Dan Holmes

1
Investigations into the physics channels
gg/qq?bbH(???) Dan Holmes, student at Bristol.
( )
2
Overview

• I looked at 1 and 2 b-jet requirement for each
  of
• SM qq/gg?bbH(??), MH125 GeV
• MSSM bbh(??). tanß30, MA144 GeV (Mh127)
• MSSM bbH(??), tanß30, MA116 GeV (MH128)

b
b
H
gg
?
?
qq
H
b
b

• Motivation measurement of Yukawa b-H coupling,
  ratio of H-b/H-t etc.., Higgs verification, MSSM
  verification
• I used CMSIM vers 4.801 with PYTHIA vers 6.205
• Also Spiras HQQ 1.1 (Hbb production x-sect) and
  HDECAY 3.0 (BR of H???).

3
Background

• The principle background for this channel is
  qq/gg??? a jet (or two) coming from ISR/FSR
  ..PU etc...
• Secondary background from qq/gg?qq two gammas,
  either real or faked by pions.
• treating the two ? from the second background as
  independent, I can put an upper limit on
  background (2) of lt 5 events per 100fb-1 as long
  as bTaggt0 for 2 jet case and gt1 for 1 jet case MH
  windowlt/-3GeV. carry these constraints
  forward to optimisation stage

4
What have I done? (since last talk)

• Used latest Spira progs. to predict where the
  biggest effective x-sect for my channels lies in
  MA-tanß space. ..latest HQQ, LEP max mixing
  scenario
• Re-assessed NLO x-sect for channels, introduced K
  factor of 1.6 (Spira hep-ph/0309204, 17th Sept.)

5
eg. MSSM h using HQQ, HDECAY.
Effective x-sect vs MA. Several values of tanB
from 5 to 30.
So biggest x-sect is 1.20 for MA 144, tanb30
(Mh127.1)
6
chose the following
...Spira then suggests multiplication by K1.6
for NLO/LO for all three channels.
7
Production

• I made some events 200K for each of smH, mssmH
  and mssmh. Pythia?CMSJET
• I made some background 8 million qq/gg??? and 4
  million qq/gg?qq.
• All included 17.3 superimposed min bias events
  as PU. Limited to recycling a PU sample of 25k
  events due to file size.

8
some pics

• signal (left) , principal background (right)

..both decay exponentially. The background is
bigger but decays faster.
9
Apply some cuts and then optimise them

• basic CMS acceptance cuts
• di-? trigger, Pt ?s gt 40,25.
• both ? isolated in tracker and ECAL.
• CMSJET modified UA1 type jet-finder. (cone r0.5)
• 1 or 2 jets Pt gt 20.
• FATSIM b-tagging utility. ff-bar back limits
  bTaggt0 for 2-tags, gt1 for 1 tag.

• Optimisation cuts investigated
• Pt?1, Pt?2, delta PtGam (Pt?1-Pt?2)
• Pt Jet1, Pt Jet2, deltaPt Jet (PtJet1-PtJet2)
• b-tagging efficiency
• Higgs mass window size
• Pt H
• topological cuts ..not so useful, but maybe
  useful for background (2).

10
optimisation procedure

• Define an optimisation parameter.
• Chose the Bityukov S2(v(sb) -vb).
  Significance for 50 chance of discovery at a
  planned experiment. s and b taken for 100 fb-1 of
  data
• BUT this is an approximation which can break
  down at low s,b, so check at all times the
  behavior of the function against a Poisson
  P-value and an event counting likelihood method.
• Optimise signal samples against the main ??
  background scan over N-d cut space, follow the
  significance gradient and define islands.
• Scan step size is not too small otherwise we are
  sensitive to stat fluctuations.

11
smH signal vs ?? background
similar stuff for mssm cases, signif rises to
about 2.7 for h.

• the signal is very small
• for the 1-tag case, some significance can be
  found (2 to 2.7 depending on sm/mssm). For 200
  fb-1 i can get it over 3...
• for such a small signal optimisation, very
  accurate knowledge of the background is needed,
  perhaps PYTHIA is not adequate.

12
recently..

• ..compHEP (Mikhail Dubinin)
• 5k evts irreducible back (bb??) 70 fb
• 50k jetjet?? (light quarks/glu) 60 pb
• ..i ran it through cmsjet (pythia
  hadronisation)..added PU.

..doesnt look good.. ..but i dont have enough
events to try and do some optimisation...
20 events from the bb ?? back.
13
conclusion

• i will have a look using alpGen ...hopefully i
  can make myself some big enough stats to finish
  this before xmas..
• maybe more tuning (loosening!) of jetfinder
  params, isolation ? of etc.. will make some
  difference..? (clutching? ...straws?)
• dont bank on this being a major CMS channel.

14
spare stats slide

• Bityukov method can be calc exactly but it is
  just too slow to do it for every step value,
  check that my approx is still valid against
  behavior of following
• P-Value Sngtsb e-bbn/n!
• 1/ I calculate 1st 4 terms explicitly but i just
  cant for big n.
• 2/ I try the approx bn/n! (b/n)(b/n-1)(n/n-2)..
  . but even this breaks down with big nums..
• Counting type likelyhood stat LLs/(LsLb)
• where Lii/i accept cuts
• I cant do a proper likelihood method as my
  cuts are too interdependent.