p0 ALL analysis in PHENIX

1
p0 ALL analysis in PHENIX
K. Tanida (RIKEN/RBRC) RSC meeting 11/18/03

• Outline
• Relative luminosity- estimation of systematic
  error- ALL in ZDC/BBC
• LocalPol
• Counting number of p0s
• Result

2
ALL
same helicity ? opposite helicity
(P) Polarization (L) Relative Luminosity (N)
Number of pi0s
3
How to get relative luminosity?

• Principle RN/N- -- from any counter
  (trigger)
• BBCLL1 trigger - coincidence of two Beam-Beam
  Counters- small background ( 10-4)- high
  statistics (detects 1/2 of all collisions)-
  ALLBBC expected to be small, need confirmation

BBC-South
BBC-North
?? 3.0 3.9 ?f 2p
144.35 cm
4
Method of accuracy estimation

• Cross-check by ZDCLL1 trigger
  (Zero Degree Calorimeter) - low background
  ( 10-4) - statistics moderate ( 3 of BBCLL1)
• N(ZDCLL1)/N(BBCLL1) must be constant
• Crossing-by-crossing, fill-by-fill ratio
  calculation Blue - - - -....
  Yellow - - - -....? We have all spin
  combinations in each fill
• GL1p scaler counts both BBCLL1 and ZDCLL1for
  each crossing
• Required accuracy constant to 10-3 level for
  each fill.

5
crossing-by-crossing ratio
c2/dof 346/45
Significant deviation from constant. Why?
6
Possible explanation detector acceptance
run 87693

• Extreme cases - Flat acceptance ? N
  area - Narrow acceptance ? N height

crossing 74
counts
crossing 90
BBC Z vertex (cm)

• ZDCLL1 Z-vertex resolution 20 cm (wider
  acceptance)
• BBCLL1 5 cm
  (narrower) (both have Z-vertex cut at 30 cm)

7
Correlation with Z-vertex width
c2/dof 55.1/45
cm
Z-vertex width cm
We understand the reason now ? correction
8
After correction
c2/dof 54.8/45
Relative luminosity accuracy lt 0.09 (stat.
limited)
9
Alternative correction

• Actual vertex width is not yet available in most
  runs(will come soon)
• Alternative width estimation? use ZDCout/ZDCin
  ratio - ZDCin ZDCLL1 counts with Z-vertex
  cut of 30 cm - ZDCout ZDCLL1 outside of
  ZDCin. (both are counted
  crossing-by-crossing)? wider vertex distribution
  ? larger ratio

10
Actual accuracy estimation-- Bunch fitting

• Crossing-by-crossing analysis, taking care of
  possibleeffect of ALL in BBC and ZDC
• For i-th crossing, calculate r(i)
  NZDCLL1(i)/NBBCLL1(i)corrected for vertex width
• Fit r(i) by r(i) C1ALLBBC/ZDCPB(i)PY(i)f
  it parameter C, ALLBBC/ZDC
• You will get C, ALLBBC/ZDC, their errors, and
  c2- accuracy is obtained as dALLBBC/ZDC (x
  PBPY)- c2 gives a good check for systematic
  errors.

11
c2 vs stat. error before correction
1 point/fill
w/o vertex width correction
correlating ? ssyst 0.002
c2/dof 1(ssyst/sstat)2 ssyst 0.002
c2/dof
statistical error (deLL, sstat)
12
c2 vs stat. error after correction
fill 3735
correlation is weak, but still some systematic
effect remains
bad fill
c2/dof 1(ssyst/sstat)2 ssyst 0.001
c2/dof
statistical error (dALL, sstat)
13
Result BBC/ZDC ALL

• Systematic error seen ? c2 correction required
• If c2/dof is larger than 1, then enlarge
  statistical error by sqrt(c2/dof) conservative
  correction
• Averaged result ALLBBC/ZDC (1.81.8) x 10-3
  (without vertex width correction, dALL is 2
  larger)
• This corresponds to relative luminosity accuracy
  of dR 2.5 x 10-4? Achieved the goal
• ZDC-BBC ALL is 0 consistent? we can use both for
  reference of ALL 0

14
Some notes

• We reached dALLBBC/ZDC1.8 x 10-3- this is
  dominated by statistics of ZDC, so its rather
  conservative estimation- actual relative
  luminosity is given by BBC
• In p0 analysis, vertex position dependence of
  acceptance is the same as BBCLL1 because- BBC
  hits are required in the analysis- PHENIX
  central arm acceptance is almost flat for p0
  within the region we are using (30 cm)

15
Remaining tasks for rel. luminosity

• Apply offline Z-vertex cut - better resolution
  5 cm ? 2 cm (effect is found to be small)
• Full accuracy estimation using real vertex
  distribution. - correction for other than vertex
  width?
• Study effect of multiple collisions - negligible
  in Run3, significant for higher luminosity.
• Z-vertex dependence of acceptance
  (efficiency)for BBCLL1, p0 and other reaction
  channels. - reaction channel specific relative
  luminosity. - BBCLL1 itself is almost perfect
  for p0 case, but may be not for other
  channels.

16
Vertex distribution (BBCLL1)
reject
reject
counts
accept
offline Zvertex cm
changes rel. lumi. by lt 0.04 per fill
(negligible)
17
PHENIX Local Polarimeter

• Forward neutron transverse asymmetry (AN)
  measurements
• SMD (position) ZDC (energy)

f distribution
SMD
Vertical ? f p/2 Radial ? f 0 Longitudinal
? no asymmetry
ZDC
18
Spin Longitudinal Component
ST is measured with PHENIX Local Polarimeter
Up-Down asymmetry
Left-Right asymmetry
19
p0 counting
Data collected with high pT photon trigger Based
on EMCal Threshold 1.4 GeV/c Rejection factor
110 Analyzed data sample 42.7M events (0.215
pb-1) sqrt(ltPbPygt)26 Minimum Bias data To
obtain unbiased ?0 cross section at low pT For
high pT photon trigger efficiency study
20
p0 reconstruction for ALL
2-3 GeV/c Bckgr17
1-2 GeV/c Bckgr45
Results obtained for four pt bins from 1 to 5
GeV/c Pi0 peak width varies from 12 to 9.5
MeV/c2 from lowest to highest pt bins Background
contribution under pi0 peak for ?25 MeV/c2 mass
cut varies from 45 to 5 from lowest to highest
pt bins
3-4 GeV/c Bckgr7
4-5 GeV/c Bckgr5
21
p0 counting for ALL
N?0 ?25 MeV/c2 around ?0 peak (and also ?15 and
?35 MeV/c2 for cross checks) Nbck1 Two 50
MeV/c2 wide areas adjacent to ?0 peak Nbck2
250 MeV/c2 wide area between ?0 and ? peaks
N?0 and Nbck accumulated statistics
pt GeV/c N?0 15 MeV/c2 N?0 25 MeV/c2 N?0 35 MeV/c2 Nbck1 Nbck2
1-2 1278k 1777k 2129k 1470k 3478k
2-3 874k 1059k 1146k 335k 989k
3-4 176k 201k 208k 27k 83k
4-5 34k 38k 39k 3.9k 12k
22
N?0 statistical error estimation
dk/dNev
dk/dNev
Nbck2
N?0

• In multi-?0 events
• Comparison to naïve approach

Fluctuation enhancement (squared) due to
multi-particle events
pt GeV/c N?0 15 MeV/c2 N?0 25 MeV/c2 N?0 35 MeV/c2 Nbck1 Nbck2
1-2 1.134 1.220 1.302 1.360 1.708
2-3 1.047 1.081 1.114 1.178 1.406
3-4 1.015 1.026 1.037 1.101 1.222
4-5 1.008 1.015 1.021 1.059 1.162
23
ALL measurements
same helicity ? opposite helicity

• Procedure
• Collect N and L for and ? configurations (sum
  over all crossings) and calculate ALL for each
  fill
• Average ALL over fills use ?2/NDF to control fit
  quality use bunch shuffling to check syst.
  errors

24
ALL Calculations
ALL averaged over fills
1-2 GeV/c
2-3 GeV/c
1-2 GeV/c ALL -2.81.2 ?2/ndf 64/48 2-3
GeV/c ALL -2.21.5 ?2/ndf 34/48 3-4
GeV/c ALL -0.23.3 ?2/ndf 49/48 4-5
GeV/c ALL -2.37.4 ?2/ndf 39/48
4-5 GeV/c
3-4 GeV/c
25
AL check for yellow beam
pt GeV/c 15 MeV/c2 25 MeV/c2 35 MeV/c2
1-2 0.001?0.004 0.000?0.003 0.000?0.003 0.002?0.004 0.000?0.003
2-3 0.001?0.004 0.000?0.004 0.000?0.004 0.002?0.007 0.002?0.005
3-4 0.007?0.009 0.011?0.009 0.008?0.009 -0.033?0.025 -0.010?0.015
4-5 -0.001?0.021 0.004?0.020 0.008?0.020 0.020?0.064 0.050?0.039
All are zeros within 1.5?
26
AL check for blue beam
pt GeV/c 15 MeV/c2 25 MeV/c2 35 MeV/c2
1-2 -0.001?0.004 0.001?0.003 0.000?0.003 -0.002?0.004 -0.000?0.003
2-3 0.001?0.004 0.000?0.004 0.002?0.004 0.009?0.007 0.000?0.005
3-4 0.004?0.009 0.006?0.009 0.006?0.008 -0.004?0.024 -0.036?0.015
4-5 -0.024?0.021 -0.016?0.020 -0.016?0.019 -0.011?0.062 0.013?0.038
All are zeros within 1.5?, except
27
?0 ALL result
pT GeV/c (rbck) (Background subtracted)
1-2 -0.028?0.012 (45) -0.006?0.014 -0.046?0.025
2-3 -0.022?0.015 (17) -0.035?0.027 -0.019?0.019
3-4 -0.002?0.033 (7) 0.094?0.092 -0.009?0.036
4-5 -0.023?0.074 (5) 0.38?0.24 -0.045?0.079

• Polarization scaling error ?P 30 is not
  included
• Enters the ALL quadratically
• Analyzing power AN(100 GeV) AN(22GeV) is
  assumed
• ?P30 combined stat. and syst. error for
  AN(22GeV) (AGS E950)
• pT smearing correction is not included

28
Summary

• Relative luminosity systematics is mostly due
  to crossing by crossing variation of Z-vertex
  width and acceptance difference ?
  correction applied
• Relative luminosity accuracy dALL 1.8 x
  10-3 achieved
• Relative luminosity is reaction channel
  dependent - For p0 analysis, BBCLL1 is good
  enough.
• Preliminary result for p0 ALL obtained
• Various null tests all OK statistical and
  systematic errors are well studied.