1. LKr inefficiency measurement
Giuseppe Ruggiero
CERN
Presentation to the Referees
14/11/2005

2. Overview Goal: Requested LKr inefficiency:
Demonstrate with DATA (NA48/2, 2006?) that the LKr matches our request in terms of inefficiency (~10-5 level measurement)
Goal:
Define a method for measuring the LKr inefficiency on DATA during real data taking with an uncertainity < 10-5
E range
Inefficiency
< 1 GeV 1
(1,3) GeV
10-4
(3,5) GeV
Linear between
> 5 GeV
10-5
Requested LKr inefficiency:

3. Simulation
LKr + passive material in front of the calorimeter simulated using GEANT4 (not the standard simulation of NA48, it is used as a specific tests on the effects due to the material)
g from p+p0
MeV
Inefficiency (energy deposited by g < 400 MeV) = (1 ± 0.8) × 10-6
Inefficiency due to photo – nuclear interaction of the photon with the stesalite support 5 cm thick.

4. LKr inefficiency measurement: method1
Select p+p0
Select at least one cluster not matched to the track in LKr
Assume this cluster originated from the lowest energy g
Look for clusters around the region where the other g should be 2
Select p+p0Dalitz
Reconstruct the three tracks with the spectrometer
Look for clusters around the region where the other g should be 3
Select p+p0p0Dalitz
Reconstruct the three tracks and the p0
Look for clusters around the region where the other g should be

5. p+p0 method DATA Simulation (used for cross check purposes)
NA48/2 Test 75 GeV (August 2004)
K+, PK=75 GeV/c, DPK/PK=2%, parallel beam (divergence < 50 mrad)
1/8 nominal beam intensity (low accidental activity)
Simple trigger for p+p0: no m, 1 track in the Hodoscope
Simulation (used for cross check purposes)
Standard simulation of NA48 based on GEANT3 and modified for the test run conditions
Selection (main criteria, most part of the cuts not listed)
PK=75 GeV/c along Z assumed (good approximation for the test run)
Only 1 track reconstructed in the spectrometer and inside of the detector acceptances.
No muons in the Muon Veto
(ELKr / P measured in spectrometer)>0 (further muon suppression)
(ELKr / P measured in spectrometer)<0.8 (against e+/e-)

6. p+p0 selection Statistics: p+p0p0
About 5×105 events selected within the p+p0 peak
Resolution vs momentum
p+p0p0

7. Photon selection Number of clusters in LKr < 4
Cluster matched to the track if the cluster closest to the track is at < 20 cm from the track.
No in time hits in the large photon vetoes (AKL)
g candidate: minimum energy cluster
Energy of the cluster enough to have a good resolution (>3 GeV)
Cluster in time with the track (within 1.5 ns)
Cluster with a shape consistent with the shape of an electromagnetic cluster
g expected :
Inside of the LKr acceptance
Isolated from any other cluster (> 20 cm)
Energy > energy of the g candidate
expected position
Cluster of the
expected g
g candidate
Pion cluster
Track

8. Limits of the method 20 < Rlkr < 90 cm Eg > 10 GeV MC MC
Beam hole effect
External edges effect
Photon cluster looked for in
a 55 cm radius circle around the expected position
20 < Rlkr < 90 cm
Eg > 10 GeV

9. Results MC
DATA
h(20 < RLKr < 90 cm) = (5 ± 2) × 10-5
h(20 < RLKr < 90 cm) = (1.08 ± 0.05) × 10-3
No dependence on RLKr in
[20,90] cm
Inefficiency due to the g conversions in the chambers before the magnet: the e± track are not reconstructed in the spectrometer and the e- and/or e+ cluster is detected in the LKr.
Inefficiency vs energy g
No significant dependence

10. Results
Inefficient events
DATA MC
No significant dependence of the inefficient events on the LKr surface

11. Systematic effects: Background
Km2-Km3
Km2 rejected by mVeto, photon requirement and kinematics, Km3 by mVeto.
Fraction of Km3 events with respect to p+p0 if only the Muon Veto is used: ~10-5 (from MC and assuming ~10-3 inefficiency for Muon Veto).
Further suppression by requiring E/P> 0 (1 order of magnitude, at least).
Can be reduced at a negligible level in P-326 (RICH+MAMUD)
Check on data: inefficiency measured by removing the cut on E/P> Result: | h(no m cut) – h| = (0.4 ± 1.3)×10-5
Ke3
Fraction of e+ with E/P<0.8 ~5×10-4 (number taken by existing NA48 measurement)
Fraction of events with respect to p+p0: ~1.5×10-5 (MC)
Warning: this background depends on the LKr inefficiency (e ID with LKr)
Can be reduced at a negligible level in P-326 (RICH)

12. Systematic effects: Background
p+p0D
Fraction of events with respect to p+p0: ~4×10-5 (MC)
Most of these events are tagged as “efficient” because the e+ and/or e- are detected in LKr close to the expected g: effect on inefficiency <10-5
Can be reduced of an order of magnitude using cuts on the hit multiplicity in the chambers and/or Hodoscope
p+p0g
Fraction of events with respect to p+p0: ~7×10-3 (MC)
Most of the events with soft radiative photon: kinematics close to the p+p0
Most of these events have the radiative g around the expected g position
Effect on inefficiency <5×10-6
Accidentals (to be studied in more details):
Inefficiency measured on data opening the track – cluster time cut
Inefficiency increases of (2±1)×10-5

13. Other systematic effects
Resolution effects (to be studied in more details):
Photon outside the LKr while predicted inside.
Depends on the cut defining the region where the g is looked for.
Reduced with an hermetic veto coverage of the region outside the LKr:
Will be provided in P-326
Inefficient large angle veto and no small angle coverage in NA48.
Effect ~10-5 (MC) for Eg>10 GeV, increases steeply at lower Eg.
Pion clusterization in LKr:
A relic of pion cluster far from the track impact point on LKr can be identified as the minimum energy photon candidate in events where at least one of the two photons is outside the LKr acceptance
Probability of mismatching: ~2×10-5 (measured with DATA in K+++- decays with 2p at MIP and full MC)
Can be reduced at ~10-6 with more tight cut on track-cluster distance (40cm): it costs in statistics (problem for the 2004 test run).

14. Systematic effects: g conversions
Photon conversions in Kevlar window, DCH1 and 2
Fraction of events with at least one conversion before the magnet: ~5×10-3 (MC)
Most of the events tagged as “efficient” because the e± cluster is found around the expected g position.
Residual effect on inefficiency: ~5×10-5 (MC)
Cuts on DCH and Hodoscope hit multiplicity can help
Can be further reduced in P-326 using the RICH e-
Kevlar Window
DCH1
DCH2
DCH3
DCH4
LKr e± g e+
Beam pipe p g
Hodoscope
Magnet

15. Systematics: summary Sources Effect Background 2×10-5 Accidental
Resolution
10-5
Pion showering
g conversions
5×10-5
Total
~6×10-5
Can be reduced of ×2 in NA48
Negligible in P-326
Need more studies
More detailed study in NA48
Negligible in P-326
Can be reduced at
negligible level also in NA48
Can be reduced in NA48.
Probably negligible in P-326.
More studies needed
The observed inefficiency in DATA (10-3) cannot be explained by the uncertainties in the method of measurement
The measured inefficiency cannot be a reconstruction effect because in MC and in DATA the same reconstruction algorithm is used
Can be a READOUT inefficiency

16. Study of the READOUT inefficiency
Use of the SPY:
Alternative readout used for monitoring the Neutral Trigger chain
Calorimeter read by grouping the cells in a 64×64 matrix
Method:
In events without a cluster in the “expected” LKr region (“LKr inefficient events”) look for SPY_X and SPY_Y intersection in the region where the expected cluster should be.
If an intersection (SPY hit) corresponding to the expected photon is found, the LKr does not see the cluster because of a READOUT inefficiency
The expected cluster is found in SPY if at least one hit in SPY closer than 8 cm to the expected cluster position exists.
Limit of the method:
The SPY can be inefficient by itself.

17. SPY inefficiency Method: Hits in SPY Fraction
In events for which the 2 photons and the pion have the clusters reconstructed in LKr, look how many times the SPY give a hit corresponding to each of that clusters.
Hits in SPY
Fraction
No hit
2.2%
1 hit
0.5%
2 hits
14.4%
3 hits
83%
Global inefficiency of the SPY 2.2%
hSPY~6% for single g
Expected probability not to see one over the three clusters in SPY: ~7.5% per cluster.
Measured probability not to see one over the three clusters in SPY: 8±3%

18. SPY analysis: results h(LKr) < 1.7×10-4 (Eg > 10 GeV) Results:
Probability to miss in SPY the hit corresponding to the expected g: a = 8 ± 3%
N(total events) = 3×105
N(inefficient events) = 3×102
N(hit in expected g for the inefficient events) = 2.5×102
Because of the SPY inefficiency:
If we take into account the effect of the inefficiency we can extract the number of events in which the LKr was intrinsically inefficient:
N(Intrinsic LKr ineff)=N(inefficient events) – N(hit in expect. g)/(1 - a)
Statistical error and error on a added linearly. Error on a the most important one.
We have always to consider ±6×10-5 as systematic uncertainty of the method
h(LKr) < 1.7× (Eg > 10 GeV)
h(LKr) = (1.0 ± 0.5)× (Eg > 10 GeV)

19. p+p0D and p+p0p0D methods p+p0p0D
The measurement on these sample does not suffer for pion clusterization, backgrounds and accidentals
Same problems from resolution and conversions as using p+p0
Lower photon energy region can be addressed
Measurement of the inefficiency cleaner than using p+p0
p+p0p0D
Analysis at earlier stage
Analysis performed on the normal NA48/2 data 60 GeV/c
First look to data suggests the presence of the READOUT inefficiency like in p+p0
Check with SPY not done yet
Eg<10 GeV difficult to handle
s(distance expected- real g ) cm
E expected g (GeV)

20. Summary
A method to measure the LKr inefficiency on DATA using p+p0 events developed in detail:
Estimated systematic uncertainty ~6×10-5
Room for improvements with the NA48 layout (> factor 2)
Sensitivity <10-5 with the P-326 layout (accidentals to be understood)
Hint for a READOUT inefficiency at the level of ~10-3
Method to decouple the intrinsic and READOUT ineff. developed:
Use of an existing alternative READOUT (SPY)
Intrinsic LKr inefficiency 10-4
Sensitivity limited by the uncertainty of the p+p0 method and by the inefficiency of the SPY
Other methods with lower systematic uncertainty developed
Analysis at earlier stage
Confirm the ~10-3 READOUT inefficiency

21. Outlook Existing DATA: 2006 RUN:
Complete the analysis exploiting all the methods
Understand the READOUT inefficiency
Set a method to measure the inefficiency at Eg<10 GeV
2006 RUN:
Solve the READOUT inefficiency (hardware level)
Improve the SPY inefficiency
Take data to measure the LKr inefficiency with a sensitivity close to 10-5
Possible scan of the LKr inefficiency vs beam intensity