Observation of Single Top Quark

Presentation on theme: "Observation of Single Top Quark"— Presentation transcript:

1 Observation of Single Top Quark Production @Tevatron
2009/11/10

2 Tevatron フェルミ国立加速器研究所 円周6.3km 陽子・反陽子 √s = 1.96 (0.98+0.98) TeV ２つの測定器
CDF(Collider Detector at Fermilab) D0 ~1km p

3 Top Quark Production Pair Single 測定の難しさ
strong interaction (gtt vertex) q g t b W+ W- 1995 Single weak interaction (Wtb vertex) u g b W*+ t W+ l ν t-channel d s-channel 2009 測定の難しさ B.G. に対してCross section が小さい。 理論値 σst ~ 2.9 pb tt pair (σ ~ 7 pb) の約40% Single top event である特徴が少なく、バックグランドとの分離が難しい。 Cut base event selection した後でさえ、S/N比が5~6% → 様々な多変数解析(Multivariate technique ) が行われた。 解析が難しい原因：Pair-production(4本) に比べてジェットの数が少ない（2、3本） S/N 比が悪いため、単純に事象数の計測から求めることができない。 → わずかな運動量分布の違いを利用した多変数解析

4 Motivation |Vtb| の直接測定 いままではtop の崩壊から間接測定 top quark の偏極特性の観測
Source of single ~100 % polarized top quarks ヒッグス粒子探索のためのB.G. 理解 WH モードと終状態が同じ 新物理の探索 W’ or H+ and FCNC (ug → t) W+ q q’ H b l ν u g b t W+ l ν W’+ u d t b W+ l ν

5 Signal events tree-level matrix-element generator を用いてシミュレーションした。
(MADEVENT) t → bW (100%) S/N 比をよくするため、W がleptonic に崩壊したイベントを探した。 u g b W*+ t t-channel d s-channel

6 B.G. events QCD multi-jet events W + HF jets
q(e) QCD multi-jet events q’ b W l ν W + HF jets + mistags ( mistakenly b-tagged light-flavor jets) → Prediction derived from data control samples q g t b W+ W- tt-pair W l ν Di-boson → MC prediction scaled to the total integrated luminosity

7 Selection Cut l + missing-Et + 2 or 3 jets Veto
e or μ: Isolated leptons with Pt > 20 GeV ν : missing-Et > 25 GeV jets : Et > 20 GeV and |η| < 2.8 At least one jet identified as b-tagged (displaced secondary vertex algorithm) Veto Di-lepton To reduce the Z + jets, tt-pair and di-boson multi-jet events without a leptonic W decay photon and cosmic ray u g b W*+ t W+ l ν t-channel (W+3-jets) d s-channel (W+2-jets)

8 Selection Cut (no lepton)
レプトンを検出できなかった事象を探す ( → MJ analysis で用いる) missing-Et + 2 or 3 jets no lepton : veto events selected for “l + missing-Et + jets”. ν : missing Et > 50 GeV ( ←大量のQCD B.G. を除去) jets : Two jets within |η| < 2.0, at least one of which has |η| < 0.9 1st jet Et > 35 GeV, 2nd Et > 25 GeV angular separation between the two jets, ΔR > 1.0 Veto four or more jets with Et > 15 GeV in |η| < 2.4 To reduce the QCD multi-jet and tt-pair B.G.. B-tagging Neural network (NNQCD) to reduce the QCD multi-jet B.G.. missing-Et とmissing-Pt の絶対値 Angle between missing-Et and missing-Pt Azimuthal angle missing-Et or missing-Pt and jet directions → Removing 77 % of the QCD B.G. while keeping 91 % of the signal acceptance.

9 Yield and Multivariate Analysis
Selection cut をした後でもB.G. が圧倒的にdominant。 Multivariate technique Likelihood function for t-channel (LF) for s-channel (LFS) Matrix element (ME) Neural network for l + missing-Et + jets (NN) for missing-Et + jets (MJ) Boosted decision tree (BDT) Improved b-tag (bnn) neural network tool (NEUROBAYES) Trained to distinguish b jets from charm and light-flavor jets based on secondary vertex tracking information. Sig. B.G. ME … 生成過程の不変振幅をもちいた手法 LFSは部分集合のデータ。

10 Likelihood Function for t-channel (LF)
Combine several sensitive variables into a single one 2つのlikelihood function を用意した。 2-jet イベント用 (L2j) … 7 input variables 3-jet イベント用 (L3j) … 10 input variables ビンの中のシグナルの割合 signal B.G. Likelihood function

11 Likelihood Function for t-channel (LF)
b-tagged jet is assumed as a jet from top decay. For double-b-tagged event, L2j is set to zero. 7 Input variables for L2j Improved b-tagging (bnn) Reconstructed top mass (Mtop) Di-jet mass (Mjj) lepton charge x pseudo-rapidity of jet not assigned to be the b from top decay (Q x η ) Total scalar sum of the transverse energy in the event (Ht) cosθ*lj (θ : between the lepton and jet not assigned to be the b from top decay, in the top quark rest frame) t-channel matrix element

12 Likelihood Function for s-channel (LFS)
2つ以上のb-tagged jets を要求 (→ 609 events) W+HF と tt-pair がmain B.G. Kinematic fitter を使用。 z component of neutrino momentum b-jet that most likely came from the top decay Important input variables Output of the kinematic fitter Invariant mass of the two b-tagged jets (Mbb) Transverse momentum of the bb system Reconstructed top mass (Mtop) Total scalar sum of the transverse energy in the event (Ht) Leading jet transverse momentum Missing-Et

13 Matrix Element (ME) Compute an event probability for signal and B.G. based on calculations of the SM differential cross sections. PDF (parton distribution functions) Transfer function (detector resolution effects) Lepton and jets 4-vectors Partonic quantities Event Probability Discriminant B-tagging by NN

14 Neural Network (NN) Network are developed using NEUROBAYES analysis package. Important input variables Improved b-tagging (bnn) Reconstructed top mass (Mtop) Di-jet mass (Mjj) lepton charge x pseudo-rapidity of jet not assigned to be the b from top decay (Q x η ) Total scalar sum of the transverse energy in the event (Ht) cosθ*lj (θ : between the lepton and jet not assigned to be the b from top decay, in the top quark rest frame) Transverse mass of the W boson

15 Boosted Decision Tree (BDT)
Decision tree method : Sequence of binary splits using the discriminating variable which gives best sig-bkg separation. Boosting algorithm : Events misclassified during DT training are given a higher weight in the next training. Use over 20 input variables

16 Data Check The modeling of each input variable was checked by the B.G. dominated data control samples. l + missing-Et + jets analysis l + b-tagged 4 jets samples Enriched in tt-pair events 2 or 3 jets samples (no b-tagging) 3 jets samples are enriched in W+jets and QCD events Mismodeling がないか2000以上の分布を確認した。 small discrepancies Angle between two jets in the untagged lepton + 2-jets samples Modeling of jets with rapidity greater than 2.4 → systematic uncertainties に含めている。

17 Output from “l + missing-Et + jets” Analysis

18 Combine Combine the LF, ME, NN, BDT, and LFS channel
Superdiscriminant method (NEAT : Neuro Evolution of Augmenting Topologies) Neuro-evolution network Best linear unbiased estimator (BLUE) 同一のデータを用いて、複数の方法で1つの物理量を測定した（相関がある）場合、どのように結果をまとめればよいか？ 相関なし… “ 1/σi2 ”のウェイトを掛けて平均をとる。 相関あり…σ2 が最小となるようにαi を決める。

19 Combined Output 一番右側のビンで、 significance が5σ 以上

20 Analysis of events with Missing-Et and Jets (MJ)
レプトンを検出できなかったイベントを解析 Data : 2.1 fb-1 Use a neural network Important input variables Invariant mass of the miss-Et Second leading jet Scalar sum of the jet energies missing-Et Azimuthal angle between miss-Et and the jets

21 Measurement of cross section
Simultaneous fit over two exclusive channels (SD and MJ) Bayesian binned likelihood technique を使用。 →Cross section : 2.3 ( stat+syst) pb → |Vtb| = 0.91 ± 0.11 (stat + syst) ± (theory) SD discriminant MJ discriminant

22 Significance and Sensitivity
信号の有意性を調べるためにQ値を定義 実際のデータが観測されたQ値が観測される確率 p = 3.10x10-7 → 5.0 σ (significance) data = b, s+b の場合のQ値分布

23 Summary Single top production を観測した。
Data : 3.2 fb-1 (ppbar-collision, √s = 1.96 TeV) Significance : 5.0 σ Sensitivity : 5.9 σ Cross section : 2.3 ( stat+syst) pb Cross section から | Vtb | を求めた。 |Vtb| = 0.91 ± 0.11 (stat + syst) ± (theory) |Vtb| > 0.71 (95% C.L.)