COMPTON POLARIMETRY

1
COMPTON POLARIMETRY

• Electron Differential Method
• Electron Integrated Method
• Photon Integrated Method
• (work in progess in Syracuse)

2
Electron Detector
4 planes of 48 mstrips (650 ?m wide)
Trigger 2 planes among 4
Ydet
Compton events (10 mm)
3rd dipole
HD?0 (300mm)
X
?0 (56.3 mrad)
?e
Bdl
D4102mm
3
Differential Method
Compton Rate
Compton Asym
Calibrate with Compton edge --gt Fit asymmetry
curve versus energy
4
Calibration
Y
3rd dipole
HD?0 (300mm)
X
?0 (56.3 mrad)
?e
Bdl
D4102mm
X lt--gt E conversion depends on external
parameters Ebeam, Bdl, D and Ydet.
5
Ydet
Previous discrepancy between YdetRate and
YdetAsym 200 ?m
Goal2
6
Ydet
Rate Method Fit of the Compton
cross-section instead of ratio of the two last
Strips. Explained most of the discrepancy.
Asym Method Few refinements and minor
bugs fixed
7
Ydet
2005 4He run Agreement lt 25 ?m !!
Common syst. error from Bdl?
8
0Xing Method

• Linear fit of the zero crossing
• of the Compton Asym using
• 4 strips
• Integrate from the asym
• from that point and compare to
• Anamywing power with same
• threshold.
• Absolute calibration, the only
•  input  is QED.
• Previous syst. calibration error
• is converted into stat. error of
• the linear fit !

Hydrogen run (12457)
Dream method, systematics free?!?
9
Back to real world

• Extrapolation to closest strip edge back with
  Bdl, D and Ydet
• BUT the lever arm of this extrapolation is lt 1/2
  strip (325mm) instead of the full
• Compton spectrum (10mm) --gt Syst error peanuts
• 0Xing point outside of the detector
• energy range or at the edge
• Larger error from the fit because of larger
• lever arm, but a priori stat. error only
• (red lines, strips 1-4).
• Syst. offset in the extracted 0Xing because
• of the curvature of the asym curve (green
• line, strips 4-7). Corrected using simulated
• curve with YdetRate calibration.
• Small correction. Small syst. error.
• Sensitivity to edge effects. Could be
• kept small in principle but major issue
• for the HAPPEx run.

Helium run (11743)
10
Tracking in e- Detector
D
C
B
A

• Small misalignment can cause very large
• efficiency loss in the first strips of upstream
• planes
• must be corrected before comparing the
• experimental weighted sum of the strips
• to the mean analyzing power.

Ydet

• Few reference runs are chosen to study individual
  electron tracks
• Compute the geometrical edge effects from the
  angle of the tracks
• Use 3 and 4 planes events to compute intrinsic
  detection efficiency (?)

11
?
Compton Edge
4He run (11823)
Edge effect in first 2-3 strips Then efficiency
flat 95
2 correction in Pe !
12
Ydet Cross-Check
Complementary check with different syst. errors
?Ydet 45 ?m
13
Two Independent Methods

• Differential Method
• Main error calibration, depends on Ydet,
  Bdl,D,Ebeam
• Good stat. Accuracy
• Low sensitivity to edge effect and ?

• 0Xing Method
• Sub syst. error if the crossing is well inside
  the det range
• Incertainty on external parameters converted to
  stat. Error
• Very sensitive to ? variation in the 0Xing range

--gt Powerful cross-check of Compton analysis
14
4He Results
15
4He PullPlots

• Cuts
• HWP state
• ?Pegt0
• No transverse runs
• Fit Stability

16
Fit Instabilities

• Cut FitProbgt0.001
• 50 of the helium runs
• 25 of the hydrogen runs
• Large fraction but randomly
• distributed --gt No effect on mean
• value.

17
4He Mean Values
Plane Pe_Fit Pe_0Xing
Pe_0Xing_No_Prob_Cut (250
runs) (399 runs) A
83.81 84.20 83.97
B 83.96 84.03
84.16 C 84.11 83.55
83.49 D 84.07
84.01 83.78 ltPegt 83.99
0.05 83.95 0.12 83.78 0.10 (stat.)
Very stable and perfect agreement ! Cut on Fit
Prob has not effect on the mean value
18
LH2 Results
Supposed to be the easiest because higher Ebeam
and 0Xing in strip 2-3 instead of 0-1 for helium

• 2 discrepancy
• Agreement clearly varies in time
• Best candidate is e variations

19
? vs Time
Run 12400
Run 12602
Run 12900
LH2
Run 12298
Run 11702
Run 12048
4He
20
LH2 New Pass

• Better but not convincing
• Large ? variation on a run to run basis?
• Wrong estimate of ? but then 4He results are
  really lucky?
• Differential method already good enough

21
LH2 Pull Plot
0Xing
0Xing method is more consistent with constant
poalrization because of the intrinsic larger stat
error.
Differential
22
LH2 Mean Values
Plane Pe_Fit Pe_0Xing
Pe_0Xing_No_Prob_Cut (273
runs) (390 runs) A
86.67 84.90 84.90
B 86.73 84.59
84.53 C 86.67 85.20
85.22 D 86.96
85.94 85.86 ltPegt 86.75
0.04 85.15 0.06 85.12 0.05 (stat.)
New pass with ? reference runs every 50 runs
A 86.60 84.64 B
86.70 85.32 C 86.59
85.66 D 86.90
86.37 ltPegt 86.70 0.04
85.55 0.06 (stat.)
Points to remaining problems in the determination
of ?
23
Error Budget (preliminary)
If the ? error is pinned down, who is going to
trust my 0Xing numbers?
24
Online / Offline
Offline Analysis 83.99 0.87syst 0.05stat
25
Online / Offline
Offline Analysis 86.70 0.81syst 0.04stat
26
Conclusion

• In good shape for the final publication
• Pe(4He) 83.99 0.87syst 0.05stat
• Pe(LH2) 86.70 0.81syst 0.04stat
• Still some work to establish final error bar (?,
  beam quality cuts, beam spot size,radiative
  corrections)
• Few remarks about e- det for the PRex
• Remote controlled vertical motion to bring e- det
  as close as possible to beam axis (0xing expected
  _at_ 3.5mm).
• Check angle of tracks --gt adjust Vcorrector in
  3rd dipole or e- det angle
• Smaller distance between planes to reduce
  sensitivity to track angle.