexpected W H → ℓνb ¯ b events N is given by: N = ǫ · Z L dt · σ ( p ¯ p → W H ) · B ( H → b ¯ b ) , (9)where ǫ , R L dt , σ ( p ¯ p → W H ), and B ( H → b ¯ b ) are the event detection efficiency, integratedluminosity, production cross section, and branching ratio, respectively. The production crosssection and branching ratio are calculated to NLO precision [5]. The acceptance ǫ is broken22 et Multiplicity 1 jet 2 jets 3 jets ≥ ± ± ± ± W b ¯ b ± ± ± ± W c ¯ c ± ± ± ± W c ± ± ± ± t ¯ t (6.7pb) 6.9 ± ± ± ± ± ± ± ± Z → τ + τ − ± ± ± ± W QCD 84.2 ± ± ± ± ± ± ± ± secvtx b -tag that passes the NNfilter as a function of jet multiplicity.Jet Multiplicity 2 jets 3 jets ≥ ± ± ± W b ¯ b ± ± ± W c ¯ c ± ± ± W c - - - t ¯ t (6.7pb) 10.4 ± ± ± ± ± ± Z → τ + τ − ± ± ± W QCD 1.4 ± ± ± ± ± ± secvtx b -tagged jets as a functionof jet multiplicity. et MultiplicityW+1 jet W+2 jet W+3 jet 4 jet ‡ W+ N u m be r o f E v en t s ‡ W+ N u m be r o f E v en t s ObservedW+Heavy FlavorMistagNon-W QCD tt fi Diboson/Z (6.7pb)+Single TopttBackground Error
Jet MultiplicityW+1 jet W+2 jet W+3 jet 4 jet ‡ W+ N u m be r o f E v en t s ‡ W+ N u m be r o f E v en t s Observed W+Heavy FlavorMistagNon-W QCD tt fi Diboson/Z (6.7 pb)+Single TopttBackground Error
FIG. 3: Number of events as a function of jet multiplicity for events with exactly one secvtx b -tagbefore(left) and after(right) applying the NN b -tagging requirement. Jet MultiplicityW+1 jet W+2 jet W+3 jet 4 jet ‡ W+ N u m be r o f E v en t s ‡ W+ N u m be r o f E v en t s ObservedW+Heavy FlavorMistagNon-W QCD tt fi Diboson/Z (6.7pb)+Single TopttBackground Error
FIG. 4: Number of events as a function of jet multiplicity for events with at least two secvtx b -tagged jets. down into the following factors: ǫ = X ℓ = e,µ,τ ( ǫ z · ǫ trigger · ǫ lepton ID · ǫ b tag · ǫ kinematics · B ( W → ℓν )) , (10)where ǫ z , ǫ trigger , ǫ lepton ID , ǫ b tag , and ǫ kinematics are efficiencies to meet the requirements ofprimary vertex, trigger, lepton identification, b -tagging, and kinematics. The major sources24 Higgs mass(Gev/c110 115 120 125 130 135 140 145 150 A cc ep t an c e ( % ) ‡ FIG. 5: The summary of acceptance of the process
W H → ℓνb ¯ b in W+2jet bin for various b -taggingstrategies as a function of Higgs boson mass. of inefficiency are the lepton identification, jet kinematics, and b -tagging factors; each isa factor between 0.3 and 0.45. The factor ǫ z is obtained from data, and the others arecalculated using Monte Carlo samples. The total signal acceptances for various b -taggingoptions including all systematic uncertainties as a function of Higgs boson mass are shownin Fig. 5.The expected number of signal events is estimated by Eq. 9 at each Higgs boson masspoint. The expectations for various b -tagging strategies are shown in Table VII. The NN b -tagging filter keeps about 90% of signal while it removes 35% of the total background in W +2 jet events as shown in Fig. 3.The total systematic uncertainty on the acceptance stems from the jet energy scale, ini-tial and final state radiation, lepton identification, trigger efficiencies, and b -tagging. A 2%uncertainty on the lepton identification efficiency is assigned for each lepton type (CEM elec-tron, CMUP and CMX muon), based on studies of Z boson events. For each of the high p T lepton triggers, a 1% uncertainty is measured from backup trigger paths or Z boson events.The initial and final state radiation systematic uncertainties are estimated by changing theparameters related to ISR and FSR from nominal values to half or double the nominal [30].The difference from the nominal acceptance is taken as the systematic uncertainty. Theuncertainty in the incoming parton energies relies on the eigenvalue uncertainties provided25 iggs Mass Expected Signal Events(GeV/ c ) Pretag 1 tag 1 tag with NNtag ≥ ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± W H → ℓνb ¯ b signal events with systematic uncertainties forvarious b -tagging options, where “tag” and “NNtag” stand for secvtx b -tagging and NN b -tagging,respectively. in the PDF fits. An NLO version of the PDFs, CTEQ6M, provides a 90% confidence intervalof each eigenvector [31]. The nominal PDF value is reweighted to the 90% confidence levelvalue, and the corresponding reweighted acceptance is computed. The differences betweennominal and reweighted acceptances are added in quadrature, and the total is assigned asthe systematic uncertainty [9].The uncertainty due to the jet energy scale uncertainty (JES) [15] is calculated by shiftingjet energies in W H
Monte Carlo samples by ± σ . The deviation from the nominal accep-tance is taken as the systematic uncertainty. The systematic uncertainty on the secvtx b -tagging efficiency is based on the scale factor uncertainty discussed in Sec. IV. When NN b -tagging is applied, the scale factor uncertainty is added to that of secvtx in quadra-ture. The total systematic uncertainties for various b -tagging options are summarized inTable VIII. VII. OPTIMIZATION OF SEARCH STRATEGIES
The search strategy is optimized by calculating a signal significance defined as S/ √ B ,where S and B are the number of expected signal and background events. In this analysis, S and B are counted within a window which gives the best significance in dijet mass dis-tribution for the particular Higgs mass hypothesis being considered. The window itself is26 ource uncertainty (%)1 Tag 1 Tag & NNtag ≥ < < < b -tagging 3.8% 5.3% 16%Total 5.8% 7.2% 19%TABLE VIII: Systematic uncertainties for various b -tagging requirements. The labels “Tag” and“NNtag” refer to secvtx and NN b -tagging, respectively. optimized by varying the window peak and width for each b -tagging strategy. A comparisonof the significance for various b -tagging options, shown in Fig. 6, provides an a priori metricthat predicts which selection gives the best result.Requiring the NN filter improves the sensitivity by about 10% in the sample of eventswith exactly one b tag. The significance in double-tagged events is almost the same asthat in events with at least one tag and no NN filter. Combining the two results thereforeyields another sensitivity improvement. This combined use of two separate b -tagged samplesprovides a significant improvement as shown in Fig. 6. The total significance increases by20% moving from “ ≥ ≥ secvtx b -tagged jet passingthe neural network filter or at least two secvtx b -tagged jets. VIII. LIMIT ON HIGGS BOSON PRODUCTION RATE
As shown in section VII, there is no significant excess number of events over the SMbackground expectation. Because the dijet mass resonance is a useful discriminant for theHiggs boson signature, we use a binned likelihood technique to fit the observed dijet massdistributions in Figs. 7 and 8, and set an upper limit on the
W H production cross section27 Higgs Mass (GeV/c110 115 120 125 130 135 140 145 150 BS / ‡ ‡ ‡ FIG. 6: Comparison of significance obtained from various b -tagging strategies. “Tag” and “NNTag” represent secvtx and NN b -tagging, respectively. The filled circles correspond to the com-bined analysis which treats disjoint samples with exactly one NN b -tag and two secvtx tagsseparately. times H → b ¯ b branching ratio. A. Binned Likelihood Technique
The number of events in each bin follows the Poisson distribution: P i ( n i , µ i ) = µ n i i e − µ i n i ! ( i = 1 , , · · · , N bin ) (11)where n i , µ i , and N bin represent the number of observed events in the i -th bin, the expectationin the i -th bin, and the total number of bins. The Higgs production hypothesis is constructedby setting µ i to µ i = s i + b i , where s i and b i are the number of signal and expected backgroundevents in the i -th bin. This quantity s i can also be written as a product: s i = σ ( p ¯ p → W ± H ) · B ( H → b ¯ b ) · ǫ W H · Z L dt · f W Hi (12)where f W Hi is the fraction of the total signal which lies in the i -th bin. In this case, σ ( p ¯ p → W ± H ) · B ( H → b ¯ b ) is the variable to be extracted from data. An upper limit on the Higgs28 Dijet Mass (GeV/c0 50 100 150 200 250 300 350 400 450 500 ) E v en t s / ( G e V / c Dijet Mass (GeV/c0 50 100 150 200 250 300 350 400 450 500 ) E v en t s / ( G e V / c Observed W+Heavy FlavorMistagNon-W QCD ttfi Diboson/Z (6.7pb)+Single TopttBackground Uncertainty ) =115 GeV/c H
10 (m · WH FIG. 7: Dijet mass distribution in W +2 jets events including exactly one secvtx b -tagged jet thatpasses the NN b -tagging filter. The contributions of the various background sources are shownin histogram, while the hatched box on the background histogram represents the backgrounduncertainty. boson production cross section times branching ratio σ ( p ¯ p → W ± H ) ·B ( H → b ¯ b ) is extractedby using a Bayesian procedure with a likelihood defined by: L = N bin Y i =1 P i ( n i , µ i ) = N bin Y i =1 µ n i i e − µ i n i ! . (13)The background prediction b i includes contributions from the various background sourcesdescribed in Sec. V: b i = N T OP f T OPi + N QCD f QCDi , (14)where f T OPi and f QCDi are the fractions of the total number of top (including t ¯ t and singletop) and QCD backgrounds (including W+jets, non- W , and diboson) in mass bin i . Thereare systematic uncertainties in the estimates of both the number of signal events and theexpected background. Such uncertainties modify the likelihood to be L ( σ · B ) = Z N QCD Z N TOP Z N WH N bin Y i =1 µ n i i e − µ i n i ! × G ( N QCD , σ
QCD ) G ( N T OP , σ
T OP ) G ( N W H , σ
W H ) dN QCD dN T OP dN W H (15)where the G ( N, σ ) factors are truncated Gaussian densities constraints using the estimatednumbers of events and the associated uncertainties. We assume a uniform prior for σ · B Dijet Mass (GeV/c0 50 100 150 200 250 300 350 400 450 500 ) E v en t s / ( G e V / c Dijet Mass (GeV/c0 50 100 150 200 250 300 350 400 450 500 ) E v en t s / ( G e V / c Observed W+Heavy FlavorMistagNon-W QCD ttfi Diboson/Z (6.7pb)+Single TopttBackground Uncertainty) =115 GeV/c H
10 (m · WH FIG. 8: Dijet mass distribution in W +2 jets events including at least two secvtx b -tagged jets. and integrate the likelihood over all parameters except σ · B . A 95% credibility level upperlimit on σ · B is obtained by calculating the 95 th percentile of the resulting distributions.To measure the expected sensitivity for this analysis, background-only pseudo-experiments are used to calculate an expected limit in the absence of Higgs boson production.Pseudo-data are generated by fluctuating the individual background estimates within totaluncertainties. The expected limit is derived from the pseudo-data using Eq. 15.The likelihoods from events with exactly one secvtx b -tagged jet passing the NN b -tagging filter and events with at least two secvtx b -tagged jets criteria are multipliedtogether. The systematic uncertainties associated with the pretag acceptance, luminosityuncertainty, and uncertainty of the b -tagging efficiency scale factor are considered to be100% correlated between the two selection channels. Background uncertainties, specificallyon the heavy-flavor fractions and b -tagging scale factor, are also completely correlated. The“=1 tag w/ NNtag” selection combined with “ ≥ Higgs Mass (GeV/c110 115 120 125 130 135 140 145 150 ) ( pb ) b b fi BR ( H · H ) – W fi p ( p s -2 -1 ObservedExpected Upper Limit) -1 CDF Run II (319 pb) -1 D0 Run II (440 pbStandard Model (NLO)
FIG. 9: 95% confidence level upper limit on σ ( p ¯ p → W H ) ·B ( H → b ¯ b ) with an integrated luminosityof 1 fb − obtained from the combined likelihood between events with exactly one secvtx b -tagpassing the NN b-tagging and events with at least two secvtx b -tagged jets. The previous CDFdata [7] and recent D0 data [8] are shown for comparison. higher than the expected limit, but this is consistent with a statistical fluctuation in thedijet mass distributions (see Fig. 7) around m H = 115 GeV /c . Such a fluctuation is muchlarger than the expectation for SM Higgs boson production in this channel.The search sensitivity is improved significantly with respect to previous searches, about30% beyond the expectations from simple luminosity scaling. The two main effects are theseparation of the b -tagged data sample into single- and double-tagged events, and the NNfilter applied to the single-tag sample. IX. CONCLUSIONS
We have presented a search for the standard model Higgs boson in the ℓνb ¯ b final stateexpected from W H production. The event selection includes an additional neural network b -tag filter to reduce the background contributions from light flavor and charm quark jets.This improvement, along with a total dataset corresponding to 1 fb − , allows us to improvethe upper limit on Higgs boson production. We set a 95% confidence level upper limit on the31 (pb)bb fi WH) BR(H fi · p(p s N u m be r o f P s eudo - E x pe r i m en t s = 110 GeV/c H mMean = 2.2 pbRMS = 0.8 pb ) (pb)bb fi WH) BR(H fi · p(p s N u m be r o f P s eudo - E x pe r i m en t s = 115 GeV/c H mMean = 2.2 pbRMS = 0.8 pb ) (pb)bb fi WH) BR(H fi · p(p s N u m be r o f P s eudo - E x pe r i m en t s = 120 GeV/c H mMean = 2.0 pbRMS = 0.7 pb ) (pb)bb fi WH) BR(H fi · p(p s N u m be r o f P s eudo - E x pe r i m en t s = 130 GeV/c H mMean = 1.8 pbRMS = 0.7 pb ) (pb)bb fi WH) BR(H fi · p(p s N u m be r o f P s eudo - E x pe r i m en t s = 140 GeV/c H mMean = 1.7 pbRMS = 0.6 pb ) (pb)bb fi WH) BR(H fi · p(p s N u m be r o f P s eudo - E x pe r i m en t s = 150 GeV/c H mMean = 1.5 pbRMS = 0.6 pb FIG. 10: Results of 95% confidence level limits obtained from the combined likelihood in pseudo-experiments. The arrows indicate the observed limits. iggs Mass Upper Limit (pb)GeV/c Observed Expected SM110 3.9 2.2 ± ± ± ± ± ± σ ( p ¯ p → W H ) · B ( H → b ¯ b ) at 95 % C.L.,compared to the SM production rate calculated at NNLO. production cross section times branching ratio varying from 3.9 to 1.3 pb for Higgs bosonmasses 110 to 150 GeV /c . Acknowledgments
We thank the Fermilab staff and the technical staffs of the participating institutions fortheir vital contributions. This work was supported by the U.S. Department of Energy andNational Science Foundation; the Italian Istituto Nazionale di Fisica Nucleare; the Ministryof Education, Culture, Sports, Science and Technology of Japan; the Natural Sciences andEngineering Research Council of Canada; the National Science Council of the Republic ofChina; the Swiss National Science Foundation; the A.P. Sloan Foundation; the Bundesmin-isterium f¨ur Bildung und Forschung, Germany; the Korean Science and Engineering Foun-dation and the Korean Research Foundation; the Science and Technology Facilities Counciland the Royal Society, UK; the Institut National de Physique Nucleaire et Physique des Par-ticules/CNRS; the Russian Foundation for Basic Research; the Comisi´on Interministerial deCiencia y Tecnolog´ıa, Spain; the European Community’s Human Potential Programme; theSlovak R&D Agency; and the Academy of Finland. [1] P. W. Higgs, Phys. Rev. Lett. , 508 (1964).
2] R. Barate et al. (ALEPH, DELPHI, L3, and OPAL collaborations and the LEP WorkingGroup for Higgs boson searches), Phys. Lett.
B565 , 61 (2003), hep-ex/0306033.[3] J. Alcaraz et al. (ALEPH, DELPHI, L3, and OPAL collaborations and the LEP ElectroweakWorking Group), Tech. Rep. CERN-PH-EP-2006-042 (2006), hep-ex/0612034.[4] T. Han and S. Willenbrock, Phys. Lett.
B273 , 167 (1991).[5] A. Djouadi, J. Kalinowski, and M. Spira, Comput. Phys. Commun. , 56 (1998), hep-ph/9704448.[6] A. Abulencia et al. (CDF collaboration), Phys. Rev.
D75 , 012010 (2007), hep-ex/0612015.[7] A. Abulencia et al. (CDF collaboration), Phys. Rev. Lett. , 081802 (2006), hep-ex/0605124.[8] V. M. Abazov et al. (D0 collaboration), Phys. Lett. B (to be published), hep-ex/0712.0598.[9] D. Acosta et al. (CDF collaboration), Phys. Rev. D71 , 052003 (2005), hep-ex/0410041.[10] D. Acosta et al. (CDF collaboration), Phys. Rev.
D71 , 032001 (2005), hep-ex/0412071.[11] L. Balka et al., Nucl. Instrum. Methods
A267 , 272 (1988).[12] S. Bertolucci et al., Nucl. Instrum. Methods
A267 , 301 (1988).[13] M. G. Albrow et al., Nucl. Instrum. Methods
A480 , 524 (2002).[14] F. Abe et al. (CDF collaboration), Phys. Rev.
D45 , 1448 (1992).[15] A. Bhatti et al., Nucl. Instrum. Methods
A566 , 375 (2006), hep-ex/0510047.[16] G. Ascoli et al., Nucl. Instrum. Methods
A268 , 33 (1988).[17] T. Dorigo (CDF collaboration), Nucl. Instrum. Methods
A461 , 560 (2001).[18] E. J. Thomson et al., IEEE Trans. Nucl. Sci. , 1063 (2002).[19] D. Acosta et al. (CDF collaboration), Phys. Rev. Lett. , 091803 (2005), hep-ex/0406078.[20] A. Abulencia et al. (CDF collaboration), Phys. Rev. Lett. , 082004 (2006).[21] C. Peterson, T. R¨ognvaldsson, and L. L¨onnblad, Comput. Phys. Commun. , 185 (1994).[22] A. Abulencia et al. (CDF collaboration), Phys. Rev. D74 , 072006 (2006), hep-ex/0607035.[23] A. Abulencia et al. (CDF collaboration), Phys. Rev. Lett. , 082004 (2006), hep-ex/0606017.[24] M. L. Mangano, M. Moretti, F. Piccinini, R. Pittau, and A. D. Polosa, J. High Energy Phys. , 001 (2003), hep-ph/0206293.[25] G. Corcella et al. (2001), hep-ph/0201201.[26] J. Campbell and R. K. Ellis, Phys. Rev. D65 , 113007 (2002), hep-ph/0202176.[27] M. Cacciari, S. Frixione, M. L. Mangano, P. Nason, and G. Ridolfi, J. High Energy Phys. ,068 (2004), hep-ph/0303085.
28] B. W. Harris, E. Laenen, L. Phaf, Z. Sullivan, and S. Weinzierl, Phys. Rev.
D66 , 054024(2002), hep-ph/0207055.[29] T. Sj¨ostrand et al., Comput. Phys. Commun. , 238 (2001), hep-ph/0010017.[30] A. Abulencia et al. (CDF collaboration), Phys. Rev.
D73 , 032003 (2006), hep-ex/0510048.[31] J. Pumplin et al., J. High Energy Phys. , 012 (2002), hep-ph/0201195., 012 (2002), hep-ph/0201195.