Practical Guide to Using SiPMs Stefan Gundacker

Transcription

1 Practical Guide to Using SiPMs Stefan Gundacker 7th of June 2015

2 The Silicon Photo Multiplier (SiPM) Array of self quenched Geiger-mode APDs (microcells, SPADs) connected in parallel. Microcell or single photon avalanche diode (SPAD) Different SPAD sizes, for example: 3mm 25x25μm2, 50x50μm2 or 100x100μm2 Nowadays the trend goes to replace PMTs by SiPMs, because of insensitivity to magnetic fields, high gain, high photon detection efficiency, robustness, low operation voltage, excellent timing properties,... This trend is observed in high energy physics CMS: ECAL readout with avalanche photodiodes and in positron emission tomography (PET) time of flight and PET/MR. There are new developments going on digital SiPM from Philips which are more like digital photon counters. This tutorial, however, will mainly concentrate on analog SiPMs. 2

3 The Silicon Photo Multiplier (SiPM) Array of self quenched Geiger-mode APDs (microcells, SPADs) connected in parallel. Microcell or single photon avalanche diode (SPAD) Different SPAD sizes, for example: 3mm 25x25μm2, 50x50μm2 or 100x100μm2 SiPM has exceptional photon counting ability. 4 SPAD 3 SPAD 2 SPAD 1 SPAD Photon detection efficiency (PDE) and single photon time resolution (SPTR) are important quantities to characterize the SiPMs. 3

4 Content General description of avalanche photodiodes (APD) - photodetection - from the APD to the Geiger-mode APD - avalanche quenching and triggering - basic noise sources in a silicon photomultiplier (SiPM) Getting a SiPM to work: equivalent circuit, biasing Amplifier concepts for the SiPM signals (voltage vs transimpedance) How to treat the SiPM signals: filtering Practical guides and some results for different measurements with SiPMs - Photon detection efficiency (PDE) - Single photon time resolution (SPTR) - Coincidence time resolution (CTR) 4

5 Principle of photodetection 1st step: Photoconversion: photoelectric effect produced electron hole (e/h) pair Band gap (T=300K) = 1.12 ev (~1100 nm) photon electron hole Beer-lambert law: I(λ, z)=i(λ) e α( λ) z I(λ): initial photon flux z: penetrated distance in Si I(λ,z): photon flux at distance z α(λ): absorption coefficient inverse absorption coefficient (um) Absorption length (1/α) in Silicon 1E+5 1E+4 1/α ~ 63% absorbed 1E+3 1E+2 1E+1 1E+0 1E-1 1E-2 1E wavelength (nm) Green MA, Keevers MJ. Progress in Photovoltaics 1995, 3:

6 Principle of photodetection 1st step: Photoconversion: photoelectric effect produced electron hole (e/h) pair Band gap (T=300K) = 1.12 ev (~1100 nm) photon electron hole 2nd step: Photoelectron collection: once created, the electron/hole pair can be lost (absorption, recombination) If only collection of the produced electron/hole pairs PiN Diode 3rd step: Multiplication: The primary electron/hole pair is amplified (photodetector with internal gain, e.g. avalanche photodiodes, SiPMs) 6

7 Operation ranges 3rd step: Multiplication: The primary electron/hole pair is amplified (photodetector with internal gain, e.g. avalanche photodiodes (APDs), SiPMs) 7

8 Operation ranges 3rd step: Multiplication: The primary electron/hole pair is amplified (photodetector with internal gain, e.g. avalanche photodiodes (APDs), SiPMs) Geiger-mode avalanche 8

9 Multiplication: avalanche modes Only electrons contribute to the avalanche. Avalanche self quenched. Electrons and holes contribute to the avalanche. Avalanche self sustained, external quenching necessary (resistor). 9

10 Operation ranges 3rd step: Multiplication: The primary electron/hole pair is amplified (photodetector with internal gain, e.g. avalanche photodiodes (APDs), SiPMs) Ionization coefficients α for electrons and β for holes electrons SiPM region holes APD region 10

11 Quenching the avalanche Geiger-mode photon avalanche diodes γ Serial quenching resistor and parasitic (or engineered) quenching capacitance zoom into one cell CD: diode capacitance RS: silicon substrate serial resistor VBD: breakdown voltage A B: avalanche triggered, discharge of CD to VBD, asymptotic growth of current B C: voltage drop on Rq quenches the avalanche C A: CD recharges via the quench resistor (fast recharge provokes high afterpulsing) 11

12 Avalanche trigger probability in SiPMs Ionization coefficients α for electrons and β for holes Avalanche trigger probability SiPM region G. William et.al, IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. ED-19, NO. 9, Sept Because of the higher impact ionisation is the avalanche trigger probability always higher for electrons than for holes. 12

13 Two SiPM technologies: p on n and n on p p on n, e.g. Hamamatsu, Ketek, FBK-NUV More sensitive in the blue and near UV (because of electrons produced near the p++ layer triggering the avalanche) n on p, e.g. FBK-RGB More sensitive in the red-green (because of electrons produced near the p+ layer triggering the avalanche) 13

14 Two SiPM technologies: p on n and n on p electrons p on n, e.g. Hamamatsu, Ketek, FBK-NUV More sensitive in the blue and near UV (because of electrons produced near the p++ layer triggering the avalanche) electrons n on p, e.g. FBK-RGB More sensitive in the red-green (because of electrons produced near the p+ layer triggering the avalanche) 14

15 The SiPM: Array of Geiger-mode APDs Array of self quenched Geiger-mode APDs (microcells, SPADs) connected in parallel. Microcell or single photon avalanche diode (SPAD) avalanche photo diode in Geiger-mode 3mm 4 SPAD 3 SPAD 2 SPAD 1 SPAD Each avalanche in SPAD gives same signal (1 or 0) photon counting SPAD signal always the same, no matter how many photons initially triggered the avalanche. Photon counting and good energy resoltuion with scintillator due to the high number of SPADs per SiPM feasible. 15

16 Noise sources in a SiPM Array of self quenched Geiger-mode APDs (microcells, SPADs) connected in parallel. SPAD can have avalanche due to thermal or via tunnelling produced electron/hole pair dark count 3mm Breakdown production of a large number of charge carriers some of them are trapped in deep trap levels and eventually released at later times provoking new avalanches in the same microcell afterpulsing (A 8 C lower temperature reduces dark count rate by a factor of 2.) Avalanche produces light! these photons can produce avalanches in neighbouring cells optical crosstalk 16

17 Noise sources in a SiPM Dark counts: pulses triggered by non-photo-generated carriers (thermal / tunnelling generation in the bulk or in the surface depleted region around the junction) Afterpulses: carriers trapped during the avalanche can produce delayed secondary pulses Crosstalk: avalanche in one cell probability that an avalanche produced photon triggers another avalanche in a neighboring cell without delay 17

18 I-V curve and breakdown voltage Very important parameter characterizing the bias voltage when Geiger-mode avalanche starts to occur. Parameters like photon detection efficiency (PDE), single photon time resolution (SPTR), dark count rate, crosstalk or afterpulse are dependent on overvoltage not bias voltage. Breakdown voltage strongly temperature dependent ~50mV/K for Hamamatsu MPPC 2V overvoltage 18

19 Equivalent circuit and biasing of a SiPM 19

20 Equivalent circuit of a SiPM S G-APD (Diode) S. Seifert, IEEE TNS VOL. 56, NO. 6, Dec Signal rise time is product of diode capacitance (CD) and silicon substrate serial resistor (RS): τr=cd*rs= in the order of hundreds of picoseconds. Quench capacitance (Cq) provokes fast initial signal, good for timing Signal fall time is product of diode capacitance (CD) and quenching resistor (RS): τf=cd*rq= in the order of tens of nanoseconds. 20

21 Biasing a SiPM Power supply ripples blocked by capacitance Power supply filter Cathode Optional, suppresses low frequency hum (50Hz pick-up). Anode 21

22 Energy measurement with simple setup LSO SiPM Source QDC Measured with Hamamatsu MPPC 3x3mm2 and 25um cell size coupled to standard LSO scintillator. 22

23 Saturation and energy measurement with LSO Collected charge is proportional to the number of fired cells Nfired cells: ( N fired cells =N total 1 e N photon PDE N total ) Measurements with Hamamatsu MPPC 3x3mm2 (25um, 50um, 100um cell size) coupled to standard LSO scintillator for different gamma energies: Ntotal: total number of SPADs Nphoton: impinging photons on SiPM Formula only correct for light pulses which are shorter than pixel recovery time, and for an ideal SiPM (no crosstalk and no afterpulsing) 23

24 Amplifier concepts for SiPMs 24

25 Voltage vs transimpedance amplifier Very stable Source resistor (RS) easy changeable Low RS needed for fast SiPM signal low signal to noise ratio Very low input impedance preferred readout for SiPMs Tendency to oscillations, less stable Gain is defined by Rf Gain=Rf/Rg typically 10 to

26 Practical voltage amplifier ( Gain= = ) signal signal Differential readout suppresses low to medium high frequency pick-up. 26

27 SiPM 2x2mm2 vs 3x3mm2 with voltage amplifier Hamamatsu through silicon via (TSV) MPPC 2x2mm2 50um cell size (2.5V overvoltage) 10ns/div & 100mV/div Hamamatsu through silicon via (TSV) MPPC 3x3mm2 50um cell size (2.5V overvoltage) 10ns/div & 100mV/div SPAD signal for voltage amplification becomes lower if device size (capacitance) gets larger. Due to capacitive divider (quenching and SiPM terminal capacitance). 27

28 Practical transimpedance amplifier Gain= = Differential readout suppresses low to medium high frequency pick-up. 28

29 SiPM 2x2mm2 vs 3x3mm2 with transimp. amplifier Hamamatsu through silicon via (TSV) MPPC 2x2mm2 50um cell size (2.5V overvoltage) 10ns/div & 100mV/div Hamamatsu through silicon via (TSV) MPPC 3x3mm2 50um cell size (2.5V overvoltage) 10ns/div & 100mV/div SPAD signal for transimpedance amplification stays the same for different device sizes, if input impedance of amplifier is lower than device impedance. 29

30 Quenching capacitance Example on Hamamatsu LCT (low crosstalk) 3x3mm2 SiPM at 50V bias voltage and 15 C Cq provokes fast signal at the start Recharging via Rq is slow 30

31 Filtering the SiPM signals Reducing this capacitor down to 100pF gives a τ of 1ns, suppresses baseline fluctuations strongly. Generally this is enough. Pole-zero compensation: Also possible to play with this capacitor, e.g. 100pf gives a τ of 5ns. Avoiding overshoot in the tail of signal A. Gola, et.al, IEEE TNS, VOL. 60, NO. 2, APRIL

32 Filtering with high-pass or pole zero Example on STM 4x4mm2 SiPM at 32V bias voltage and 15 C Direct signal High-pass filtered with 100pf on 50 Ohm (input oscilloscope) 32

33 Filtering with high-pass or pole zero Example on STM 4x4mm2 SiPM at 32V bias voltage and 15 C Direct signal High-pass filtered with 100pf on 50 Ohm (input oscilloscope) 33

34 Excellent single photon resolution with filtering Example on STM 4x4mm2 SiPM at 30V bias voltage and 15 C Pedestal 27 photoelectons 34

35 Putting the readout into practice 35

36 NINO a fast preamplifier discriminator Developed for the Alice TOF project to readout multigap resistive plate chambers (MRPCs). Common gate cascode amplifier with low input impedance of 40Ω. Differential input: F. Anghinolfi et.al, IEEE Trans. Nucl. Sci., vol. 51, pp , October

37 NINO a fast preamplifier discriminator Subsequent cascade amplifiers with total gain of 1296 allow operation as discriminator. The threshold is applied via a current difference in the NINO input cascode stage. F. Anghinolfi et.al, IEEE Trans. Nucl. Sci., vol. 51, pp , October

38 NINO time over threshold principle 500μm Time-over-threshold discrimination: Leading edge gives time information and pulse width is a function of the energy. F. Powolny et. al., IEEE Trans. Nucl. Sci., vol. 58, pp , June

39 Other electronics for SiPM readout FLC_SiPM MAROC SPIROC PETA BASIC SPIDER RAPSODI TOFPET Asic STiC Orsay Orsay Orsay Heidelberg Bari/Pisa Siena/Pisa/Oslo Krakow Both developed in the Lisbon frame of the EndoHeidelberg TOFPET-US project 39

40 Practical discussions of measurements with SiPMs 40

41 Photon detection efficiency (PDE) 41

42 Setup of the PDE bench Measures constantly the light intensity impinging on the SiPM (calibration factor known) Calibrated photodiode, light source and SiPM are perpendicular to each other. Pulsed light source: LED with 10ns pulse width Data acquisition: charge measurement Changing the LED allows to measure at different wavelength 42

43 Analysis and measurement 1pe Poisson statistics: 2pe 3pe Number of detected photoelectrons (npe) estimated by the probability of detecting zero photoelectrons (no crosstalk and afterpulsing influence!): 4pe Correction for dark counts in detection window: 43

44 Example of results Hamamatsu TSV 3x3mm2 at 410nm wavelength Hamamatsu TSV 3x3mm2 1.5V, 2.3V, 3.0V and 3.5V 3.5V 3.0V 2.3V corrected for dark counts 1.5V PDE = geometric fill factor * avalanche trigger probability * quantum efficiency quantum efficiency = light transmission at surface * (1 - e/h recombination prob.) * photon absorption probability in depleted layer 44

45 Single photon time resolution (SPTR) 45

46 Laser Tests on SiPM Femtosecond laser (400nm) with 200fs pulse width LeCroy Oscilloscope DDA 735Zi 40Gs/s In collaboration with Physical Chemistry Department University of Geneva B.Lang, E.Vauthey S. Gundacker et.al, NIMA, vol. 718, pp , August

47 Laser Tests on SiPM Selecting only events in first photoelectron S. Gundacker et.al, NIMA, vol. 718, pp , August

48 Laser Tests on SiPM 100μm type can not be operated at optimum bias voltages, due to high dark count. S. Gundacker et.al, NIMA, vol. 718, pp , August

49 Laser Tests on SiPM FWHM CTR =280ps FWHM CTR =93ps 100μm type can not be operated at optimum bias voltages, due to high dark count. SPTR in coincidence FWHM = 280ps (85ps sigma * 1.414*2.35) although, for 4 photons it is 140ps S. Gundacker et.al, NIMA, vol. 718, pp , August

50 Electronic noise and SPTR Timing jitter due to electronic noise: σe σ j= ( dv /dt ) Threshold F. Powolny, PhD Thesis Single SPAD of 50um up to SiPM of 3x3mm2 show similar intrinsic timing, when accounted for electronic noise. F. Acerbi, et.al, IEEE TNS, VOL. 61, NO. 5, OCTOBER

51 Coincidence time resolution (CTR) with 511keV gammas 51

52 Components of the radiation detector 52

53 Measurement setup for TOF-PET Coincidence measurement for the 511keV photon pair: Crystal: LSO:Ce,0.4% Ca 2x2x5mm3 producer Agile, fully wrapped in Teflon and coupled with optical grease SiPMs: -) Hamamatsu 25μ -) Hamamatsu 50μ -) Hamamatsu 100μ Mounted in temperature stabilised (20 C) dark box Data acquisition: LeCroy Oscilloscope DDA 735Zi with 3.5GHz Bandwith and 40Gs/s SiPM and electronics (NINO chip) assembled on one board. S. Gundacker et.al, IEEE vol. 59, no. 5, pp , Oct

54 Data Analysis Crystal: LSO:Ce,0.4% Ca 2x2x5mm3 SiPM: Hamamatsu 50μ Acquire pulse width and delay trends of both channels oscilloscope. After selecting photopeak on both sides deduce the time resolution. time with the we vertical projection 54

55 Data Analysis Crystal: LSO:Ce,0.4% Ca 2x2x5mm3 SiPM: Hamamatsu 50μ Acquire pulse width and delay trends of both channels oscilloscope. After selecting photopeak on both sides deduce the time resolution. time with the we vertical projection FWHM=142±4ps horizontal projection 55

56 Photopeak selection NINO pulse width = log(energy) Bias overvoltage 1.2V Bias overvoltage 2.2V Bias overvoltage 2.5V analog (linear) analog (linear) analog (linear) NINO NINO NINO 56

57 SiPM Scintillator 511 kev gamma rays LSO:Ce codoped 0.4% Ca (2x2x3mm3) Photodetector current New electronics for SiPM readout MPPC from Hamamatsu, active area 3x3mm2 t Readout Electronics NINO analogue amplifier time information energy information Leading edge discrimination gives time information. SiPM pulse amplitude gives energy information. 57

58 New electronics for SiPM readout High-pass: down to 100pF (or pole zero compensation) Two different amplifiers: instrumentation amplifier for energy plus transimpedance amplifier for timing! (they do not influence each other) High-pass filter before NINO prevents baseline shifts 58

59 Developed board for SiPM readout voltage amplifier SiPM NINO chip S. Gundacker et.al, JINST, August JINST 8 P

60 Measurement setup Coincidence measurement for the 511keV photon pair: Photodetector (SiPM): MPPC from Hamamatsu, active area 3x3mm2 and microcell (SPAD) size 50μm. (3600 microcells) Mounted in temperature stabilised (20 C) dark box Data acquisition: LeCroy Oscilloscope DDA 735Zi with 3.5GHz Bandwith and 40Gs/s Crystals: LSO:Ce codoped 0.4% Ca 2x2x3mm3, 2x2x5mm3, 2x2x10mm3, 2x2x20mm3 producer Agile, wrapped in Teflon and coupled with optical grease to SiPM. S. Gundacker et.al, JINST, August JINST 8 P

61 Measurement setup CERN setup Detector 1 Detector 2 61

62 3 Results with 2x2x5mm LSO:Ce codoped Ca Old board with NINO pulse width energy encoding: New board with analog energy encoding: Improvements due to a better photopeak-selection and ability to operate SiPM at higher overvoltages. 62

63 3 Results with 2x2x3mm LSO:Ce codoped Ca Crystals: LSO:Ce codoped 0.4% Ca 2x2x3mm3 SiPMs: Hamamatsu MPPC 50μ CTR = 108±5ps FWHM Acquire SiPM pulse height and mutual leading edge delay time of both channels. After selecting the photopeak on both sides we deduce the time resolution. S. Gundacker et.al, JINST, August JINST 8 P

64 Measured CTR as a function of crystal length Bias overvoltage scan: Crystal: LSO:Ce codoped 0.4% Ca with 2x2mm2 cross section Photodetector: Hamamatsu S P MPPC 20mm length => CTR=176±7ps FWHM 10mm length => CTR=143±7ps FWHM 5mm length => CTR=123±8ps FWHM 3mm length => CTR=108±5ps FWHM S. Gundacker et al, NIMA, vol. 737, pp , February

65 Outlook: State of the art Measurements done at CERN with FBK-NUV SiPMs: CTR of 140ps FWHM with 20mm long LSO:Ce codoped 0.4%Ca crystal. CTR of 85ps FWHM with 3mm long LSO:Ce codoped 0.4%Ca crystal. M. V. Nemallapudi et.al, Phys. Med. Biol. 60 (2015)

66 Outlook: State of the art Measurements done at CERN with FBK-NUV SiPMs: SiPM + Crystal Na22 Source position 66

67 Summary Some advantages of SiPMs High gain ( ) with low bias voltage, but needs amplifiers and shapers because of the long recharging time Single photon detection with good timing resolution and excellent photon resolution Some disadvantages of SiPMs Very good coincindence time resolution (CTR) in a TOF-PET like configuration achievable Insesitivity to magnetic fields up to 7T proven High photon detection efficiency up to 40% Very robust, mechanically and against strong light sources High dark count rate at room temperature and other correlated noise, as crosstalk and afterpulses Precise and linear measurement of light intensity difficult (corrections for crosstalk, afterpuls, recovery time and saturation) High temperature dependence of breakdown voltage and thus the gain the Up to now, relatively small devices Needs amplifiers and shaping to deal with long decay time of signals Low cost in mass production possible A lot of R&D still ongoing --> higher PDE, better SPTR to be expected 67

68 References Some SiPM references (incomplete list): RICH 2013: Status and Perspectives of Solid State Photo-Detector, Gianmaria Collazuol NDIP 2014: Tutorial SiPMs, Véronique Puill SiPM workshop, , CERN: State of the art in SiPM s, Yuri Musienko SiPMs: parameters and applications, Elena Popova Physics of semiconductor devices 3rd edition, S.M Sze (John Willey & Sons) Advances in solid state photon detectors from D. Renker and E. Lorenz Silicon Photomultiplier - New Era of Photon Detection from Valeri Saveliev 68

69 Do you have any questions?

70 Analog and digital concept of a SiPM - each cell connected to common readout - each cell (diode) acts as a switch when a photon is being detected - output is sum of every cell charge pulse - ideally time of every photon detected is being registered S. Gundacker, et.al, NIM A 787 (2015)

71 Maximum likelihood time estimation (MLTE) Bayes' theorem: The probability of a certain positron emission time θ under the condition of a measured data set D we want to maximize. Unbiasedness demands prior to be one. Likelihood function is known through the covariance matrix C. Normalization As likelihood function we suppose a multivariate Gaussian distribution: With the covariance matrix: S. Gundacker, et.al, NIM A 787 (2015)

72 Maximum likelihood time estimation (MLTE) Bayes' theorem: The probability of a certain positron emission time θ under the condition of a measured data set D we want to maximize. Unbiasedness demands prior to be one. Likelihood function is known through the covariance matrix C. Normalization Using Bayes' theorem one can derive the most likely time of positron annihilation: This is effectively a weighted average of the measured time stamps D with the coefficients. S. Gundacker, et.al, NIM A 787 (2015)

73 MLTE for zero photon travel spread (PTS=0) PTS zero, although same LTE (absorption) as for 2x2x3mm3 crystal S. Gundacker, et.al, NIM A 787 (2015)

74 MLTE leads to same CTR as analog time estimator For 2x2x20mm3 crystals S. Gundacker, et.al, NIM A 787 (2015)

75 Single cell signal and acquisition system noise CTR Osci = dv / dt noise =1.45mV dv /dt =179 V / s Data acquisition: LeCroy Oscilloscope DDA 735Zi with 3.5GHz Bandwith and 40Gs/s CTR(Osci)=27ps FWHM S. Gundacker et al, 2013 JINST 8 P

76 Influence of light output measured light output [photons/mev] Light output measured with a Photonis XP2020Q PMT measured CTR [ps] CTR corrected for light output If we correct for the light loss in the crystal (i.e. for photon statistics by multiplying with the square root of the light output loss) the CTR deterioration starts to saturate for longer crystals. S. Gundacker et al, NIMA, vol. 737, pp , February