1. Optimization of Detectors for Time of Flight PET Marek Moszyński, Tomasz Szczęśniak, Soltan Institute for Nuclear Studies, 05-400 Otwock-Świerk, Poland

2. 2 Time-of-Flight PET 1980-82: the first TOF PET – LETI, Grenoble, France -rediscovery of CsF -discovery of fast component of BaF 2 M. M. Ter Pogossian - Saint Louise, USA A higher detection efficiency of BGO has blocked further developments of TOF PET based on BaF 2.

3. 3 Scintillation detector number of photoelectrons proportional to the light output of crystals and Quantum Efficiency of photoedetectors Rise and decay times of light pulse the transit time jitter of photodetector rise time of the anode pulse The fast timing depends on three major parameters of a scintillation detector:

4. 4 PMTs for fast timing in TOF PET The superior time jitter of R5320 is compensated by a high QE of XP20D0 and the screening grid at the anode For LSO crystals a high QE is more important that a very low time jitter. Time spectra measured with 10  10  5 mm 3 LSO crystal coupled to Photonis XP20D0 and Hamamatsu R5320 PMTs, as measured for the 511 keV full energy peak in relation to the small BaF 2 crystal.

5. 5 Dependence of timing on Nphe The linear dependence of the time resolution on reversed square root of the number of photoelectrons shows the importance of high quantum efficiency of the photocathode Measured time resolution versus number of photoelectrons (Nphe) collected in the PMT Measurements done for the 10x10x5 mm 3 LSO and 4x4x20 mm 3 LSO crystals

6. 6 Time jitter dependence Time resolution is normalized in the following way: where:  t - measured time resolution N - number of photoelectrons r - gain dispersion of the electron multiplier Linear growth, but with the offset resulting from a slow decay of the LSO light pulse

7. 7 Rise time dependence Only in the case of photomultipliers optimized for timing measurements like XP20D0 the lowest threshold could be achieved leading to the best time resolution

8. 8 Decay time constants The plot represents the general influence of the decay time on the time resolution The linear dependence can be easily observed giving the best time resolution for the fastest crystals However, a group of red squares reflects timing tests done with NE213, LaBr 3 and LaCl 3 scintillators characterized by the finite rise time of the light pulses.

9. 9 Hyman theory  - decay time  - width of SPE response  ’ - time jitter H predicted = 0.095, H measured = 0.09 Time resolution prediction by the Hyman theory (H   t) LSO at XP20D0:  = 46 ns  = 1 ns  ’ = 0.25 ns L.G. Hyman, “Time resolution of photomultiplier systems”, Rev. Sci. Instr., vol. 36, No. 3, pp. 193-196, Feb. 1965.  H =  t × N 1/2 /  × r

10. 10 Common light readout Following A. Kuhn et al. „Performance Assessment of Pixelated LaBr3 Detector Modules for Time- of-Flight PET”, IEEE Transactions on Nuclear Science, Vol. 53, No. 3, Jun 2006 A further improvement of the time resolution, due to the common light readout by the cluster of PMTs in the block detector, was proposed by Kuhn et al.

11. 11 Measurements with LSO

12. 12 Conclusions Ideal fast PMT demands the lowest possible time jitter, high quantum efficiency and the best anode pulse shape quality In the case of LSO the dependence on the number of photoelectrons is stronger than the one connected with the time jitter The anode pulse improvement (screening grid) is the third important parameter for the final timing properties of the detector A better light collection with a system realizing simultaneous redout of two (or more) PMTs leads to the improvement of time resolution The value of about 100 ps at FWHM time resolution for a LSO based detector (~2 cm for two detectors) seems to be a limit for near future top class PMTs

13. 13 Idea of monolithic crystal detector LYSO 20x20x20mm 3

14. 14 Results with LSO 10x10x5

15. 15 Results with LYSO 20x20x20

16. 16 Tested MPPC (SiPM) ManufacturerHamamatsu TypeS10362-33-050C SN22 Active area3×3 mm 2 Number of pixels3 600 Pixel size 50x50  m 2 Fill factor61.5 Gain7.5 x 10 5 Spectral resp.range 320-900 nm (max emission at 440nm) Q.E.70% at 400 nm Operating voltage70 ± 10 V Dark count6Mcps Capacitance320 pF Dead time of pixel15 ns MPPC (Multi-Pixel Photon Counter) ManufacturerZecotek TypeLFS-3 Size [mm³]3×3×3 Decay time [ns]39.7 Emission peak [nm] 420 Light output [photons/MeV] 30 000 Scintillator

17. 17 Time resolution Optimum is achieved for very short CFD shaping delay equal to only 2 ns 1 ns delay corresponds to Leading Edge mode of the discriminator The best results are identical in both configurations but with more „flat” values for 2 ns

18. 18 Time resolution Timing varies with HV due to: increase of the Nphe, higher probability of Geiger breakdown Large time resolution improvement (45%) is observed in the range of 1.2V The mean time resolution of a single MPPC+LFS detector: 200 ± 13 ps. Value is slightly worse than the best results, below 170 ps, obtained with fast photomultiplier XP20D0 and LSO [T. Szczęśniak, et al, IEEE TNS, vol. 56, p.173]

19. 19 MPPC response signal Each micro-cell of the SiPM matrix has its own capacitance and total capacitance of the detector together with input resistance of readout electronics leads to significant time constant that defines the rise time of the output signal. To reduce an influence of the 320pF MPPC capacitance, the current preamplifier (constructed at SINS) with very low input resistance was used in the timing study.

20. 20 Time jitter Despite of perfectly separated single photoelectron peaks the time jitter is determined by the mean number of photoelectrons Another method of photoelectron number selection is needed instead of Single Channel Analyzer gates

21. 21 Time jitter Average pulse 1 photoelectron Average pulse 3 photoelectrons

22. 22 Time jitter After subtraction of noise contribution (blue) the time jitter dependence (green) is perfectly proportional to the statistics of the photoelectron number MPPC 059C SN22 time jitter is equal to 900 ps Comparable to that reported in NIM by Fermilab

23. 23 Number of phe and ENF Strong Nphe dependence on HV High nonlinearity even for low photon numbers About 16 000 phe/MeV for 69.2 V used during timing measurements, on XP2020Q: 6000 phe/MeV The statistical contribution  st to the pulse height resolution (PHR)  st =2.355 x (ENF / NPHE) 1/2 Measurements with LED allowed calculation of the maximal ENF for a given voltage

24. 24 Comparison to photomultipliers Time resolution is normalized in the following way: where:  t - measured time resolution N - number of photoelectrons r - square root of ENF  - decay time constant Normalized time resolution shows similar speed of the tested MPPC and PMTs. Adventages: high quantum efficiency, high photoelectron number Disadventages: slow rise time (large capacitance of 320 pF), high time jitter, high ENF

25. 25 Conclusions The time jitter of a 3x3mm 2 MPPC is 2 times higher comparing to fast PMTs and at the level of 900ps for 2.3V overvoltage. Jitter is increasing with lowering of bias voltage. High QE of MPPC leads to high photoelectron number, increasing with HV and 2.5 higher comparing to PMT. Nphe is affected by excess noise factor from 1.4 up to 4, dependent on HV. Despite poor rise time, poor time jitter and high ENF the time resolution of 200ps for a single detector is only slightly worse comparing to classic PMTs due to high Nphe. Optimal timing properties correspond to highly nonlinear range of HV