The Search for Z?bb at DO

1
The Search for Z?bb at DO

• Amber Jenkins
• Imperial College London
• DO Winter Physics Workshop
• 28 February 2005

On behalf of Per Jonsson, Andy Haas, Gavin Davies
and myself
2
The Z-gtbb Story

• Motivation for the search
• How to trigger on Z?bb events
• How efficient are our triggers?
• The Data. The Monte Carlo.
• Choosing the signal box
• Subtracting the background
• Looking in data
• Conclusions and outlook

3
The Story Begins

• Z-gtbb is not revolutionary new physics. But its
  observation at a proton-antiproton collider is
  very important.
• For calibration of the b-JES, relevant to much of
  D0 physics
• As a crucial test of our jet energy and dijet
  mass resolution
• As an ideal testbed for the decay of a light
  Higgs.
• Back in the heyday of Run I, DO lacked the
  tracking or b-tagging capabilities needed for
  Z-gtbb. CDF claimed to observe 91 30 19 events
  using their Silicon Vertex Detector.

4
The Challenge

• In Run II, however, Z-gtbb is within our reach.
  With good muon detection, b-tagging the
  prospect of the Silicon Track Trigger, all the
  tools are at our disposal.
• The challenge is to fight down the massive QCD
  background swamping the signal.
• Triggering is crucial. We need to achieve
  sufficient light-quark rejection such that
  trigger rates are acceptable at high luminosity.

5
Triggering on Z?bb
Hope to incorporate a L2 STT Zbb trigger term in
v14

• The natural Z-gtbb trigger would be a low energy
  jet trigger. However, rates would be
  unmanageable.
• Ideally we would trigger on dijet events with
  displaced vertices at Level 2. This will soon
  be possible with the STT.
• In the meantime, we rely on semileptonic
  decay of b-jets to 1 or more muons
• Use single-muon dimuon triggers at
  Level 1
• Require additional jet, ? track terms at
  Levels 2 and 3
• Capitalise on power of our Impact Parameter
  b-tagging at L3

Ultimately we are limited by the BR(b??) ? 10
6
v12 and v13 Trigger Selection

• The analysis exploits data collected with both
  v12 and v13.
• For v12, pre-existing muon triggers are used 61
  in total.
• For the v13 trigger list we have designed 5
  dedicated triggers which optimise Z-gtbb signal
  efficiency while achieving required background
  rejection for luminosities up to 80E30.
• They went online last summer.

7
The v13 Suite of Triggers
v13 Trigger Description Good Luminosity Collected in PASS2 Data (pb-1)
DMU1_JT12_TLM3 L1 dimu L2 1 med ? L3 1 jetgt12 GeV, 2 ? (1 with trk-match gt 3 GeV) 47.5
ZBB_TLM3_2LM0_2J L1 1 trk-matched ?, pTgt3 GeV L2 1 med ? L3 2 loose ?, 2 jetsgt12 GeV 41.2
ZBB_TLM3_2JBID_V L1 1 trk-matched ?, pTgt3 GeV L2 1 med ? L3 IPlt0.1, 2 jetsgt12 GeV, PVZlt35 41.2
MUJ1_2JT12_LMB_V L1 1 ?, 2 trig towersgt3 GeV L2 1 med ?, 1 jetgt8 GeV L3 1loose ?, 2 jetsgt12 GeV, IPlt0.05, PVZlt35 44.0
MUJ2_2JT12_LMB_V L1 1 ?, 1 trig towergt5 GeV L2 1 med ?, 1 jetgt8 GeV L3 1 loose ?, 2 jetsgt12 GeV, IPlt0.05, PVZlt35 42.7
8
How Effective are our Triggers?
Offline cuts 1 tight muon 2 or more good jets
jet ??? lt 2.5 jet pT gt 15 GeV 1st 2nd leading
jets are taggable and tight-SVT b-tagged.
v13 Trigger L1L2L3 Trigger Efficiency w.r.t. Offline Cuts () L1L2L3 Trigger Efficiency w.r.t. Offline Cuts () L1L2L3 Trigger Efficiency w.r.t. Offline Cuts ()
v13 Trigger ??(jet1,jet2) gt 2.85 radians ??(jet1,jet2) gt 2.90 radians ??(jet1,jet2) gt 3.0 radians
DMU1_JT12_TLM3 16 15 16
ZBB_TLM3_2LM0_2J 17 18 21
ZBB_TLM3_2JBID_V 67 70 72
MUJ1_2JT12_LMB_V 87 88 89
MUJ2_2JT12_LMB_V 88 88 89
9
Data and Monte Carlo Samples
Data and Monte Carlo Samples

• Data
• Run selection Remove bad CAL/MET runs, runs with
  bad luminosity blocks select good CAL/MET runs.
• v13 Dataset (70 pb-1) PASS2 Higgs skim 45M
  events containing our Z-gtbb triggers collected
  from June 2004 ? August 2004.
• v12 Dataset (200 pb-1) PASS2 BID skim 90M
  events containing at least 1 loose muon 1 0.7
  cone jet
• PYTHIA Monte Carlo signal plus bb and
  light-quark inclusive QCD backgrounds covering a
  wide pT spectrum (see next slide).
• Full jet energy scale corrections are applied to
  all samples using JetCorr v5.3.

10
Monte Carlo Samples
PYTHIA-Generated Sample Number of Events
Z-gtbb (p14.03.02) 84,000
bb QCD, 5ltpTlt20 GeV 202,240
bb QCD, 20ltpTlt40 GeV 123,750
bb QCD, 40ltpTlt80 GeV 148,750
bb QCD, 80ltpTlt160 GeV 72,500
bb QCD, 160ltpTlt320 GeV 23,250
Light-quark QCD, 5ltpTlt10 GeV 200,000
Light-quark QCD, 10ltpTlt20 GeV 200,000
Light-quark QCD, 20ltpTlt40 GeV 300,000
Light-quark QCD, 40ltpTlt80 GeV 250,000
Light-quark QCD, 80ltpTlt160 GeV 300,000
Light-quark QCD, 160ltpTlt320 GeV 50,000
p14.07.00
11
Kinematic Handles

• We investigate which kinematic tools provide best
  discrimination between signal background.
• To eliminate essentially all of the light-quark
  QCD, it is sufficient to require 2 b-tagged jets
  in each event.
• There are few handles which really cut back the
  bb background
• After b-tagging, the main difference between
  Z-gtbb bb QCD background is colour flow in the
  events
• Expect more colour radiation in QCD processes
• Expect pattern of radiation to be different
• Study number of jets per event, njets
• Study azimuthal angle between 2 b-quark jets,
  ??12

12
Using dphi as a Discriminator
Passing v13 triggers
njets gt 2
njets 2

• data
• Zbb MC
• bb MC

njets 3
13
Offline Event Selection

• Initial dataset is cleaned up by
• removal of noisy jets
• requiring a tight offline muon
• After triggering we apply
• - jet ??? lt 2.5
• - jet pT gt 15 GeV
• - ensure 1st 2nd leading jets are
  taggable tight-SVT b-tagged
• - require njets gt 2
• - require ?? gt 3.0
• We are completing the tuning of these cuts,
  after v13 trigger selection.

• - Zbb MC
• bb MC

Mass window cut
14
Signal Peak in Monte Carlo
Z mass low!
15
Signal Event Predictions
Trigger Predicted Number of Signal Events
ZBB_TLM3_2JBID_V 134
ZBB_TLM3_2LM0_2J 39
MUJ1_2JT12_LMB_V 166
MUJ2_2JT12_LMB_V 166
DMU1_JT12_TLM3 30

• Calculate no. of signal events expected per
  trigger, accounting for luminosity,
  cross-section, trigger and offline efficiency
• Predictions are for ?? gt 3.0 and njets gt 2

16
Background Subtraction
njet 2, dphi gt 3.0

• Key to this analysis is understanding the
  background
• The method (a la CDF Run I)
• Define 2 regions
• IN SIGNAL ZONE evts which pass
  njets 2 dphigt3.0
• - OUTSIDE ZONE all other evts
  (which fail above IN
  ZONE conditions)
• 2. Calc. Tag Rate Function (TRF) of
  double/single tagged evts OUTSIDE ZONE
• 3. Expected Bkg IN ZONE
  TRF (single-tagged
  evts IN ZONE)
  i.e.
  Nexp,IN Nobs,IN (Nobs,OUT/Nobs,OUT)
• 4. IN ZONE, Subtract Expected Bkg from
  Observed Events. An excess around 90 GeV is
  bias-free evidence for a signal.

IN ZONE
3.0
OUTSIDE ZONE
dphi
njet
2
3
17
1. The Invariant-Mass Based TRF

• Construct an invariant-mass based TRF
• The single-tag mass histogram IN ZONE is
  multiplied bin-by-bin by this TRF to give a
  background estimate
• Background then subtracted

18
1. The Invariant-Mass Based TRF (contd)
Estimated Background
Excess
High mass excess
19
2. Correction for ?? Dependence of TRF

• Ratio of double to single only tight SVT tagged
  events with dphi lt 3.0 (black) and dphi gt 3.0
  (red)
• Peaked increase in the Z region for dphigt 3.0
• The probabilities do not, however, converge at
  higher masses
• This implies we are underestimating the bkgd in
  this region

20
2. Correction for ?? Dependence of TRF (contd)

• We observe a linear dependence of TRF on ??, in
  both MC and data

• We derive the TRF correction outside the signal
  zone.
• The correction is 15 for 2.8lt??lt3.1.
• Note that the signal estimate is conservative.
  We assume Zbb is only found in signal zone,
  which is not accurate.

Treat as if no signal in this region this is
conservative
21
3. The Jet-Based TRF

• We improve the background estimation by moving to
  using a jet-based TRF a la the Hbb analysis (see
  Andys talk)
• We consider events where the 1st leading jet is
  single-SVT tagged
• For these events, we then consider 2nd ldg jet.
  In three eta regions for the 2nd leading jet, we
  calculate the TRF as function of ET.
• This generates a TRF per jet. It is still based
  on events outside the signal zone.
• Each event is then weighed accordingly. This is
  likely to be a more accurate method as it
  provides a finer resolution to the correction.
• We still include the 15 correction for TRF
  dependence on dphi.

22
Background Subtraction Using this Refined Method

• Excess seen in v12 data

S/?(SB) 5
Estimated background
490 ? 22 events
Background shape well-modelled
23
Conclusions

• We are searching for Z? bb in v12 and v13 data
• Different methods to estimate the background are
  being tested.
• Out of 9294 selected double-tagged events, we
  observe an excess after background subtraction of
  490 ? 22 events.
• Analysis cuts are being finalised.
• The Analysis Note is being prepared for group
  review.

24
To Do List

• Complete tuning of cuts in v13 triggers
• Include extra 100 pb-1 of data from v11 runs and
  before
• Systematic errors
• Update note
• Take to Wine Cheese seminar

25
Sources of Error

• JES
• Btagging
• Trigger efficiencies
• Luminosity
• Jet reco/ID
• Statistics outside signal zone
• TRF dep on dphi