1. Evidence for direct detection of ionizing particles in SiPMs
14th Topical Seminar on Innovative Particle and Radiation Detectors: IPRD16
Evidence for direct detection of ionizing particles in SiPMs
R. Santoro1, M. Caccia1, S. Gentile2 and S. Veneziano3
1) Università degli Studi dell’Insubria, Dip. Scienza ed Alta Tecnologia - COMO
2) Università LA SAPIENZA, Dip di fisica - ROMA
3) INFN – ROMA

2. Introduction
SiPM is a well know light sensitive detector
Different teams are studying a protocol to qualify the detector according to the specific applications (i.e. V. Arosio’s Talk )
The typical measurements included in the protocols are:
Dark Count Rate, Cross Talk, After pulse, Photon Detection Efficiency, Gain, Spectral response
This talk will focus on the SiPM sensitivity to ionizing particles, an effect that would be better to consider for those applications were both light and ionizing particles are present
R. Santoro
IPRD16, October 2016, Siena (Italy)

3. Experimental setup
All the results shown in this talk are based on the following setup:
Caen desktop control and amplification unit (SP5600A)
3-stage amplification with leading edge discrimination
Caen desktop digitizer (DT5720)
250 Ms/s sampling rate, 2V dyn-Range and 12bit resolution
Hamamatsu SiPMs (different area and different pitch size)
The analysis is based on
A series of measurements used to qualify the SiPMs in terms of DCR and Xtalk
Spectra analysis based on data collected with and without 90Sr
R. Santoro
IPRD16, October 2016, Siena (Italy)

4. SiPM qualification in the dark
Dark Count and Xtalk 15°C
typical Signal
Threshold scan
SiPM (S ) 3x3 mm2 with 50μm pixel size
Vop=2.6V over breakdown
DCR = 43 KHz
Xtalk = 45 % 1.5ph-e / 0.5ph-e )
R. Santoro
IPRD16, October 2016, Siena (Italy)

5. SiPM qualification in the dark
Dark Count and Xtalk 15°C
typical Signal
Threshold scan
SiPM (S ) 3x3 mm2 with 50μm pixel size
Vop=2.6V over breakdown
DCR = 43 KHz
Xtalk = 45 % 1.5ph-e / 0.5ph-e )
If I measure the mean value and its std in each step …
Zoom
R. Santoro
IPRD16, October 2016, Siena (Italy)

6. SiPM qualification in the dark
Dark Count and Xtalk 15°C
typical Signal
Threshold scan
SiPM (S ) 3x3 mm2 with 50μm pixel size
Vop=2.6V over breakdown
DCR = 43 KHz
Xtalk = 45 % 1.5ph-e / 0.5ph-e )
… I have a different way to show the threshold scan (staircase)
R. Santoro
IPRD16, October 2016, Siena (Italy)

7. SiPM qualification in the dark: CrossTalk
If you deal with low Xtalk probability, any approximation is good, while…
… if you are looking for small effects and you have large Xtalk probability, a proper xTalk description is mandatory!
Data-sheet
Overvoltage in use corrected for ΔT = 10°C
R. Santoro
IPRD16, October 2016, Siena (Italy)

8. SiPM qualification in the dark: CrossTalk
If you deal with low Xtalk probability, any approximation is good, while…
… if you are looking for small effects and you have large Xtalk probability, a proper xTalk description is mandatory!
R. Santoro
IPRD16, October 2016, Siena (Italy)

9. SiPM qualification in the dark: CrossTalk
If you deal with low Xtalk probability, any approximation is good, while…
… if you are looking for small effects and you have large Xtalk probability, a proper xTalk description is mandatory!
Geometrical Chain
Branching Poissonian
Definitely better!
R. Santoro
IPRD16, October 2016, Siena (Italy)

10. SiPM response to a β source
Threshold scan without source compared to threshold scan with 90Sr
The 2 trends are scaled for the counting rate measured at 0.5 ph-e
An excess of counting rate at higher threshold is quite evident
R. Santoro
IPRD16, October 2016, Siena (Italy)

11. SiPM response to a β source
Threshold scan without source compared to threshold scan with 90Sr
The 2 trends are scaled for the counting rate measured at 0.5 ph-e
An excess of counting rate at higher threshold is quite evident
I’m using the Borel distribution to check if the 90Sr data set has an extract effect with respect to the DCR
This is a clear indication that the response to a β source cannot be justified with an high Xtalk probability
R. Santoro
IPRD16, October 2016, Siena (Italy)

12. Spectrum method Digitized Signal Typical Signal Typical Spectrum
When a cell in the SiPM is fired, an avalanche starts with a multiplicative factor of about
The analogue signal is proportional to the number of fired cell
Typical Signal
Digitized Signal
If integrated into a fixed gate, it is proportional to the number of fired cells
Typical Spectrum
A small number of fired cells allows to see the different peaks, but this effect disappears when the cells number increases
The peak-to-peak distance allows to measure the light intensity even when the peaks are not distinguishable anymore
R. Santoro
IPRD16, October 2016, Siena (Italy)

13. SiPM response to a β source
Spectrum without source compared to spectrum with 90Sr
Trigger set at 16mV (≈1.5 ph-e)
The two spectra are normalized in time before the bin-to-bin subtraction
R. Santoro
IPRD16, October 2016, Siena (Italy)

14. SiPM response to a β source
Spectrum with 90Sr after background subtraction
Fit performed using a fat tail distribution: a Moyal function* (Gaussian Landau convolution)
Most Probable Value = 22.4 ph-e
Assuming a cross-talk probability of 60%, we get a mean cluster size of ≈ 9 cells
(Npixels=MPV*(1-xTalk))
*  J. E. Moyal “Theory of Ionization Fluctuations” Phil. Mag. 46 (1955), 263
R. Santoro
IPRD16, October 2016, Siena (Italy)

15. Is the effect dependent on SiPM pitch size?
Threshold scan without source compared to threshold scan with 90Sr
SiPM (S ) 3x3 mm2 with 10μm pixel size
The data are normalized wrt the trigger rate at the same threshold.
The effect is still visible but we cannot be quantitative
The cluster size cannot be measured because we don’t see the peak-to- peak distance
R. Santoro
IPRD16, October 2016, Siena (Italy)

16. Same analysis with new sensors (just arrived!)
Threshold scan without source VS threshold scan with 90Sr
Detector size 1.3 x 1.3 mm2, (Hamamatsu S PE / S PE )
Two different pitch sizes (50 and 25 μm)
Low xTalk probability
New technology (we don’t know if they can be compared with the previous)
XT=3%
XT=1.4%
Pitch size = 50 μm
Pitch size = 25 μm
R. Santoro
IPRD16, October 2016, Siena (Italy)

17. Same analysis with the new sensors (just arrived!)
Spectrum without source VS spectrum with 90Sr
Detector size 1.3 x 1.3 mm2 (Hamamatsu S PE)
No background subtraction is required
XT = 1.4%
Ncells≈ 6.74
Pitch size = 25 μm
Background
90Sr
Threshold
Comparable with the previous sensor
R. Santoro
IPRD16, October 2016, Siena (Italy)

18. Looking at the digitized signal
Detector size 1.3 x 1.3 mm2, (Hamamatsu S PE)
xTalk probability 1.4 %
Signal digitized at 250 Ms/s
Typical DCR Event
Typical Event with 90Sr
R. Santoro

19. Looking at the digitized signal
Detector size 1.3 x 1.3 mm2, (Hamamatsu S PE)
xTalk probability 1.4 %
Signal digitized at 250 Ms/s
Primary Avalanche
Delayed Avalanches
Typical Event with 90Sr
Filter applied to count the peaks
R. Santoro

20. Time distribution analysis (DCR)
Detector size 1.3 x 1.3 mm2, (Hamamatsu S PE)
xTalk probability 1.4 %
Signal digitized at 250 Ms/s
Typical DCR Event
Number of delayed avalanches distribution plot
Primary Avalanche Peak Height distribution plot
MPV ≈ 1phe
MPV ≈ 0
R. Santoro

21. Time distribution analysis (90Sr)
Detector size 1.3 x 1.3 mm2, (Hamamatsu S PE)
xTalk probability 1.4 %
Signal digitized at 250 Ms/s
Typical Event with 90Sr
Primary Avalanche Peak Height distribution plot
Number of delayed avalanches distribution plot
MPV ≈ 4 phe
MPV ≈ 2
R. Santoro

22. Conclusions and Outlook
The SiPM sensitivity to direct detection of ionizing particles has been shown
The effect has two main components
Cluster effect: a certain number of cells start the avalanche at the same time
Delayed effect: a certain number of avalanches start after the primary ones
A more exhaustive analysis is on the way although a collaboration with SiPM producers will be beneficial to allow the proper description of the observed phenomena
R. Santoro
IPRD16, October 2016, Siena (Italy)

23. Spares
R. Santoro
IPRD16, October 2016, Siena (Italy)

24. Spectrum with / without source
Spectrum without source compared to spectrum with 90Sr
Trigger set at 16mV (≈1.5 ph-e)
The two spectra are normalized in time before the bin-bin subtraction
R. Santoro
IPRD16, October 2016, Siena (Italy)

25. Is the effect sensible to the SiPM pitch size?
SiPM (S ) 3x3 mm2 with 10μm pixel size
Fit with a fat tail distribution: a Moyal function (Gaussian Landau convolution)
Measured MPV = 250 a.u.
the cluster size cannot be measured since don’t have the peak-peak distance
χ2/dof = 2.95
Signal depends on the cell capacitance
Assuming all is the same except the area, I presume the signal to be ≈25 time smaller
If this assumption is correct I could use the previous dpp and I could make a guess on the MPV in ph-e
If in addition I presume to have the same xTalk, the expected mean cluster would be
MPV*(1-xT)= 11 cells
Really too much even for an educated guess!
R. Santoro
IPRD16, October 2016, Siena (Italy)

26. New sensors (just arrived!)
Spectra with 90Sr
Smaller detector size (1.3 x 1.3 mm2)
Different pitch sizes (25 and 50 μm)
Low xTalk probability
New technology (we don’t know if they can be compared with the previous)
Pitch size = 50 μm
Pitch size = 25 μm
XT = 3%
Ncells≈ 6.8
XT = 1.4%
Ncells≈ 6.74
R. Santoro
IPRD16, October 2016, Siena (Italy)