Single-photon detectors in micro-electronics technology part...
Transcript of Single-photon detectors in micro-electronics technology part...
Single-photon detectorsin micro-electronics technology – part 2
Claudio PiemonteChief Scientist, Fondazione Bruno [email protected]
Claudio Piemonte, FBK 1
Low-level light detectiontechnologies
Claudio Piemonte, FBK 2
The problem: processing of extremely weak signals
Current signal:
1 pair/photon
Detector noise:
fluctuation of leakage
current
Electronics noise:
shot + thermal noise
Need of a detector with internal amplification
to reduce the impact of electronic noise!
Dominating
noise
Claudio Piemonte, FBK 3
Recap
Solid-state photodetectorswith internal gain.
Claudio Piemonte, FBK 4
Photodiode
Detectors with gain
Claudio Piemonte, FBK 5
n+
p+
p
-
+ Efield
multiplication
region
drift
region
p
n+
p+
p
-
+
p
APD
PDAPD
1 photo-carrierImpact ionization:
- linear mode: gain M
- Geiger mode: avalanche
Turn-off
probability
GM-APD/SPAD: model
Claudio Piemonte, FBK 6
CD
RS
VBD
RQVBIASVD
SPAD
n+
p+
p
-
+
multiplication
region
drift
region
p
t
ID
t0
t2
t
IINT
~(VBIAS-VBD)/RQ
t1
~(VBIAS-VBD)/RQ
t0
Photon
detection:
Avalanche
triggering
probability
IINTID
P10 – Turn-off probability
Claudio Piemonte, FBK 7
t
IINT
t1t0
t
IINT
t1t0
good quenching (high Rq)
bad quenching (medium Rq)
Fluctuation
(VBIAS-VBD)/RQ < 20uA
(VBIAS-VBD)/RQ
t
IINTno quenching (low Rq)
(VBIAS-VBD)/RQ
CD
RS
VBD
RQVBIASVD
SPAD
IINTID
fast, efficient!
RQ>~300kohm
time
Current
2 photons1 photon
The SPAD response is not proportional to light
intensity (in flashes).
1 photon 2 photons
avalanche
propagates
all over the
diode
+-
+- +
-
Claudio Piemonte, FBK 8
SiPM concept
Array of tiny
independent
SPADs can
provide a
proportional
information.
Claudio Piemonte, FBK 9
SPADs are connected in
parallel.
Output analog signal is
proportional to # of photons.
Very simple technology, fully
custom, optimized.
Analog SiPM
Signal digitized at SPAD level.
Integrated Digital architecture
produces information on # of
photons and time stamp.
Deep sub-mm CMOS technology
required.
time
current
1pe
2pe
3pe
Digital SiPM
Claudio Piemonte, FBK 10
SPADs/SiPMs: when, where?1960
Theory of Geiger-mode discharge in pn junctionsHaitz, McIntyre…
1970
Idea of SPAD exploitingG-M theory Various sources
1980
First SPADsPoliMIEG&G, Canada
Some papers
from 80-90s:
Claudio Piemonte, FBK 11
SPAD/SiPMs: when, where?1995
Analog SiPMconceptRussian researchers
~2005
First SiPM deviceswith good performance
2000+
SPAD & SiPMdevelopmentsin custom and standard CMOSfew research centers
2010+
Growing industrial applications
Hamamatsu
FBK
SensL
Ketek
Philips Digital
U. Edinburgh
PoliMI
DELFT
Excelitas
PrincetonLightwave
grey = CMOS
red = custom Si
blue = custom InGaAs
Some of the main actors in SPADs/SiPMs:
ST
Claudio Piemonte, FBK 12
…from SPAD to SiPM
New issues:
- SPAD fill factor limits the efficiency
- Interactions between SPADs
- Large area
SPAD1 SPAD3
active area
SPAD2
silicon
high Ef region
active area
high Ef region
Claudio Piemonte, FBK 13
Analog SiPM
SEM picture of a 1x1mm2 SiPMmetal quenching resistor
active area
Simple technology:
• ~10 lithographic layers layers
• passive quenching
• «easily» customizible to application
Claudio Piemonte, FBK 14
Analog SiPM
high-field region
(efficiency, noise)
guard struct.
(fill factor)
isolation
struct.
(fill factor,
cross-talk)
Silicon
drift region
(operation
voltage
stability)
silicon
properties
Technology can be optimized in
all aspects to get an optimal SPAD.
Claudio Piemonte, FBK 15
Digital SiPM
Main Features:
• standard CMOS technology
• integration of «intelligence» in the sensor
• switch off noisy SPADs
• distributed TDCs for better timing
• smart validation systems
• …
• architecture is usually customized to the application
• SPAD sensor is sub-optimal
Claudio Piemonte, FBK 16
Digital SiPM (for PET) - examplesPhilips DPC
Spadnet D-SiPMdesigned by FBK
produced by ST
Claudio Piemonte, FBK 17
• STM SPAD-based TOF Proximity Sensor
From high-end applications to consumer
@Chipworks (http://ww2.chipworks.com/e/4202/6180-Time-of-Flight-Sensor-pdf/hwvfs/713665047)
Claudio Piemonte, FBK 18
Claudio Piemonte, FBK 19
Main characteristics of a SiPM
Parameters of a SiPM
1) Gain
Number of electrons produced per detected photon
2) Primary Noise (Dark Count Rate [Hz])
Cells firing spontaneously
3) Correlated Noise excess noise factor (ENF)
after-pulsing, optical cross-talk
4) Photo-detection efficiency (PDE)
5) Single Photon Time Resolution (SPTR)
Claudio Piemonte, FBK 20
Description of each parameter with reference to an FBK technology:
NUV-HD
Claudio Piemonte, FBK 21
Original technology
FBK SiPM technology roadmap
Electric fieldengineering
New cell border(trenches)
RGBNUV
RGB-HDNUV-HD
NUV-HD optimization
RGB-UHD
2005
2010
2012
2015
2012
VUV-HD NIR
Ongoing developments
Claudio Piemonte, FBK 22
• Narrow dead border region → High Fill Factor
• Trenches between cells → Low Cross-Talk
• Make it simple: 9 lithographic steps
HD technology
(C. Piemonte et. al., (2016) IEEE T. Electr. Dev., 10.1109/TED.2016.2516641)
< 2 µm
High-field region
Poly
resistor
Metal
SPAD of SiPM
Trench
< 1 µm
Claudio Piemonte, FBK 23
SPAD Pitch 15 µm 20 µm
Fill Factor (%) 55 66
SPAD/mm2 4444 2500
HD: layout features
25 µm 30 µm 35 µm 40 µm
73 77 81 83
1600 1111 816 625
High Dynamic Range High PDE
SPAD size active area
NUV
SiPM NUV-HD
SPAD pitch (mm)
Claudio Piemonte, FBK 24
Gain = IMAX*tQ = (VBIAS-VBD)*tQ = (VBIAS-VBD)*CD__ ________ __ ____________
q RQ q q
t
i~(VBIAS-VBD)/RQ
exp(-t/RQ*CSPAD)
Gain of a SPAD/SiPM
Gain = number of electrons per photo-carrier generation
Uniformity and stability of VBD is crucial!
CD
RS
VBD
RQVBIASVD
DIODE
Claudio Piemonte, FBK 25
26
SiPM Signal – NO amplifier
0.E+00
1.E-03
2.E-03
3.E-03
4.E-03
5.E-03
6.E-03
7.E-03
0.0E+00 1.0E-07 2.0E-07 3.0E-07 4.0E-07
Time (s)
Am
pli
tud
e (
V)
Thanks to the large gain it is possible to connect the SiPM
directly to the scope
VBIAS
SiPM
50W
Digital
Scope
SiPM: 1x1mm2
single SPAD fired
two SPADs fired
recovery
fast
component
Gain of a NUV-HD SiPMT
ime
co
nsta
nt (n
s) 120
100
80
60
40
20
Ga
in (
mill
ion
s)
6
5
4
3
2
1
t
i
exp(-t/RQ*CSPAD)
t
i
Over-voltage (V)
Over-voltage (V)Claudio Piemonte, FBK 27
SiPM photon counting capability
gain uniformity in SPAD and SPAD-to-SPAD
# of detected photons
T = 20 C
SiPM area 1x1mm2
Histogram of the signal area
at 4 illumination levels
Claudio Piemonte, FBK 28
SiPM illuminated
with weak pulses
VBD Uniformity
6” Silicon wafer
1532 3x3mm2 SiPMs per wafer
SPAD pitch is 35mm
Each SiPM has 6295 SPADs
Reverse IV measurement on all
SiPMs on 3 sample wafers:
Breakdown voltage
Gain uniformity (intra- and inter- device) is mostly
affected by Vbd. What is the Vbd uniformity?
Claudio Piemonte, FBK 29
VBD Uniformity
Distribution of VBD on
three wafers.
Max variation 200mV!
Claudio Piemonte, FBK 30
VBD Temperature dependenceThe mean free path of the carriers in the high-field region
increases with decreasing temperature.
6 V
5 V
33mV/C
25mV/C
DOI: 10.1109/TED.2016.2641586Claudio Piemonte, FBK 31
Dark count rate
Thermal generation
(SRH model)
Tunneling generation
Main source of
carrier generation
in silicon at room T in
standard photodiodes
TAT is an important additional
contribution in devices relying
on impact ionization
Claudio Piemonte, FBK 32
DCR measurement
(C. Piemonte et. al., (2012) IEEE NSS/MIC Conf. Rec., N1-206)
Claudio Piemonte, FBK 33
DCR measurement
Inter-arrival time histogram
Scatter plot
DCR
Scatter plot
of amplitude vs
inter-times
DCRDCR event
distribution
with Poisson fit
Claudio Piemonte, FBK 34
T=70K
(example to separate
well the components)
DCR / mm2 vs. Temperature
Standard field
Low-field
> 7 orders of
magnitude !
NUV-HD 4x4mm2 size 25mm SPAD pitch (25600 SPADs)
Claudio Piemonte, FBK 35
-0.35
-0.3
-0.25
-0.2
-0.15
-0.1
-0.05
0
0.05
-1.0E-08 1.0E-08 3.0E-08 5.0E-08 7.0E-08
Time (s)
Vo
lta
ge
(V
)
Collection of primary Events with after-pulse
The amplitude of the after-pulse
increases as the cell recovers to
its opertional condition
After-pulsing Probability
Claudio Piemonte, FBK 36
Carriers may be trapped during an avalanche,
released and may triggering another avalanche.
AP measurement
Inter-arrival time histogram
Scatter plot
AP
Scatter plot
of amplitude vs
inter-times
AP
Claudio Piemonte, FBK 37
AP in NUV-HD
Low-fieldStandard field
AP increases with decreasing temperature because the
release becomes slower..
Claudio Piemonte, FBK 38
3x10-5 photons with energy higher than 1.14eV emitted per carrier
crossing the junction.
[A. Lacaita et al., IEEE TED, vol. 40, n. 3, 1993]
cell1 cell2
Optical cross-talk
Photons are emitted during an avalanche which may
trigger another avalanche in another SPAD
Claudio Piemonte, FBK 39
substrate
activelayer
-+
external cross-talk
avalanche
--+
delayed cross-talk+
scintillator
-+direct cross-talk
high-field region
Optical cross-talk
Claudio Piemonte, FBK 40
0.E+00
1.E-03
2.E-03
3.E-03
4.E-03
5.E-03
6.E-03
7.E-03
0.0E+00 1.0E-07 2.0E-07 3.0E-07 4.0E-07
Time (s)
Am
pli
tud
e (
V)
single SPAD fired
two SPAD fired,
likely CT
CT measurement
Inter-arrival time histogram
Scatter plot
Di-CT
De-CT
Amplitude histogram
De-CT
Claudio Piemonte, FBK 41
CT in NUV-HD
Direct cross-talk Delayed events
25-30-35mm SPAD size 25-30-35mm SPAD size
Claudio Piemonte, FBK 42
Photo-detection efficiency
2016 JINST 11 P11010
Claudio Piemonte, FBK 43
SPAD
Fill Factor
Quantum
Efficiency
Triggering
Probability
QE
It depends on:
- Anti-reflective coating
- active layer thickness
- junction depth
no detection detection
Claudio Piemonte, FBK 44
NUV-HD: QEQE Measured on a photodiode produced with SiPMs
+
Simulation of the anti-reflective coating
simulation of the anti-reflective coating
measured QE
Claudio Piemonte, FBK 45
NUV-HD: QE*Pt
SPAD size is defined by
metal opening which is
within the high-field region
cSPAD (100% Fill Factor)
Claudio Piemonte, FBK 46
SPAD Pitch 15 µm 20 µm
Fill Factor (%) 55 66
SPAD/mm2 4444 2500
NUV-HD: PDE
25 µm 30 µm 35 µm 40 µm
73 77 81 83
1600 1111 816 625
35mm cell pitch
Claudio Piemonte, FBK 47
Statistical Fluctuations in the current growth:
1. Photo-conversion depth
2. Vertical Build-up at the very beginning of the avalanche
3. Lateral Propagation
t=0 pair generation
0<t>t1 drift to the high-field region
t>t1 avalanche multiplication
single carrier
current level
Intrinsic Time jitter
Claudio Piemonte, FBK 48
Single-photon time resolution (SPTR)
Claudio Piemonte, FBK 49
• 2-ps width pulsed
laser (425nm)
Measurement setup for SPTR
•Set-up time
resolution ~ 10ps FWHM
Claudio Piemonte, FBK 50
Laser
F. Acerbi, et al IEEE TNS vol. 61, n. 5, 2014 , pp. 2678 - 2686
• SPAD (50x50um2)
• 1x1mm2 SiPM of 50x50um2
• 3x3mm2 SiPM of 50x50um2
Example of SPTR
Claudio Piemonte, FBK 51
Noise contribution to SPTR
Signal shape
𝜎𝒕𝜎𝑎𝑓𝑡ℎ′
Is the electronic noise the origin of SPTR deterioration
moving to large areas?
Claudio Piemonte, FBK 52
SPTR: ultimate limit?
IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 20, NO. 6,
NOVEMBER/DECEMBER 2014
Claudio Piemonte, FBK 53
We designed a small
SPAD with masked
edges.
Metal ring all-around active area
Better signal extraction
more uniform signal shape
SPTR: analysis at a single SPAD level
uncovered edge
uncovered edge
covered edge
uncovered edge
Claudio Piemonte, FBK 54
20 ps FWHM!!!
Main numbers in one table
ParameterNUV-HD technology 1mm2 40mm SPAD
DCR (20C) <100 kHz
Di-CT <20%
De-CT <5%
AP (20C) <5%
PDE 60% (400nm)
SPTR 80 ps FWHM
Claudio Piemonte, FBK 55
PMT vs SiPM for NUV light
compactness ↓ ↑
robustness ↓ ↑
sensitivity to magnetic fields
↓ ↑
low operating voltage ↓ ↑
PDE ↑ ↑
DCR ↑ ↓
dynamic range ↑ ↓
SPTR ↑ ↑
cost ↑ ~↑
market competition ↓ ↑Claudio Piemonte, FBK 56
Hot topics in SIPM/SPAD
• High Dynamic Range
• NIR sensitivity
FBK is working on those aspects.
UHD technology is an example.
Claudio Piemonte, FBK 57
UHD Technology
Reduction of all the feature sizes
Circular active area (smallest cells)• No corners (with lower field)
• Hexagonal cells arranged in
honeycomb configuration
D = 0.5 D = 0.75
D = 1 D = 1.25
L < 1um
Claudio Piemonte, FBK 58
New design
RGB-UHD
12.5mm
10mm
7.5mm
HD technology: small cells
L
RGB
RGB-HD
20mm
25mm
15mm
30mm
12mm
cell pitch (mm) cells/mm2
12 7000
15 4500
20 2500
25 1600
30 1100
cell pitch (mm) cells/mm2
7.5 20530
10 11550
12 7400
RGB-HD
RGB-UHD
Claudio Piemonte, FBK 59
Comparison to other technologies
40 um cell
RGBSiPMs
25 um cell
RGB-HDSiPMs
<7.5 um cell
RGB-UHDSiPMs
Claudio Piemonte, FBK 60
…new design is not enough…
F.Acerbi et al., presented at IEEE NSS, Strasbourg 2016Claudio Piemonte, FBK 61
Border region plays an important role for small cells.
We have to develop a new guard ring (NGR) especially
for the smallest SPAD (5um)
SEM images
Microcells separated by
trenches (picture taken during
processing)
Finished SiPM
(10 um)
Claudio Piemonte, FBK 62
Oscilloscope waveforms
7.5 um cell 10 um cell
The 5um SPAD does not work with the standard guard
ring while it works with NGR!!
Claudio Piemonte, FBK 63
RGB-UHD: PDE
• Optimization of the technology
• Development of even smaller cell size
5 mm SPAD pitch
46200 cells / mm2
Measurement at CERN
by Y. Musienko
Claudio Piemonte, FBK 64
RGB-UHD: Signal
Claudio Piemonte, FBK 65
66
Thanks to all collaborators
Fabio Acerbi
Giacomo Borghi
Gabriele Faes
Nicola Furlan
Alberto Gola
Marco Marcante
Stefano Merzi
Claudio Piemonte
Giovanni Paternoster
Antonino Picciotto
Veronica Regazzoni
Nicola Zorzi