santoro romualdo

Massimo Caccia
Universita’ dell’Insubria @ Como
massimo.caccia@uninsubria.it
Multi Pixel Photon Counters:
introducing the digital age in
low light detection
The Malmo lectures, October 19-21, 2015
Outline
1. Introduction
2. A review of the basic parameters
2.1 Photon Spectrum & Gain
2.2 Dark Counts & Cross Talk
2.3 After Pulsing
2.4 Photon Detection Efficiency
2.5 Gain Stabilization
2.6 What’s next?
3. Applications
3.1 Volcanoes & Nuclear Threats
3.2 More about Nuclear Waste
3.3 Radiation protection
3.4 Homeland Security
3.5 Response to a constant flux
3.6 PET
3.7 BioImaging
4. Conclusions
1.Introduction
Photomultiplier Tubes: amazing devices since
1934
Since then, the PMT continuously evolved, serving the industrial and
scientific community with a wealth of different design & specification
Assembling the ATLAS Tile
Calorimeter at LHC, CERN, Geneva
The Daya Bay reactor neutrino
experiment
CT / PET/ Combined images
Devices, facilities, experiments as beautiful as the ENIAC, back to 1946,
based on another solid rock technology
But soon after, a bright discovery changed the world….
Photon absorption and avalanche ignition in a
Single Photon Avalanche Photodiode: (SPAD)an
artist’s view
Ceci n’est pas un PMT…..
Pioneering development by S. Cova et al at Politecnico di MIlano
… and when you get to an array, a matrix of
SPAD, you get to the main subject of this talk,
with a clearly growing interest in the scientific &
industrial community:
No. of papers in Google
Scholar with the exact
match of “silicon
photomultiplier” in the
title/abstract.body
Year
# papers
2000-2001
11
2002-2003
31
2004-2005
82
2006-2007
211
2008-2009
366
2010-2011
603
2012-2103
1117
E. Charbon & S. Donati, OPN Optics & Photonics News, February 2010
Multi Pixel Photon Counters, a.k.a.
Silicon Photon Multipliers (SiPM)
SiPM = High density (~104/mm2 ) matrix of
diodes with a common output, reverse
biased, working in Geiger-Müller regime
When a photon hits a cell, the generated
charge carrier triggers an avalanche
multiplication in the junction by impact
ionization, with gain at the 106 level
Courtesy http://www.yk.rim.or.jp/~reyhori/pages/galacc2_e.html
genuine Photon Number Resolving Detectors… (1/2)
Multi Pixel Photon Counters, a.k.a.
Silicon Photon Multipliers (SiPM)
genuine Photon Number Resolving Detectors… (2/2)
 SiPM may be seen as a collection of
binary cells, fired when a photon in
absorbed
 “counting” cells provides an
information about the intensity of the
incoming light:
With an unprecedented
Photon Number Resolving
power:
SiPM: simplified electrical model
- C. Piemonte, NIM A 568 (2006) 224–232
- S. Seifert et al., IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 56, NO. 6, 2009
- P. Hallen, bachelor thesis, Aachen University, 2011
 exponential recharge with
t = RqCD
 Gain:
Vbias - VBreakdown ) CD
(
G=
e
Typical values:
 Rq ~ 200 kΩ
 CD ~ 100 fF (30x30 μm2 )
 τ ~ 20 ns
 Vbreakdown ~ 70V
 G ~ 106
As of today, MPPC come in different
flavours……….
So, being able to understand what’s being offered on this
Menu a’ la carte is more and more important…
The MPPC line-up by size
Linear array, 1x4 mm2, 4 channels
Single Channel
Squared array, 4 side buttable, Trough
Silicon Vias technology
 1x1 mm2
 3x3 mm2
 6x6 mm2
Ceramic package & SMT
 4x4
elements,12x1
2 mm2
 8x8
 16x16 on
the way
The MPPC line-up by cell pitch
Koei Yamamoto, HAMAMATSU, March 2014
100 μm is available as well!
2. A review of the basic parameters
Characterization protocol



I-V measurements (leakage current, quenching resistor, breakdown voltage)
Noise measurements (vs over voltage and vs temperature):
 dark counting rate (DCR) vs bias voltage
 optical cross-talk (DCR vs threshold)
 afterpulse
Analysis of (Poissonian photon) spectrum (vs temperature)
 resolution power (how many photons can I distinguish?) & gain
 working point optimization (at low and large flux)
 noise measurement (not DCR; essentially system noise and cell-to-cell
variations)
 optical cross-talk (deviations from the Poissonian distribution)

linearity & dynamic range

Spectral response (PDE vs λ, PDE vs temperature)

timing properties and time resolution (currently O(100ps))
2.1Photon Spectrum & Gain
Light source: CAEN ultra-fast LED (SP5601) OR Pico Quant PDL 800 light at ~ 510 nm (green)
being kAMP is the
ampli factor
∆peak

> 20 peaks
@ room T
Peak width:




System noise
Cell-to-cell gain
variation (process
uncertainties)
Ileak fluctuations
Spurious hits in the
QDC integration
time
Gain ~ 106
width ∞√cells
Xtremely good
resolution power!
Telling the difference between a good & poor
detector
Guess who is who?
• the front-end is the same
• the light source is the same
• the cell-to-cell variation is
definitely not the same
… and the resolution power
as well..
Gain vs overvoltage for the current products
Vbias - VBreakdown ) CD
(
G=
e
2.2 Dark Count Rate & Xtalk
Threshold scan
The Dark Counts (DCR) measure
the rate at which a Geiger
avalanche is randomly initiated
by thermal emission.
> 0.5 ph
0.5 ph
1.5 ph
2.5 ph
> 1.5 ph
DCR @ fixed threshold (0.5 phe-)
> 2.5 ph
an avalanche generation can fire another
cell by a photon; measuring the DCR for
different thresholds is possible to define and
evaluate the Optical Cross talk as:
Where is the optimal working point?
Where are the MPPC today? (1/2)
DCR vs overvoltage
 Dark Count rate: ~ 100 kHz/mm2 with the possibility to reduce it
considerably by cooling:
The dark count decreases by half for every approx. 8 °C drop
in temperature
Where are the MPPC today? (2/2)
 Optical cross talk:
Larger pixels have a higher gain  higher probability to
have a secondary photon to trigger a neighbouring cell
Not yet the end of the story…. Introducing
TRENCHES
2.3 After-pulsing
- Y. Du, F. Retiere (NIM A 596 (2008) 396–401
- Eckert et al., arXiv:1003.6071v2 (2010)
- M. Caccia et al., JIINST 9 T10004 (2014)
Delayed avalanches triggered by the release of a charge carriers that
has been produced in the original avalanche and trapped on an
impurity; the release of the trapped carriers is characterizated by a
typical decay time ~200 ns
Current after-pulsing figures:
Resulting from a continuous
improvement in the material
& the technology
2.4 Photon Detection Efficiency (PDE) (1/2)
PDE
Obviously, knowing the number of impinging photons is due but not
necessarily trivial
Quite Often, <n>ph is measured using an
integrating sphere and a reference
detector
[e.g. Otte et al. NIM A567 (2006) 360–363]
[but not only; see for instance
Y. Musienko, et al.,NIM A 567 (2006) 57]
Photon Detection Efficiency (PDE) (2/2)
Once the number of impinging photons is known, the number of photoelectrons can be measured by looking at the 0 photon peak
probability:
Eckert et al., arXiv:1003.6071v2, 2010)
Account for the bias on the estimate of npe due to the DCR
The method, originally due to Otte et al.(NIM A567 (2006) 360–363), has
the advantage of being insensitive to cross-talk and afterpulse..
PDE vs Gain
Spectral response
Photon Detection Efficiency of the MMPC:
2.5 Temperature variations of the gain


Pico Quant PDL 800 – light @ ~ 510 nm
Cooling Box + Temperature control
Gain vs bias voltage
Gain vs temperature
Temperature drift of V breakdown:
~ 60 mV/oC
Gain stabilization vs. T
Breakdown Voltage
rescaled accounting for
the temperature variation
tomorrow
Cross Talk (%)
today
DCR (kcps)
yesterday
AfterPulse (%)
2.6 What’s next? A significant noise reduction…
The MPPC can be operated at higher over-voltage values
The avalanche triggering probability is approaching the
asymptotic value
The PDE stability against T variations is increased!
A new high fill factor
design O(80%)
2.7 What matters most in the quality of the spectrum?
[a simulation game by dr. Romualdo Santoro @Uni.Insubria]
Examples of simulated signals,
corresponding to:
 SNR=10
 Signal decay time constant: 70 ns
 CellToCell gain variation: 0.1 p.e.
 Signal sampling time: 1 ns
 Integration gate =320 ns
Is the DCR relevant?
DCR=100 kHz
DCR=0, Xtalk=0
after pulse=0
DCR= 1 MHz
R. Santoro
35
How about the other stochastics
terms?
DCR=100 kHz, Xtalk=2%
+
after pulse = 2%
DCR=0, Xtalk=0
after pulse=0
DCR= 1 MHz, Xtalk=20%
+
after pulse=20%
R. Santoro
36
It does not look so dramatic, unless the development time of the
signal gets longer [because of the sensor OR the crystal scintillation
time!]
Tau = 70 ns
gate = 330 ns
Tau = 150 ns
gate = 750 ns
R. Santoro
DCR=1 MHz, Xtalk=20%
after pulse=20%
DCR=1 MHz, Xtalk=20%
after pulse=20%
37
2.8 HAMAMATSU vs TheRestOfTheWorld
[I might have added Excelitas & PHILIPS digital counting …]
HAMAMTSU
S13360-1350CS
Sensor
dimension [mm2]
sensL
c series
KETEK
PM1150
+trenches
1.3x1.3
1.0x1.0
1.2x1.2
Cell pitch [μm]
50
50
50
Vbreakdown [V[
53 ± 5
24.5 ± 0.2
25 ± 3
Voperational [V]
Vbr +3
(Vbr +2.5) ±2
(Vbr +5)
Gain@Vop [*106]
1.7
6
8
PDE@Vop
40
35
>50% (!)
λ@PDEmax [nm]
420
420
420
DCR [kHz]
90
30
<400
Xtalk [%]
1
10
15
<3
0.6
NA
AfterPulse [%]
IMHO (InMyHumbleOpinion)……
A good example...
…. sensL is your main competitor today, not only because of the quality of the sensors
but for a few clever design features, packaging, tiling (even if they do not offer TSV
yet, to my knowledge), stability of the main figures [I cannot comment about pricing,
you know it better than I do…]
A pictorial view of why a fast signal makes a difference:
the outcome of a filtering procedure we apply to digitized
signals while characterizing the sensors in our lab
Testing the Excelitas 6x6 mm2, 50 micron pitch cells
And timing may be expected to benefit as well, due to the reduced time walk
3.Applications
bioluminescence
Quantum
standards
Detector Calibration
Single molecule
detection
Hyper-spectral
imaging
imaging
Primary
radiometric
scales
Metrology
biotech
Single photon
Sensitive device
Imaging
Military
Meteorology
Nuclear Medicine
Night Vision
Remote Sensing
Lidar
Security
IR detectors
Environmental monitoring
Robust imaging devices
Single photon sources
radioactivity
Space
Ch. Telescope Array
computing
Medical Physics
Electronics
Entertainement
cryptography
Quantum info processing
Lighting
Displays
imaging
Chem-Bio Threat
Summary Table from http://www.photoncount.org
3.1 Volcanoes & Nuclear Threats:
MU-RAY: a geo-particle physics experiment
• G. Ambrosi et al , NIM A 628 (2011) 120–123
• http://mu-ray.fisica.unina.it
GOAL: to probe internal structures of volcanoes using almost horizontal
muons cosmic muons
MOTIVATION:
Foreseen eruption dynamics mostly depends on:
• Gas content
• Conduit dimensions and shape
• Chemical composition of magma
Mt. Vesuvius
Stromboli
METHOD: investigate the inner structure
of the volcano by cosmic ray muon
radiography
MU-RAY: not exactly a piece of cake…
Volcano model
Data rate
MU-RAY: design and construction
 3 x,y stations, 2 planes each, 4m2 area
 4 modules/plane
 32 scintillating bars/module
 1 SiPM/bar
768 SiPM, read-out by the
SPIROC chip
MU-RAY: state of the art
First plane done!
Mounting the planes
Photoelectrons/cosmic/plane
A nice muonic image of the mt. Vesuvio
Beyond Volcanoes…
Partnership with the UK National Nuclear Lab & Sellafiedl ltd
for imaging Nuclear Waste silos, filled up in 50’s and 60’s
Cosmic muons allow as well to inspect
something smaller…
[Riggi& Riggi, The MUONPortal collaboration, http://muoni.oact.inaf.it/]
Two possible designs of the extruded
scintilator bars, embedding wls fibers
Scanning a 6x3x3 m3 volume [in a
reasonable time]
Design:
• 8 planes (4x + 4y, crossed strips)
• 6 modules/plane [3x1m2 each]
• 100 strips/module [300x1x1cm3 each,
sensors @one hand only]
 4800 SiPM’s
Current results [simulation; 1 test module
constructed so far]
Results for ~10’ scan
[i.e. 5x105 muons]
My Point of view: this technique may
be very interesting for spotting orphan
sources in scrap material
3.2 More about Nuclear Waste
Paolo Finocchiaro, Nuclear Physics News,
http://dx.doi.org/10.1080/10619127.2014.941681
 Goal: online monitoring of radiation
emitted by nuclear waste drums
 Method: “annular” detector, made
out of a plastic scintillating fiber
connected to SiPM at both ends
Italian National Institute of Nuclear Physics (INFN) & SOGIN s.p.a
Qualification results:
 60Co, 35 kBq
 22Na, 243 kBq
 241Am, 36 MBq
 137Cs, 231 kBq
• efficiency for depositing at least 50 keV ~ 0.5%
• mean deposited energy ~ 180 keV, i.e. 1800 photons (light yield ~
104 photons/MeV)
• mean detected signal ~ 40 photo-electrons
• random coincidence rate ~ 1Hz
3.3 radiation protection [by a couple of
significant players in the field (1/2)]:
Personal gamma radiation detector based
on a SiPM coupled to a CsI crystal (battery
operated, 120h up time)
PDS-GO by Mirion Technologies,
Munich, Germany
https://www.mirion.com/
Radiation (2/2):
TN15 TM by KROMEK, Sedgefield
County Durham, UK
www.kromek.com
3.4 Homeland security: Silicon photomultiplier readout
of a scintillating noble gas neutron detector (2012-2014)
[M. Caccia et al. ANIMMA2013, MODES-SNM DOI: 10.1109/ANIMMA.2013.6727974
http://www.modes-snm.eu/
Project Objectives
Special Nuclear Material detection by
identifying a flux of fast neutrons in a highpressure tube filled with high pressure (180bar)
4He scintillating gas
Main drive:
• increased light detection efficiency
• engineering optimization
MPPC for MODES
Qualification of a prototype detector
Signal by a neutron
Signal by a gamma
Main Result
Corresponding to:
 fast neutron detection efficiency ~3%
 gamma rejection at the 106 level
3.5 Response to a constant flux:
Dosimetry in mammography
C. Cappellini et al., 2008 IEEE Nuclear Science Symposium Conference Record & NIM 607 (2009), 75–77
Dosimetry in mammography is
utmost important and this is
somehow proven by the ongoing
debate on the relevance of
mammography screening
…but currently existing
instruments are limited:
Standard Termo-Luminescent Detectors
require to be analyzed after examination,
degrade with time
 MOSFET detectors suffer from low stability
and degrade with each irradiation
 Ionization chamber devices need
relatively high voltage (cannot be used in
contact with the patient), not tissue
equivalent

precise measurements of the
actual dose being received by a
patient without distorting the X-ray
beam
and
introducing
any
artefacts in the image
Some functional requirements:
 dose rate range (2 ÷ 150 mGy/s)
 dose range (0.5 mGy - 180 mGy)
 sensitivity (5%)
 overall accuracy (±10%)
 tolerance to environmental
variation & stability
Prototype qualification
Conceptual design of the prototype tested @ PTW – secondary standard lab for
dosimetry:
Scintillator
(tile or fiber)
Ligth guide
Ø 1 mm
clear fiber
FC connector &
SiPM
Electronics
Trace plot: typical mammo SiPM output
PHISYCAL OBSERVABLE:
“buffered” signal sum
Sum of samples signals selected
by an edge detector algorithm
+ left & right buffer
(continuous photons flux – pulse duration of each
sample100 ms)
BUFFERED
SUM
 proportional to the DOSE
~TIME
summary of the results
Two different set-up (optimized for dynamic range & λ):
 1mm scintillator tile
 Blue scintillator fiber
coupled with MPPC (400 cells, 1x1 mm2)
Irradiation: 0,22 ÷ 217 mGy/s
Fiber +
Hamamatsu
MPPC
Precision(%)
2.31 ± 0.03
SensitivityA (mGy/s)
2.05 ± 0.03
MDSB (mGy/s)
0.458 ± 0.007
Linear Dinamic range
(mGy/s)
> 200
ASensitivity:
BMDS:
Precision / system gain
minimum signal distinguishable from the noise
3.6 Time-of-flight Positron Emission Tomography
(TOF-PET) & LIDAR
M. Conti, Physica Medica (2009) 25, 1-11
No indication about
the position of the
β+ along the Line
Of Response (LOR)
THE functional imaging tool, based on the
detection of pair of γ rays emitted backto-back by the annihilation of the β+
emitted by the 18F, chemically bound to
FDG
Identify the position of the β+
along LOR by the differenc ein the
time of arrival of the photons
TOF-PET is a HOT topic: 1510 papers in 2008-2013 (Google
Scholar) + significant investments by funding agencies &
companies
The gain in the image quality between a conventional and a TOF-PET
system may be quantified as [ Conti]:
• D = volume being inspected
• c = speed of light
• CTR – Coinicidence Time resolution
 Current machines
 TARGET
[S. Gundacker et al., NIM A 737 (2014) 92–100]
SiPM do play a role, since the timing resolution of a sensor my roughly be
written as:
s t = (output signal fluctuations) / (signal slope)trigger
Exemplary illustration of results obtained with the HAMAMATSU SiPM [R.
Vinke et al. NIM A 610 (2009) 188–191 ]:
Time resolution
wrt a laser shot
(worth thinking
of it for LIDAR)
[1x1 mm2, 50 micron cell pitch]
Timing properties of the sensor are not the full story and the
scintillator does play a role [S. Gundacker et al., NIM A 737 (2014) 92–100]:
mm2,
•[3x3
50 micron cell pitch]
• 2x2x10 mm3 LSO crystal
Actual resolution, accounting as well
for the Photon Travel Spread (PTS)
resulting by the point-of-interaction
and scintillation light time spread
Actually, HAMAMATSU is well aware of the relevance of crystals:
Small animal PET/CT scanning is also a significant market
(valued $790 million in 2012, and estimated to grow at an annual
growth rate of 14.5% over the next five years)
 The price for different smallanimal PET systems ranges
between $400,000 and
$1,200,000, depending on the
PET system configuration
 No. of crystals/scanner:~ 30000
JOURNAL OF NUCLEAR MEDICINE
TECHNOLOGY • Vol. 40 • No. 3 •
September 2012
*
• based on a single pair of detectors
(LYSO + SiPM)
• detectors mounted on rotating
structure with 2 degrees of freedom,
allowing reconstruction of source
position
• axial FOV: small animals (mice/rats)
• system geometry removes parallax
errors, eliminating the need of DOI
measurement
* Patent pending
(University of Aveiro)
• allows highly granular detector
assemblies for enhanced
performance
easyPET provides a very cost-effective solution for entry level
systems, due to the extreme reduction in the nr. of detectors and
complexity of the overall apparatus
• 2D prototype designed,
engineered and
commissioned
• Arduino UNO
microcontroller
• MATLAB interface: control
and online imaging
• Two 22Na sources, 5 µCi
• 2.7 mm Ø, 9 mm apart
• forward projection
(no filtered reconstruction)
Licensing under way for didactic/educational purposes
• position resolution
< 1.5 mm FWHM,
uniform over the whole
FOV
• Phantom with 18FDG
• 5 and 2 mm Ø, 2 mm
apart
• < 3 µCi
• forward projection
(with average filter)
3.7 Bio-imaging (last but not least), e.g. flow cytometry
Techshot Inc & Purdue University
4. Conclusions
What is SMART in a SiPM?
Not yet the end of the story…. Please turn the page one more time…
The live show is based on a modular kit co-developed by my team
and CAEN s.p.a.
More info typing in Google “CAEN SiPM kit”.
The second link is addressing the info on the company web site.
Guess where is the first link taking you?