PILEUP studies for Hbb

1
PILE-UP studies for H-gtbb

• Marek Taevský (Physics Inst. Prague)
• FP420 meeting - Manchester 10/12 2006

Irreducible background Pile-up
2
H-gtbb and H-gtWW in SM

• Both the signal and bg studied at detector level
  using FAMOS.
• The following packages used in the analyses
• - Fastcalorimetry,FastTsim, FastBtag, FastJets,
  
• - FastMuon, FastMuonTrigger, FastTotem (just
  Roman Pots)
• Jet algorithm
• o) Iterative cone, Cone radius 0.7
• o) Jet energy scale corections applied to
  detector level jets
• Signal for H-gtbb composed of two main categories
• 1) b-dijets 85 of signal, challenging channel
  topic of the talk
• 2) muons from B-decays .. 15 of signal, clean,
  no problems

3
Roman Pot acceptances for Totem and CMS
4
Irreducible bg to H-gtbb

• Sources of bg itself without considering PU
• Generated and fast simulated
• 1. DPEMC - Excl. DPE bb (ptgt30 GeV)
• (s 520 fb, 60k)
• 2. DPEMC - Incl. DPE dijets (ptgt30 GeV)
• (BPR model s 30 nb, 170k, CF model s
  13.2 nb, 130k)
• Reduction factor from double b-tag
  requirement 500
• 3. PYTHIA - Non-diff. dijets (30ltptlt50)
• (s156 µb, 1 mil.)
• Jet selection efficiency 5.6
• Reduction factor from double b-tag
  requirement 560
• 
  
• 4. PYTHIA Non-diff. dijets (50ltptlt80)
  
• (s 20 µb, 1 mil.)
• Jet selection efficiency 24.6
• Reduction factor from double b-tag
  requirement 310

5
P(g/b) and P(anything/b)

• Red. factors from double b-tag are 500, 560 and
  310 for Incl. DPE dijets
• (ptgt30 GeV), non-diffr.dijets (30ltptlt50) and
  non-diffr.dijets (50ltptlt80), resp.
  
  
• Why the red.factors from double b-tag requirement
  arent close to 104
• (corresponding to P(g/b)1 as expected)?
• Because in the inclusive dijet processes, there
  are
• - gluons light quark dijets which may be
  mis-b-tagged
• - bbbar pairs other sources of b-jets which are
  correctly b-tagged
• 
  
• 
  
• 
  So P(anything/b)4-5 looks realistic
• 
  
• 
  Some rough numbers
• 
  P(g/b)2.5 for eb55 (CMS)
• 
  P(g/b)1.3 for eb60 (ATLAS)
  

Pythia 6.4
Physics sources of b-jets in inclusive dijet
production
6
Phojet generation of PU events

• All processes 118 mb
• Non-diff.inelastic 68 mb
• Elastic 34 mb
• Single Diffr.(1) 5.7 mb
• Single Diffr.(2) 5.7 mb
• Double Diffr. 3.9 mb
• DPE 1.4 mb
• Number of pile-up events per bunch crossing (BX)
  ? NPU
• Lumi x cross section x bunch time width x LHC
  bunches/filled bunches
• 1034cm-2s-1x104cm2/m2 x 10-28m2/b x 110mb x
  10-3b/mb x 2510-9s
• X 3564/2808 35
• 51033 17.6 , 21033 7.0, 11033 3.5,
  11032 0

7
PU mixing to S and B for H-gtbb

• Generated and fast simulated
• 1. PHOJET MB s 118 mb, 200k
• Prob. to se a proton in one RP per interaction
• 420 1.0, 220 3.1 - These numbers are
  crucial for the PU study! Unfortunately, they are
  model-dependent. Which data to take to tune our
  MC models??? ZEUS? CDF?
• 2. PYTHIA Non-diff. dijets properly mixed
  with PU events
• NPU 3.5 7.0
  17.6 35.0
• --------------------------------------------------
  ---------------------------
• pt 30-50 50-80 30-50 50-80 30-50
  50-80 30-50 50-80
• Stat. 3.1 M 2.5 M 2.5 M 2.4 M 420k
  440k 428k 399k
• 3. DPEMC Excl. DPE bb (pthatgt30 GeV), 30k
• 4. DPEMC Incl. DPE bb (pthatgt30 GeV), 90k

8
For too eager and too unpatient
CDF points just tentative!

• Just to have a look at what correction factors we
  might get.
• (before KMR come up with their precise
  calculations)
• The main idea from KMR concentrate on CDF
  measurement rather than
• on ZEUS because of closer type of collision and
  closer Ecm.
• The xL spectrum expected to be almost Ecm
  independent between
• 1.8 TeV and 14 TeV. Still need to check the CDF
  data. What about
• t-ranges of CDF and ZEUS measurements and of
  Phojet?

9
CDF measurement of dsSD/d?
Practically no Ecm Dependence for ?lt0.03 and a
weak Dependence around ?0.1 CDF data may be
used for rough estimates
10
Fake rates of protons in RPs

• Mix PU events with signal or bg using FAMOS
• Sum RP acceptances over all possible proton pairs
  in all PU events
• in one BX and then look at mean over all
  signal or bg events.
• NPU properly smeared using Poisson dist.
• E.g. RRP420 ltSiNPU(n) SjNPU(n)
  AL420(i)xAR420(j)gtn90k signal or bg events
• Rate of PU events with 2 ps
  seen in opposite 420 RPs

11
Fake rates

• Analytical formula for Rate of p pairs seen in
  opposite RPs per BX
• µ AccNPU, µLR AccLRNPU
• NRP/BX(binomial) 2exp(-µ)cosh(µ) 1 1
  exp(-µLR)
• 
  Courtesy
• NRP/BX(simple) NPU(NPU-1)AccAcc
  Sasha Kupco
• 
  (Prague)
• L3 (NRP/BX)/Q, Q red.fact.from quartic det.
  based on st10 ps
• Q 30 for 220 and 420,
  for NPU 3.5
• 13 for 220 and 23
  for 420, for NPU 35

12
Selection cuts for H-gtbb at mh120 GeV
13
Control plots
Excl.DPE H-gtbb Incl.DPE dijets
Non-diffr.dijets PU 35
14
Effect of PU on DPE processes

• Event yields for Lumi 30 fb-1
• Excl. DPE H-gtbb practically immune to PU
• Excl. DPE bb
• NPU 0.0 3.5 7.0
  17.6 25.0 35.0
  
• --------------------------------------------------
  --------------------------------
• conf 420 comb420 comb420 comb420
  comb420 comb420 comb
• --------------------------------------------------
  --------------------------------
• 1 8 2 8 1 6 2
  6 2 6 --- ---
• Incl. DPE dijets (BPR model) Red.factor from
  double btag 500
• NPU 0.0 3.5 7.0
  17.6 25.0 35.0
  
• --------------------------------------------------
  --------------------------------
• 20 10020 80 70 30 230
  170 200 270 --- ---
• Incl. DPE dijets (CF model) Red.factor from
  double btag 500
• 6 8

Large model dependences for Incl.DPE dijets. New
H1 data on diffr.pdf may lower the effect even
further
15
Consistency checks
Andys results from last FP420 meeting at CERN
16
Consistency checks

• Nev for Lumi 30 fb-1
• Andy This
  analysis
• Excl.DPE H-gtbb 1.6 0.6
• Excl.DPE bb 3.0 1.0
• Excl.DPE gg 2.4 --
• Incl.DPE bb 2.4 --
• Incl.DPE dijets 0.015 6.0

17
Non-diffr. Dijets PU

• NPU 3.5 7.0
  17.6
  35.0
• --------------------------------------------------
  -----------------------------------------------
• pt 30-50 50-80 30-50 50-80
  30-50 50-80 30-50 50-80
• Stat. 3.11 m 2.52 m 2.46 m 2.39 m
  420k 440k 428k 399k
• --------------------------------------------------
  -----------------------------------------------
• Scaling 67.2 19.2 85.0 20.2
  497.4 110.0 488.2
  121.0
• --------------------------------------------------
  -----------------------------------------------
• All cuts
  420comb420comb420comb420comb
• But c10 1.00.00.94.3 2.02.21.05.43.0
  0.3 1.0 7.5 15.0 9.0 7.0 10.9
• --------------------------------------------------
  -----------------------------------------------
• Reduction from double b-tag 560 (30ltptlt50) and
  310 (50ltptlt80)
• Reduction from quartic40
• Nr.of surviving events per L30fb-1
• 70 - 2080170190201101500130110830
  730044008501300
• 170 490
  2600 13900

18
Cuts flow - efficiencies

• Signal BG Pythia NPU
  0.0
• --------------------------------------------------
  
• pt 30-50
  50-80
• Stat. 10k 1.0m 0.96
  m
• --------------------------------------------------
  -
• Jetcuts 34.2 4.4 18.3
• --------------------------------------------------
  -
• 2 btag 36 0.18 0.32
• --------------------------------------------------
  -
• RP acc 420 comb
• 32 25 0 0
• --------------------------------------------------
  -
• All cuts 1 1.6 0.0 0.0
  
• NPU 3.5
  7.0
  17.6
  35.0
• --------------------------------------------------
  --------------------------------------------------
  -------------------------------------
• pt 30-50 50-80
  30-50 50-80 30-50
  50-80 30-50 50-80
• Stat. 3.11 m 2.52 m 2.46
  m 2.39 m 420k
  440k 428k 399k
• --------------------------------------------------
  --------------------------------------------------
  -------------------------------------

19
S/B for Lumi 30 fb-1

• 3.5 7.0
  17.6 35.0
• ----------------------------------------
  -----------
• Jetcuts 19.2/33.3E10 19.2/35.7E10
  19.2/39.2E10 19.2/35.6E10
• --------------------------------------------------
  --------
• Jetcuts 6.4 / 7.5E8 6.4 / 8E8
  6.4 / 8.5E8 6.4 / 7.7E8
• 2btag
  
• --------------------------------------------------
  --------
• 420 comb 420 comb 420
  comb 420 comb
• ----------------------------------------
  -----------
• Jetcuts 2.1 1.6 2.1 1.6 2.1
  1.6 2.1 1.6
• 2btag / / / / /
  / / /
  
• RP 2.3E6 10.5E65.4E6 29.6E6 28E6
  130E6 77.8E6 293E6
• --------------------------------------------------
  --------
• All cuts 0.6 / 0.9 / 0.6 / 0.9 /
  0.6 / 0.9 / 0.6 / 0.9 /
• timing 85 80 190 300
  1600 970 8200 5700
• --------------------------------------------------
  --------

20
Multiplicity cuts at detector level

• A very preliminary look at multiplicity cuts at
  det.level (ATLFAST)
• Kt jet alg., ET,jet1,2 gt 40 GeV, ?jet1,2lt2.5,
  ??jetlt1.5, 170lt?Fjetlt190
• - pT,track gt 0.5 GeV, ?tracklt2.5
• Arbitrary normalization
• Nc Nr. of tracks outside cones of R0.5, 0.7,
  1.0 around axes of two lead. jets
• Ncperp dtto perp. to jet axes
  (p/3ltFjet-Ftracklt2p/3, 4p/3ltFjet-Ftracklt5p/3)
  
• Excl.DPE H-gtbb Non-diffr.dijets
  Looks promising

By V.Juranek
21
Summary of PU effect in H-gtbb channel

• - PU has no effect on signal.
• - PU has a little effect on jets.
• - The main effect comes from fake protons in RPs
  overlaied with
• hard jets from dijet events.
• We need to reduce BG by a factor of 100 to keep
  S/B1.
• BUT! RESULTS ARE NOT FINAL! There are a few big
  uncertainties in this study!
• New H1 Pomeron pdfs lower pdf at high z -gt
  lowers Incl.DPE contrib.
• One-sided RP acceptance for PU events (1.0 for
  420, 3.1 for 220)
• Phojet not tuned for Ecm14 TeV
• 3) B-tagging P(g/b) quite large (in CMS).
• 4) The same JES factors taken for all lumi.
  Should be lumi-dependent.
• 5) Particle multiplicity outside jets is it as
  effective as at hadron level?

22
B A C K U P S L I D E S
23
SM Excl.DPE H-gtbb Mh dependence

• Masses beyond 120 GeV not interesting in SM but
  in MSSM (x-sections may be
• increased by a factor of upto 40 in some mA,tanß
  regions for some scenarios).
• Mass windows (?M) used only for S/B studies. Two
  window widths used narrower for (420420)
• and broader for combined RP configs. Mh120
  s4201.6-gt?M4GeV and scomb5.6-gt?M10GeV.
• S/B(?M) SExcl.DPE H-gtbb, B Excl.DPE bb
  Incl.DPE bb
• SExHuMe 1.3.1, BDPEMC 2.4
• Effect of PU Signal immune
• Excl.DPE bb mutiplies by
  factor 2 indep.of Lumi
• Incl. DPE bb increases by
  2,4,11,19,21 for NPU3.5,7.0,17.6,25.0,35.0,
  respectively

24
Consistency checks

• Study of Andy Pilkington S/B 1.8 at L21033
  (NPU7) and 420420
• My study S/B1/100 at
  L21033 and 420420
• Differences between the analyses
• Andy
• looks only at HW generator level (I use Py
  det.level)
• uses the simple quadratic formula for fake rates
  of protons.
• at the lumi of interest, this means a factor
  2 to the real fake rates
• 2) uses only bb-dijets for non-diffr.bg and
  assumes double b-tag eff. 36. The latter is OK
  but one should take inclusive dijets
• and consider all sources of b-jets there
  (g-gtbb) and consider
• also misidentification of c-jets as b-jets.
• 3) applies a multiplicity cut and sees a 99
  rejection power to
• bg and only 10 loss to signal. He claims the
  multiplicity depends
• on soft underlying event tuning (its
  different between HW and Py)