Ultrafast detection in particle physics and positron emission tomography using SiPMs

Author’s Accepted Manuscript Ultrafast Detection in Particle Physics and Positron Emission Tomography Using SiPMs R. Dolenec, S. Korpar, P. Križan, R. Pestotnik

www.elsevier.com/locate/nima

PII: DOI: Reference:

S0168-9002(17)30499-0 http://dx.doi.org/10.1016/j.nima.2017.04.036 NIMA59824

To appear in: Nuclear Inst. and Methods in Physics Research, A Received date: 10 March 2017 Revised date: 18 April 2017 Accepted date: 21 April 2017 Cite this article as: R. Dolenec, S. Korpar, P. Križan and R. Pestotnik, Ultrafast Detection in Particle Physics and Positron Emission Tomography Using SiPMs, Nuclear Inst. and Methods in Physics Research, A, http://dx.doi.org/10.1016/j.nima.2017.04.036 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Ultrafast Detection in Particle Physics and Positron Emission Tomography Using SiPMs R. Dolenec∗,a,b , S. Korparb,c , P. Kriˇzana,b , R. Pestotnikb a Faculty

of Mathematics and Physics, University of Ljubljana, Ljubljana, Slovenia b Joˇ zef Stefan Institute, Ljubljana, Slovenia c Faculty of Chemistry and Chemical Engineering, University of Maribor, Maribor, Slovenia

Abstract Silicon photomultiplier (SiPM) photodetectors perform well in many particle and medical physics applications, especially where good efficiency, insensitivity to magnetic field and precise timing are required. In Cherenkov time-of-flight positron emission tomography the requirements for photodetector performance are especially high. On average only a couple of photons are available for detection and the best possible timing resolution is needed. Using SiPMs as photodetectors enables good detection efficiency, but the large sensitive area devices needed have somewhat limited time resolution for single photons. We have observed an additional degradation of the timing at very low light intensities due to delayed events in distribution of signals resulting from multiple fired micro cells. In this work we present the timing properties of AdvanSiD ASD-NUV3SP-40 SiPM at single photon level picosecond laser illumination and a simple modification of the time-walk correction algorithm, that resulted in reduced degradation of timing resolution due to the delayed events. Key words: Silicon photomultipliers, PET, Time-of-flight, Cherenkov radiation

∗ Corresponding author at: Joˇ zef Stefan Institute, Jamova cesta 39, SI-1000 Ljubljana, Slovenia. Tel.: +386 1 477 3157, Fax.: +386 1 477 3166. Email address: [email protected] (R. Dolenec)

Preprint submitted to Nuclear Instruments and Methods A

April 22, 2017

1

2

1. Introduction By measuring the time-of-flight (TOF) of the annihilation gammas in positron

3

emission tomography (PET) the quality of activity distribution images can be

4

improved. Despite many recent efforts to develop new scintillator materials,

5

the time resolution in PET is still mainly limited by the time evolution of scin-

6

tillation light production. To completely avoid this limitation, the promptly

7

produced Cherenkov light can be used instead. We previously demonstrated

8

a coincidence resolving time below 100 ps FWHM using lead fluoride (PbF2 ),

9

a pure emitter of Cherenkov light, coupled to microchannel plate photomulti-

10

plier (MCP PMT) photodetectors [1]. We also tested silicon photomultipliers

11

(SiPMs) as photodetectors due to their relatively high photon detection effi-

12

ciency (PDE), low cost and the possibility to operate them inside the magnetic

13

field of magnetic resonance scanners. SiPMs have two drawbacks for use in

14

Cherenkov PET: they need to be cooled so their dark count rate is sufficiently

15

reduced and their time resolution is slightly limited. In our previous study with

16

four SiPMs, produced by different manufacturers, we achived the best coinci-

17

dence resolving time of 300 ps FWHM with AdvanSiD ASD-NUV3S-P-40 SiPM

18

[2]. This timing improved to 190 ps FWHM if only the events, corresponding to

19

single fired micro cell (m.c.) on both sides of the coincidence were used. In our

20

latest work [3], we have observed that the multiple m.c. events have the timing

21

degraded by an additional contribution, detected with a delay of approximately

22

0.2 ns, and suggested that the most likely source for this effects is the optical

23

crosstalk. The delayed contribution degrades the timing in two ways: the spread

24

in time of the events increases and the accuracy of the time-walk correction is

25

degraded.

26

In this work we present the response of the AdvanSiD SiPM to single pho-

27

ton level laser pulses and a simple modification of the time-walk correction

28

algorithm, with which we achieved an improvement in time resolution without

29

discarding the multiple m.c. events. The experiment and the standard time-

30

walk correction algorithm are presented in Section 2. In Section 3 the results

2

31

obtained with the standard and improved time-walk correction algorithm are

32

shown. This is then followed by the summary in Section 4.

33

2. Methods

34

2.1. Experimental setup

35

An AdvanSiD ASD-NUV3S-P-40 SiPM [4] with 3×3 mm2 active area and

36

40 µm cell size was connected to a custom electronic board featuring a NEC

37

uPC2710TB preamplifier (upper limit operating frequency 1.0 GHz) [5]. The

38

photodetector was illuminated by pulses, generated by a PiLas diode laser sys-

39

tem EIG1000D with blue (404 nm) laser head. Neutral density filters were used

40

to attenuate the laser pulses to very low intensity, corresponding to approxi-

41

mately one detected photon per pulse on average. The light was guided inside

42

the light tight, temperature controlled chamber via an optical fiber and con-

43

nected to an optical system, used to illuminated the whole SiPM active surface

44

approximately uniformly.

45

The preamplified SiPM signals were lead into an ORTEC FTA820 NIM

46

amplifier with 350 MHz bandwidth and then split into a CAEN Mod.V965

47

charge-to-digital converter (QDC) and Phillips Scientific Mod.752 leading edge

48

discriminator. The measurement was triggered by the laser control unit. Time

49

relative to the trigger signal was obtained by leading the logic signal from dis-

50

criminator into a Kaizu Works KC3781A time-to-digital converter (TDC). The

51

discriminator threshold was set to 0.5 single m.c. pulse height for every mea-

52

surement.

53

All of the presented measurements were performed at a temperature of

54

−25◦ C. The SiPM breakdown voltage at this temperature was determined by

55

measuring the signal heights at different overvoltages and extrapolating to zero

56

signal height.

57

2.2. Standard time-walk correction

58

For each laser trigger, the time and charge of the signal was measured.

59

By plotting time vs. charge correlation plot (Fig. 1, top) the time-walk effect 3

60

becomes apparent. This 2-D correlation plot was divided into a large number

61

of projections along the time axis, which were individually fitted with Gaussian

62

functions. The mean values of the Gaussian functions from each projection

63

64

were forming the basis for the standard time-walk correction function fit. The √ correction function used was of the form: f (x) = p0 + p1 / x − p2 . Here, p0,1,2

65

are the fit parameters which were then used to calculate the corrected time

66

on event per event basis by subtracting the value of the correction function at

67

measured charge from the measured time.

68

2.3. Time resolution

69

The time-walk corrected time distributions were fitted with a sum of a Gaus-

70

sian and a product of an exponential and error functions. The time resolution

71

was then calculated as the FWHM of the fit function.

72

With the time-walk correction method, described above, a time resolution of

73

58 ps FWHM was measured when an ultrafast reference photodetector, Hama-

74

matsu R3809U-52 MCP PMT, was used in the experiment. According to pro-

75

ducers test sheets, the MCP PMT and the laser contribute about 25 ps FWHM

76

and 35 ps FWHM respectively. This leaves any other contributions (e.g. elec-

77

tronics, time-walk correction) at roughly 40 ps FWHM.

78

3. Results

79

In Fig. 1, top, the measured time vs. charge correlation plot is shown.

80

The peaks corresponding to one, two and three fired micro cells are clearly

81

separated while the events with a higher measured charge are visibly affected

82

by the saturation of the amplifier. The time-walk correction function follows the

83

single m.c. events well, but the situation is more complicated for the multiple

84

m.c. signals, which are composed of two contributions. In Fig. 1, bottom,

85

the two contributions to the double m.c. events are more clearly visible. The

86

first contribution, detected at shorter time, fits the expectations of true double

87

photon detection. The second contribution is registered with a delay of about

4

88

0.2 ns and also has a slightly larger spread in time than the first. As discussed

89

in our previous work [3], the most likely cause of the delayed contributions

90

is the SiPM optical crosstalk. This conclusion is also supported by previous

91

investigations of SiPMs similar to the one used in this work. The measured

92

crosstalk probabilities were reported in [6], while the degrading effect on the

93

time resolution, produced by the delayed contribution, was already observed in

94

[7]. The AdvanSiD device tested in this work is optimized for low afterpulse rates

95

and seems to have only a limited isolation of neighboring microcells. Optical

96

crosstalk could be reduced by improved trenches between the microcells.

97

The time distribution obtained from this data and the contributions from

98

single and double m.c. events are shown in Fig. 2. A time resolution of 218 ps

99

and 149 ps FWHM was obtained for all and single m.c. events, respectively.

100

The resolution for double m.c. events was degraded to 314 ps FWHM due to

101

the delayed contribution. It is also clearly visible that the delayed contribution

102

caused an suboptimal time-walk correction function fit which resulted in an

103

improper shift in time for the double m.c. events.

104

In an attempt to improve the time resolution in the presence of the delayed

105

contribution to the multiple m.c. signals, events from each measurement were

106

split into subsets, corresponding to a certain number of fired micro cells (for

107

example, such subset corresponding to two fired cells was shown in Fig. 1, bot-

108

tom). The time-walk correction function fit and time resolution calculation was

109

then performed separately for each subset. In Fig. 3, top, the so obtained time

110

resolutions for subsets corresponding to only single and only double m.c. signals

111

are shown. For only single m.c. signals the time resolution continuously im-

112

proves with the overvoltage and the best time resolution achieved at the highest

113

overvoltage was 136 ps FWHM. For only double m.c. signals the delayed contri-

114

bution significantly influences the shape of the corrected time distribution and

115

degrades the resolution.

116

The time distributions including only events, corresponding to a certain

117

number of fired micro cells, corrected with parameters obtained from the sub-

118

set of such events, can be combined. In Fig. 3, bottom, the time resolutions 5

119

obtained with such improved time-walk correction are compared to those ob-

120

tained with the standard time-walk correction, presented in Section 2.2. The

121

time resolutions obtained with the improved approach are slightly better than

122

those obtained from the standard time-walk correction over the whole data set

123

and continue to improve also at higher overvoltages. The best result achieved

124

was 173 ps FWHM with the improved time-walk correction, compared to the

125

best result with the original algorithm of 218 ps FWHM.

126

4. Summary

127

The true single photon (single fired cell) response of SiPMs depends mainly

128

on the intrinsic properties of individual micro cell and the variations in signal

129

travel time from individual cells to the common output. Some of the older large

130

area SiPMs seemed to be limited in timing especially by the latter, but the latest

131

generations of devices are much improved. We measured a time resolution of

132

136 ps FWHM with AdvanSiD ASD-NUV3S-P-40 SiPM when only single m.c.

133

signals were used. When also the multiple m.c. signals were included, the time

134

resolution degraded to 218 ps FWHM. For best possible timing at single photon

135

level illumination, multiple m.c. signals should therefore be excluded. However,

136

using all detected events is important to keep the gamma detection efficiency of

137

the Cherenkov TOF PET method as high as possible.

138

We observed a delayed contribution to the multiple m.c. signals which not

139

only degraded the time resolution but also degraded the performance of the time-

140

walk correction. We achieved a small improvement in the time resolution, while

141

keeping also the multiple m.c. signals, by performing the time-walk correction

142

separately for subsets of data corresponding to a certain number of fired micro

143

cells and combining the obtained time distributions. In this way we achieved a

144

time resolution of 173 ps FWHM.

145

The presented modification of the time-walk correction algorithm is actu-

146

ally minimal and represents just the very first step of improvement. It should

147

be possible to improve the time resolution even further by developing a time-

6

148

walk algorithm even better suited to the SiPM response. Also, discrimination

149

methods based on pulse shape (multi-threshold measurement or analysis of in-

150

dividual waveforms) might be able to separate the two contributions and reduce

151

the degrading effect the delayed contribution has on the time resolution. All of

152

this is the subject of our current investigations, which will be presented in our

153

future works.

154

References

155

[1] S. Korpar, et al., Nuclear Instruments and Methods in Physics Research

156

Section A: Accelerators, Spectrometers, Detectors and Associated Equip-

157

ment 654 (2011) 532.

158

159

160

161

162

163

164

165

166

167

168

169

[2] R. Dolenec, et al., IEEE Transactions on Nuclear Science, Vol. 64, No. 5, October 2016, 2478. [3] R. Dolenec, et al., to be published in IEEE Nuclear Science Symposium Conference Record (NSS/MIC) (2016). [4] AdvanSiD ASD-NUV3S-P Low Afterpulse Datasheet Rev. 9; 10.2014 (www.advansid.com). [5] NEC Bipolar Analog Integrated Circuit µPC2710TB Data Sheet, NEC Corporation, January 2001. [6] P.W. Cattaneo, et al., IEEE Transactions on Nuclear Science, Vol. 61, No. 5, October 2014, 2657. [7] F. Acerbi, et al., IEEE Transactions on Nuclear Science, Vol. 61, No. 5, October 2014, 2678.

7

Time [ns]

0.6

1000

0.5 800

0.4 0.3

600

0.2 400

0.1 0

200 -0.1 -0.2

0 200

300

400

500

600

700

800

900

1000

1100

Time [ns]

Charge [ADC Ch.]

350

0.4

300 0.3 250 0.2

200

0.1

150

0

100 50

-0.1

0 460

480

500

520

540

560

580

600

Charge [ADC Ch.]

Figure 1: The time vs. charge correlation plot obtained at an overvoltage of 6 V (top) and a close-up of the double micro cell events (bottom). Also shown is the time-walk correction function obtained from the fit over all events.

8

Counts

All: 218 ps

50000

1 m.c.: 149 ps 40000

2 m.c.: 314 ps

30000

20000

10000

0 -0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

Time[ns]

Figure 2: The standard time-walk corrected time distribution and resulting fits at an overvoltage of 6 V for all, single and double m.c. events.

9

Timing FWHM [ps]

350 300 250 200 150 100

1 m.c.

50 0 3

2 m.c. 3.5

4

4.5

5

5.5

6

6.5

7

Timing FWHM [ps]

Overvoltage [V] 350 300 250 200 150 100 50 0 3

Standard Improved 3.5

4

4.5

5

5.5

6

6.5

7

Overvoltage [V]

Figure 3: The time-walk corrected time resolutions obtained for single and double m.c. signals (top) and for all signals with the standard and improved time-walk corrections (bottom) at different overvoltages.

10