Photon and ?-Jet Reconstruction in the STAR Endcap EMC; Towards ?-Jet Constraints on ?G

1
Photon and ?-Jet Reconstruction in the STAR
Endcap EMC Towards ?-Jet Constraints on ?G

• motivation with focus on Endcap (EEMC)
• issues and challenges (briefly)
• simulation, data and analysis techniques
• ?/p0 shower shape discrimination with the ESMD
  shower max detector
• status and outlook

DNP08, 24 October 2008, Oakland CA
2
? - Jet Coincidence Measurements Why?

• Direct ? dominated ( 90 of yield) by QCD
  Compton process qg ? q?, with large LO gluon
  spin sensitivity

• Inclusive ? cannot compete statistically with
  incl. jet ALL but ?-jet conic. meas. a golden
  channel

• Select kinematics to optimize ?G(x) sensitivity
  high xq ? high ?fq/ fq (large quark
  polarization)

?
backward ? ? large aLL
(cross section also peaks here!)

• For ?-jet coincidences, pT?, ?? ,?jet ? x1, x2
  and the angle ? can be determined
  event-by-event.
• One uses high-x quarks (where most polarized) to
  probe low-x gluons (where they are abundant)

• above very asymmetric collisions ? ?s boosted
  into STAR Endcap EMC

2
3
STAR Endcap EMC Component Overview

• Scintillating strip SMD, 288 strips each per u
  and v planes
• WLS fiber - 16-anode MAPMTs
• 30o sectors w/ no gaps
• 1 mm peak resolution

Fully installed and operating since 2005

• Pb/Scint sampling e.m. calorimeter
• Covers 1.09 lt ? lt 2 over full azimuth
• 720 projective towers ( 22 ?0)
• 2 preshower layers, postshower layer, and
  shower max. detector (SMD)
• L0 trigger- high tower, jet patches

3
4
Dominant background to prompt ? production
p0(?)???
? - Jet challenge of rare probes

• Significant ?G(x) constraints at achievable ?L
  dt requires ? (-jet) ID well below original pT
  10 GeV/c plan.
• ?/p0 1/10 at pT10 GeV, but only 1/40 at pT5
  GeV
• how low in pT can analysis be pushed while
  retaining high efficiency and purity? - need
  clever algos for ?/p0 separation and overall
  bkgnd reduction (e.g, use shower max, preshower
  along w/ full detector response).

• charged particle vetoing from tracking with the
  STAR TPC (time projection chamber) gives out
  near middle (? 1.5) of the Endcap

30o sector tower reponse vs. preshower condition

• tower response from initial analyses shows
  strong ? dependent bkgnd yields in both data
  and simulations we hope/need to suppress these
  via cuts on the full detector response!

4
5
Photon Reconstruction for STAR (Spin) Physics
Main goal use realistic MC simulations to
discriminate efficiently effectively between
direct ? QCD background evts, compare to 2006
data

• Software tools
• isolation cuts remove events where ?
  accompanied by jet fragments
• SMD response ensure energy dist. in SMD
  consistent with single shower
• pre- / post-shower exploit differing conversion
  efficiencies / discriminate against hadronic
  showers
• away-side jet require back-to-back to reduce
  background, pT matching
• complete detector response -gt LDA

• Data samples
• MC and SMD data-driven MC of ?-jet events for 5
  lt pT lt 35 GeV/c
• Similarly MC and modified MC for QCD background
  events
• initial set 3 lt pT lt 65 GeV/c
• filtered set 3 lt pT lt 65 GeV/c
• pp_long polarized data from 2006 run use only
  events from L2_gamma trigger for now
• Note different pT samples combined with proper
  weighting, normd to 3.1 pb-1

Status of ?-jet analysis

• Emphasis to date has been on Endcap photons
  barrel (fully recon.) jets
• Two approaches 1) di-jet jetfinder approach
  w/ selection of gamma-like and recoil jets for
  addtl analysis and 2) gamma tree and jet
  tree approach, which combined produce gamma-jet
  candidates for additional cuts and algo analysis.

5
6
? - Jet Analysis and Detector Response

• initial jetfinder (di-jet) type analysis
• a sequence of cuts select gamma and away
  side jets

1. N_events 3 di-jet evts (by jet-finder) 2.
cos(phi_gamma - phi_jet) lt -0.8 g-jet
opposites 3. R_3x3cluster gt 0.9 3x3
cluster/total jet energy. 4. R_EMjet lt 0.9
neutral E fraction cut on away jet 5. N_ch0
no chrg tracks assoc w/ ? candidate 6. N_bTow
0 no barrel towers assoc. w/? candidate 7.
N_(5-strip clusler)u gt 3 min SMD strips
u-plane 8. N_(5-strip cluster)v gt 3 min
SMD strips v-plane 9. gamma-algo fail failed
tower SMD uv match, etc. 10. TowSMD match
tower SMD uv match bad, etc.

• Cuts effectively select
• jets opposite in phi
• gamma large neutral fraction, recoil jet
  lower neutral (e.g., with charged particles)
• select gammas in Endcap jets in Barrel
  region
• other detector match/response details

• early ? candidate response in the various Endcap
  detector layers

• subsequent investigations of influence of
  converting materials, assoc bkgnds,etc. suggest
  analysis vs. preshower conditions important !

cut effects
6
7
MC Simulations vs 2006 pp Data

• MC vs. data and preshower condition w/ ? in
  Endcap, jet in Barrel EMC
• di-jet analysis conditions with isolation (3x3
  tower patch)/(r0.7) gt 0.90
• data black MC ?-jetred MC QCD bkgndgreen

7 GeV
highly selected/most pure
most bkgnd counts/issues/etc

• similar but with isolation (3x3 tower
  patch)/(r0.7) gt 0.98

• Overall good agreement of data and MC similarly
  for pre, post specta, etc.

7
8
?/p0 Discrimination in Endcap SMD

• Maximum Sided Residual
• Look at transverse shower profile in Shower
  Maximum Det. (SMD)
• ? and e trans profile gt expect single peak
  (response composed of narrowwide Gaussians w/
  common centroid in each SMD (u.v) plane)
• p0??? expect double peak structure main peak
  and peaklet (e.g., as for an asymmetric p0 decay)
• Fit main peak compute residual(data fit) on
  each side of main peak gt pick maximum residual
  (p0s should have more residual than ?s)

8
9
Experimental Challenges Shower Max. Det. Response
Do we understand SMD response shape?

• find simple MC width too narrow
• separately, know from p0 finding algos, that MC
  doesnt reproduce strip fluctuations (extra
  spikey behavoir) that appear to drive low inv
  mass bkgnd
• further study reveals strong dependence on presh
  conditions (material effects), and other details!
  

How to make MC more realistic

• photon data_1

• Compile library of shower shapes from data (no
  test beam so, data in situ)
• In MC, replace all ? shower shapes (25
  stripscentral /- 12 strips) with appropriate
  shapes from library after proper energy scaling,
  translation in SMD plane and superposition on
  underlying event ? Data-driven MC
• Consistency check data-driven MC in better
  agreement with data!

9
10
Data Driven Shower Max. Det. Response Library
Separated photons from etas (????)

• use standard p0 finder with L2-gamma trigger
• try to make event selection w/o biasing shape
• turn off split, also lower floors, etc.
• but require minimum 20-strip peak separation
• soft peak isolation 70 energy in central 5
  strips

S/B 31 in range
0.45ltm??lt0.65 GeV
Example of 3-Gauss fit of DD shapes

• library shapes/replacement initially binned by
• preshower response (pre1, pre2)
• photon energy
• at present use average shape over
• SMD plane (U and V)
• Sector configuration (plane ordering)
• other effects wrt detector ?, F, etc.

10
11
Status of Isolated Photons in the Endcap EMC
MC ? -jet evts
MC QCD bkgd
2006 pp data
Purity of direct photons in data sample depends
strongly on pre-shower response.
pre10 pre20

• present filtered QCD bkgnd reject vs. ?-jet
  eff for diff preshower conds

pre10 pre2gt0
response data_peak uv
0ltpre1lt 4 MeV
4ltpre1lt 10 MeV

• looks promising BTW curves ordered reflecting
  inherent purity of sample

11
residual max( data_tail fit_tail)uv
12
Summary and Outlook
Summary

• Lots of good progress! Positive steps include
• Most essential features / dependences of 2006
  data down to PT7 GeV well reproduced by
  simulations (filtered MC sample in particular
  helped clarify)
• Significant investment of time and effort to
  generate new data-driven MC samples ? good
  reproduction of SMD response essential for all
  photon / meson /hadron discrimination
• Machinery in place (? and jet trees) to allow
  more detailed analysis, and including overall
  detector response, etc. ? eventually to fold
  into a more sophisticated algorithm optimization
  (e.g., Linear Discriminate Analysis)
• but still more to be done (re direct photon
  purity and efficiency vs. pT)
• Optimization of isolation cuts (and vs. what
  theorists calculate). Charged particle veto
  (added isolation) highly desirable, but not
  easily implemented over much of Endcap.
• Sided-residual technique is powerful, but
  requires judicious choices of fitting function,
  fit range, range ( of strips) used for residual,
  boundary between signal and bkgd, etc. more
  tweaking needed here (also expanded shape
  library)
• Engage full detector response in advanced
  analysis!
• Anxious to look at run 8 data w/ reduced
  material near IR!

12
13
Backup Slides
13