1. MEG 実験 液体キセノンカロリメータ におけるエネルギー分解能の追究 東大素粒子センター 金子大輔 他 MEG コラボレーション

2. Contents ・ MEG γ-ray detector ・ What limits E-resolution ? ・ Studied items ・ PMT Gain Stability ・ Photo-cathode effective size ・ Reflection on photo-cathode ・ Summary 2012/9/12 日本物理学会秋季大会 @ 京都産業大学 1

3. Liquid Xenon Detector ・ 846 2” PMTs immersed in 165K Liquid xenon ・ Reconstruct incident γ-ray from collected VUV scintillation photons. Performance (for signal γ) Efficiency62.8 % Energy resolution (w>2)1.7 % Position resolution (uv, w)5, 6 mm Time resolution67 ps Energy resolution is worse than Monte- Carlo simulation. at run2011, preliminary γ e+e+ 2012/9/12 日本物理学会秋季大会 @ 京都産業大学 2 u v w

4. 002 001 How to Get Energy 000 846 PMTs charge … software filtering integration waveform digitizer # of p.e. u, v, w p.e. 000 p.e. 001 … p.e. 002 Σ of p.e. weighted sum by PMT area energy solid-angle correction non- uniformity correction Gain, QE depth waveform face-factor correction positionfactor adjust absolute scale 2012/9/12 日本物理学会秋季大会 @ 京都産業大学 3 Calibration runs

5. What is problem ? Total collected photo-electron number is statistically enough. 50000 photo-electron → 0.45% in σ Actual resolution is 1.0% for Monte-Carlo, 1.7% for data. Result of large prototype test : 1.2% There must be un-understood event-by-event fluctuations ! σ UP MC DATA 2011 Energy resolution mapping 2012/9/12 日本物理学会秋季大会 @ 京都産業大学 4 response function fit function data

6. Studied items ・ Systematic error of QE estimation ・ Gain non-uniformity in a PMT A : PMT Gain stability ○ gain instability effect to resolution. B : Case of smaller PMT active area ○ if sensitive area were smaller, photon collection would fluctuate more C : Reflection at PMT photo-cathode ○ Dependence of light collection efficiency on the relative position between PMT & conversion point could be enhanced already known not to affect so much (reported in 2012 spring JPS meeting) 2012/9/12 日本物理学会秋季大会 @ 京都産業大学 5

7. A: Stability of PMT gain この間 4 分 PMT charge Charge from one PMT in LED event fluctuate about 3% in run. That time, I checked only about 10 min of MEG run and only some PMTs. I checked more precisely. For all PMT Evaluate statistical fluctuation. In spring JPS meeting… We are monitoring PMT gain in MEG physics run, by flashing PMT every a few second. 2012/9/12 日本物理学会秋季大会 @ 京都産業大学 6

8. Calculation of number of photoelectron Calculate mean & σ of N p.e. by fitting with Gaussian. Example PMT 000 Np.e. Average Np.e. for all PMT 0 1000 2000 Average for all PMTs: 757.2 2012/9/12 日本物理学会秋季大会 @ 京都産業大学 7 Subtract normal statistical fluctuation assuming Np.e. is distributed in poisson distribution

9. Result : A Average gain fluctuation is 0.78% in MEG run Statistical error is about 0.2-3 % → In other data 0.70% in LED run (beam off), 0.99% in CEX run (pion beam). remainder gain fluctuation σ add This fluctuation hardly worsen total energy resolution, because gain fluctuation is random for each PMT in a event. 2012/9/12 日本物理学会秋季大会 @ 京都産業大学 8

10. B: How PMT area affect ? If PMT’s cathode were smaller than designed size, this fluctuation would become larger. for example from ・ cathode deterioration ・ effect of B field ・ etc. It is already known that difference of relative position to PMT causes additional fluctuation in photon collection. 2012/9/12 日本物理学会秋季大会 @ 京都産業大学 9

11. Generated MC d < (PMTsize) ? d hit Event generation : signal γ (52.8MeV) u [cm] v [cm] used area Range : center 18PMT not to see non-uniformity ↓ Radius[cm]Area (ratio) Case 02.25100% Case 12.1591% Case 22.0079% Data2011 charge-exchange calibration (55MeV γ) 2012/9/12 日本物理学会秋季大会 @ 京都産業大学 10 I studied the simple case PMT cathode is concentrically smaller.

12. Result : B Blue : normal Green : 21.5mm Red : 20.0mm Black : data Resolution of real data is worse than MC, in all depth region. (0.6 ~ 0.8 % in σ up ) It is hard to explain discrepancy between MC & data with PMT’s smaller cathode size, because no visible effect is seen. 2012/9/12 日本物理学会秋季大会 @ 京都産業大学 11 depth [cm] E resolution : σ up [%]

13. C: Reflection on PMT cathode surface Current MEG Monte-Carlo simulation only consider reflection on quartz window. ↓ introduce reflection on cathode surface HIT 2012/9/12 日本物理学会秋季大会 @ 京都産業大学 12 solid:cathode refl.black: detect dashed:no cathode refl.red: reflect green: absorb liquid xenon n x 1.621 (@178nm) quartz window n q 1.610, L abs 21mm photo-cathode (sensitive detector) Rn c 1.920, In c 1.690 2.5

14. Result : C Energy resolution become worse in all depth by 0.2 ~ 0.5%. This can’t explain all discrepancy between MC & data, but is not ignorable part if this is true. Photon distribution got little closer to data. Parameters were not good ? ↓ 2012/9/12 日本物理学会秋季大会 @ 京都産業大学 13 Blue : w/o reflection Green : + reflection Red : MEG data

15. Summary I studied energy resolution of MEG γ-ray detector. Following results were obtained, ・ PMT gain is stable enough in MEG data taking. ・ PMT cathode size does not affect resolution. ・ Reflection on cathode surface can cause a part of discrepancy between MC & data. Prospects 2012/9/12 日本物理学会秋季大会 @ 京都産業大学 14 ・ Understand mechanism how reflection affects resolution. ・ Optimize optical parameters of PMT material. ・ Seek another cause of E-resolution discrepancy. ・ Improve current energy analysis method. ・ Use for upgraded detector.

16. おわり 2012/9/12 日本物理学会秋季大会 @ 京都産業大学 15

17. オマケ 2012/9/12 日本物理学会秋季大会 @ 京都産業大学 16

18. Calculate mean & σ of Np.e. by fitting histogram. Analyzed data Black: MEG physics data Red: Charge exchange calibration run Blue: LED gain calibration run Example PMT 000 Np.e. Average photo electrons MEG: 757.2 CEX: 749.1 LED: 759.5 2012/9/12 日本物理学会秋季大会 @ 京都産業大学 17

19. Fluctuation by Position Left : LED nphe mean Right Top : RMS (relative) Right Bottom : Subtracted Statistic 2012/9/12 日本物理学会秋季大会 @ 京都産業大学 18

20. B high low or 0? ※ this is imaginary picture B dependence of total gain come from position dependence of gain (collection efficiency) ? 2012/9/12 日本物理学会秋季大会 @ 京都産業大学 19

21. Photon collection (project to U direction) select event here Notice Data have lower peak & longer tail. Cathode reflection is hidden due to small statistics? From where can I cut summation for energy calculation? 2012/9/12 日本物理学会秋季大会 @ 京都産業大学 20

22. Introduction of photo-cathode reflection Black: normal MC Red : cathode reflection Average Nsum2 normal : 3.365e5 reflection : 2.387e5 No obvious difference except nearest 3 rows. Is this affect to energy resolution? ↑checking… 44.51% 3.28%0.37% -0.004% 2012/9/12 日本物理学会秋季大会 @ 京都産業大学 21

23. 2012/9/12 日本物理学会秋季大会 @ 京都産業大学 22 ratio of photo electron Blue : normal MC Green : data Ratio of PMTs near to conversion point is smaller in data. Difference in photon distribution between data & MC can be explained from this? → PMT row