Content uploaded by Piera Maccagnani
Author content
All content in this area was uploaded by Piera Maccagnani on Jul 25, 2014
Content may be subject to copyright.
Large-area low-jitter silicon single photon avalanche diodes
Massimo Ghioni
*
a, b
, Angelo Gulinatti
a
, Ivan Rech
a
, Piera Maccagnani
c
, and Sergio Cova
a, b
a
Politecnico di Milano, Dipartimento di Elettronica e Informazione, Piazza Leonardo da Vinci 32 -
20133 Milano, Italy
b
MPD Micro-Photon-Devices, via Stradivari 4 – 39100 Bolzano, Italy
c
IMM-CNR sezione di Bologna, Via Piero Gobetti, 101 – 40129 Bologna, Italy
ABSTRACT
Single photon counting (SPC) and time correlated single photon counting (TCSPC) techniques have been developed in
the past four decades relying on photomultiplier tubes (PMT), but interesting alternatives are nowadays provided by
solid-state single photon detectors. In particular, silicon Single Photon Avalanche Diodes (SPAD) fabricated in planar
technology join the typical advantages of microelectronic devices (small size, ruggedness, low operating voltage and low
power dissipation, etc.) with remarkable basic performance, such as high photon detection efficiency over a broad
spectral range up to 1 µm wavelength, low dark count rate and photon timing jitter of a few tens of picoseconds. In
recent years detector modules employing planar SPAD devices with diameter up to 50 µm have become commercially
available. SPADs with larger active areas would greatly simplify the design of optical coupling systems, thus making
these devices more competitive in a broader range of applications. By exploiting an improved SPAD technology, we
have fabricated planar devices with diameter of 200 µm having low dark count rate (1500 c/s typical @ -25 °C). A
photon timing jitter of 35 ps FWHM is obtained at room temperature by using a special pulse pick-up network for
processing the avalanche current. The state-of-the-art of large-area SPADs will be reviewed and prospects of further
progress will be discussed pointing out the challenging issues that must be faced in the design and technology of SPAD
devices and associated quenching and timing circuits.
Keywords: single photon avalanche diodes, SPAD, time-correlated single photon counting, TCSPC, photon timing
1. INTRODUCTION
There is nowadays a widespread and steadily growing interest in single photon detectors, driven by the need for ultimate
sensitivity in various scientific and industrial applications such as fluorescence spectroscopy in life and material
sciences, quantum cryptography and computing, profilometry of remote objects with optical radar techniques, particle
sizing etc. In particular, the use of fluorescence lifetime spectroscopy as both an analytical and research tool has
increased markedly in recent years.
To date fluorescence lifetime spectroscopy has found remarkable applications in
chemistry, biochemistry and
biology (for a review, see for example [1]). Most biologically relevant fluorophores exhibit
characteristic decay times ranging from picoseconds to nanoseconds. The
time-correlated single-photon counting (TCSPC)
technique [2, 3] is currently used for directly measuring fluorescence decays.
TCSPC was developed relying on
photomultiplier tubes (PMTs), that is, vacuum tube detectors with high internal gain. High performance PMTs have been
produced industrially with sophisticated technologies since the 60’s; amongst their advantages, the most valuable one is
the wide sensitive area (∼ cm
2
), which in some cases greatly simplifies the design of the optical system. PMTs also attain
remarkable performance at high counting rate and offer picosecond timing resolution with micro channel plate (MCP)
models. However, they suffer from low quantum efficiencies (QE) in the visible range: the QE of conventional bialkali
and multialkali photocathodes reaches 20-25% between 400 and 500 nm [4].
*
ghioni@elet.polimi.it, phone +39-02-23996093, fax: +39-02-2367604
Invited Paper
Quantum Sensing and Nanophotonic Devices V, edited by Rengarajan Sudharsanan, Christopher Jelen,
Proc. of SPIE Vol. 6900, 69001D, (2008) · 0277-786X/08/$18 · doi: 10.1117/12.761578
Proc. of SPIE Vol. 6900 69001D-1
2008 SPIE Digital Library -- Subscriber Archive Copy
In more recent years, single photon avalanche diodes (SPADs) have emerged as a solid state alternative to PMTs [5-8].
Besides the well known advantages of solid state versus vacuum tube devices (small size, ruggedness, low power
dissipation, low supply voltage, high reliability, etc.), SPADs provide inherently higher quantum efficiency, particularly
in the red and near infrared spectral regions. The small active area of SPAD devices is often considered a disadvantage
compared to PMTs. However, this is not true in a variety of TCSPC techniques relying on descanned confocal or near
field microscopy such as fluorescence lifetime imaging (FLIM) and time-resolved fluorescence correlation spectroscopy
[1, 3]. In these applications, the spot size at the microscope image plane can easily be made underfill the SPAD active
area, provided that the diameter is sufficiently large (≥ 50 µm). In confocal microscopy, the SPAD device can also be
used as confocal pinhole, thus simplifying the whole optical system and increasing the miniaturization potential.
Although active area diameters between 50 and 100 µm are acceptable in the aforementioned applications, larger
diameters (> 100 µm) would be highly desirable for attaining good photon collection efficiency without requiring
complex and time-consuming optical alignment and focusing procedures. Furthermore, fiber pigtailing of the detector,
often employed for having a more flexible optical system, would be simpler and more efficient for SPAD detectors with
large area.
Unfortunately, large active area, low dark counting rate, high quantum detection efficiency and small photon timing jitter
are conflicting requirements for a SPAD device [9].
Single photon counting modules based on SPAD diodes with thick depleted region (20-25 µm) and large active area (180
µm diameter) are commercially available from PerkinElmer Optoelectronics [10]: they have very good photon detection
efficiency (PDE) and low dark counting rate (DCR), but they have a moderate photon timing resolution of typically 350
ps full-width at half maximum (FWHM). Such a performance is unsuitable for tissue autofluorescence measurements
[11] or protein conformational dynamic analysis [12], where fluorescence components as short as 100 ps or less are often
met.
SPAD devices with thin depleted region (1-2 µm), fabricated in a standard CMOS silicon technology can attain time
resolution down to less than 50 ps FWHM at room temperature [13, 14]. This performance is comparable with that of the
best MCP-PMTs. Standard CMOS technology makes it possible to monolithically integrate the SPAD device and its
associate quenching circuit, taking great advantage of reduced size and low parasitic capacitances. Unfortunately, CMOS
technologies available in industrial silicon foundries are not optimized for SPAD device fabrication, resulting in reduced
photon detection efficiency and relatively high dark counting rate. So far, no CMOS-based SPAD devices with active
area diameter > 100 µm have been reported.
To overcome these limitations and provide a solid-state alternative to MCP-PMTs in demanding TCSPC applications we
have developed an improved planar silicon process that enables the fabrication of high-performance SPAD devices with
active area diameter up to 200 µm.
2. DEVICE TECHNOLOGY ISSUES
It is well known that semiconductor detectors offer better photon detection efficiency (PDE) than vacuum tube detectors
since electron emission phenomenon is not involved and the primary photocurrent flows directly within the
semiconductor. However, the situation is just opposite from the standpoint of the detector noise, that is, of the detector
dark current or dark count rate (DCR). Vacuum tube detectors have dark current per unit of sensitive area inherently
lower than semiconductor detectors. For instance, S20 photocathodes are currently obtained with thermal electron
emission lower than 1000 electrons s
-1
cm
-2
at room temperature. On the other hand, the dark current in a reverse biased
silicon junction operating at room temperature or below is dominated by thermal generation of carriers in the depletion
layer and tunneling processes [15]. The total generation rate G per unit volume is given by:
G = G
th
+ G
TAT
+ G
bbt
(1)
where G
th
is the contribution due to thermal generation via localized deep energy levels (also called traps), G
TAT
is the
trap-assisted tunneling contribution and G
bbt
is the band-to-band tunneling contribution [15-17]. It has been shown
theoretically and experimentally demonstrated that G
bbt
becomes the dominant contribution only for electric field
strengths higher than 7⋅10
5
V/cm [16, 17]. We will show in the next section that this contribution can be made negligible
by properly reducing the peak electric field within the active area of the SPAD device. In this condition, the total
generation rate per unit volume can be written as:
pn
pn
TTATth
ee
ee
NGGG
+
=+≈
(2)
Proc. of SPIE Vol. 6900 69001D-2
E - Et
E
Pure
tunneling
where N
T
is the volume density of deep levels, e
n
and e
p
are the probabilities per unit of time for the emission of an
electron or a hole respectively. Both the quality of the starting material and the technological processes used in the
device fabrication have a strong impact on N
T
and therefore on the generation rate G. Transition metal impurities are the
most common source of deep levels. Metal contamination may occur during silicon handling, high-temperature heat
treatments or ion implantations As unintentional contaminants Fe, Cu, Ti or Ni are usually found in silicon in
concentrations of ∼10
11
-
10
12
cm
−3
. On the other hand, the generation rate is strongly dependent on the electric field
acting on deep levels as well [16-19]. Three different mechanisms cause the increase of emission rates e
n
and e
p
from
deep levels under the applications of strong electric fields: the Poole-Frenkel effect, the phonon-assisted tunneling, and
the direct tunneling from deep level into conduction or valence bands (Fig. 1a, b).
Fig.1 Potential barrier for the emission of an electron from a deep energy level in external electric field for: a) charged impurities
and b) neutral impurities. The arrows show the field-enhanced emission processes.
Fig. 2 Schematic diagram showing the field-enhanced emission processes of electrons and holes from donor-like (a) and acceptor-
like (b) deep levels. Either electron emission from a donor-like level or a hole emission from an acceptor-like level is
enhanced by the Poole-Frenkel effect.
The Poole-Frenkel (PF) effect consists of the lowering of the potential barrier due to the electric field applied to a
semiconductor. This effect is present only when the deep impurity level behaves like a Coulombic well, i.e., the emission
of electrons or holes ionizes the impurity. Phonon-assisted tunneling and direct tunneling are possible for impurities in
all charge states. Hence, carrier emission rate can increase in an electric field due to carrier tunneling even in the case of
a neutral impurity (Dirac well). Whether a deep level behaves like a Coulombic or a Dirac well during the emission of a
carrier it depends on its charge state (Fig. 2a, b). In practice, if the deep level is acceptor-like, i.e. neutral when empty, it
a)
b)
a) b)
Proc. of SPIE Vol. 6900 69001D-3
Electric field (V/cm)
0
04
10
ci,
E
10
102
101
106
E
V
a)
F
2
0
0
a)
E
a)
0
a)
V
a)
1000
Li
0.6 1.2
x(pm)
1.4 1.6
behaves like a Coulombic well during the hole emission and like a Dirac well during electron emission. Conversely, if
the deep level is donor-like, i.e. neutral when filled, it behaves like a Coulombic well during electron emission and like a
Dirac well during hole emission. To evaluate the impact of the electric field on the generation rate, we have considered a
donor-like and an acceptor-like deep level with energy E
T
=E
C
-0.28 eV, which are associated to Ti and FeB impurities in
silicon respectively [20]. Since these traps are closer to the conduction band, the limiting mechanism to the generation
rate is hole emission. Therefore:
pT
eNG ≈
(3)
The field enhancement factor Γ for the generation rate is given by:
)0(e
)F(e
p
p
p
=Γ=Γ
(4)
where F is the electric field strength.
Fig. 3a shows the field enhancement factor Γ for the donor-like and for the acceptor-like deep levels as a function of the
electric field strength. The dependence of Γ on the position within the depletion layer of a typical SPAD junction is
shown in Fig. 3b, where the electric field profile is also shown for clarity. A remarkable Γ of 4000 is obtained for the
acceptor-like level with F ~ 6 10
5
V/cm, whereas the donor-like level shows a lower Γ of 70.
Fig. 3 Field enhancement factor Γ for a donor-like and an acceptor-like deep level with energy E
T
=E
C
-0.28 eV. a) Γ as a function
of the electric field strength. b) Γ as a function of the vertical position within the active region of a SPAD device.
Modern silicon technologies guarantee almost no structural defects and low metal contamination levels as well.
However, even trace levels of metal impurities that would be normally undetectable with state-of-the-art analytical
techniques might impair the DCR performance of SPAD detectors, due to the inherently high electric field within the
depletion layer.
Under the simplifying hypothesis of a single deep level impurity, the DCR expected from a SPAD can be calculated as
follows:
(
)
Aw
n
Aw
n
dx
)F()0(e)F()0(e
)0(e)0(e)F()F(
)0(e)0(e
)0(e)0(e
ANdx
)F(e)F(e
)F(e)F(e
ANDCR
eff,g
i
eff
0,g
i
x
x
T
ppnn
pnpn
pn
pn
T
x
x
T
pn
pn
T
p
n
p
n
τ
=Γ
τ
=
=
∫
η
Γ+Γ
+⋅Γ⋅Γ
+
=
∫
η
+
=
(5)
where
(
)
dx
)F()0(e)F()0(e
)0(e)0(e)F()F(
w
1
p
n
x
x
T
ppnn
pnpn
eff
∫
η
Γ+Γ
+⋅Γ⋅Γ
=Γ
(6)
Proc. of SPIE Vol. 6900 69001D-4
n++ shallow n-i--i-
isolation
H
5pm
I-
p- epilayer
p -I- buried layer
4-substrate
hv
50-200pn,
—Th
Anode
Cathode
and
eff
0,g
eff,g
Γ
τ
=τ
(7)
Γ
eff
is an effective field enhancement factor, n
i
= 1.45 10
10
cm
-3
is the intrinsic carrier concentration at room temperature,
A is the junction area, w is the depletion region width,
τ
g.0
is the low-field generation lifetime (tens of ms in high quality
silicon wafers [21]),
η
T
is the avalanche triggering probability [22], x
n
and x
p
are the depletion edges. Equation 7 shows
that the effective generation lifetime in SPAD detectors is lower than the low-field generation lifetime by a factor
Γ
eff
where, in practical cases, the value of
Γ
eff
may range from tens to thousands depending on the electric field strength and
on the trap position and kind. Simple quantitative evaluations clearly show that:
a) for developing SPADs with low DCR and fairly wide area it is necessary to have a high quality fabrication
technology leading to a consistent reduction of impurity concentration with respect to standard, high-quality silicon.
In addition, the electric field profile within the depletion region must be properly designed for ensuring a low value
of
Γ
eff
;
b) better SPAD design and fabrication technology enable a significant increase of the active area of SPAD detectors.
SPAD active area will be anyway much smaller than that of PMTs having comparable level of noise.
A couple of examples about the fabrication of planar SPAD with depletion layer width w = 1 µm can well illustrate these
points. If we target a DCR not exceeding 1000 c/s in a detector with diameter D= 50 µm, the effective generation
lifetime must be at least 25 ms. This is a good level of quality that has been already attained in the technology that we
have developed. Further improvement of the technology may be pursued, but the limitation to the maximum active area
compatible with a low DCR level will still be noteworthy. Let us assume to attain an extremely good level in the
fabrication technology, which ensures a very long lifetime
τ
g,eff
=1 s: even with such an outstanding technology, in order
to have DCR not exceeding 1000 c/s it is necessary to limit the maximum diameter of the sensitive area to D = 400 µm.
3. RECENT ADVANCES IN LARGE-AREA SPAD DETECTORS
Fig. 4 shows a schematic cross section of the SPAD structure. A double-epitaxy structure is used in order to reduce
the diffusion effects that adversely affect the time response of the detector [23]. The active n+p junction is built in the
upper low-doped p-epilayer. The buried p+ epilayer provides a low-resistance path to the side ohmic contact. A boron
implantation in the central part of the n+p junction defines the high electric field region, that is, the active area of the
detector. Deep, highly doped n+ regions connected to the bulk n-silicon guarantee electrical isolation between adjacent
SPADs.
In the last years we steadily improved the planar double-epitaxial SPAD technology [24], achieving a reliable fabrication
of high-performance SPAD devices with active area diameter up to 200 µm. The efforts were mainly directed to:
- exploit specific gettering processes, as phosphorus diffusion and p/p+ segregation gettering for removing metal
impurities from the detector active volume, therefore reducing DCR and afterpulsing effects;
Fig. 4 Cross section of the planar SPAD structure.
Proc. of SPIE Vol. 6900 69001D-5
- identify and remove or mitigate all the possible sources of contamination in the detector processing, with special
care to transition metal contamination;
- employ a lower electric field within the p-n junction depletion region to attenuate the band-to band-tunneling and
field-enhanced generation of carriers, thus making it possible to reduce more effectively the DCR by cooling the
detector. In order not to adversely affect the PDE, a reduction of the peak electric field must be compensated by a
suitable increase in the width of the electric field profile.
In the following the performance of the large-area SPAD devices recently fabricated will be reviewed, pointing out the
most significant improvements.
3.1. Photon Detection Efficiency
Fig. 5 shows the photon detection efficiency of a 200 µm SPAD device as a function of wavelength, measured at
different excess bias voltages, V
exc
. At V
exc
= 5 V the PDE has a peak of 52% at 550 nm and it is about 15% at 820 nm
wavelength. These figures are consistent with the total thickness of the epitaxial layer (about 5µm). Similar results have
been obtained with 50 and 100 µm SPAD devices. The PDE is quite uniform all over the wafer: small fluctuations were
observed (std deviation
∼5% of the average value), mainly due to the thickness fluctuations of the top oxide layer. By
increasing V
exc
to 10 V the peak PDE increases to a remarkable 68%. The PDE increases with V
exc
mainly due to the
increase of the avalanche triggering probability [22]. The small increase of the depletion region thickness gives a
negligible contribution. However, also the DCR increases at higher V
exc
, setting a trade-off that must be carefully
evaluated when seeking the maximum sensitivity.
If the detector noise is dominant, a suitable figure of merit for the sensitivity is the noise equivalent power (NEP),
which is defined as the signal power required to attain a unity signal-to-noise ratio within 1-s integration time:
DCR2
PDE
h
NEP
ν
= (8)
Fig. 6 shows the typical dependence of the NEP on V
exc
for a 50 µm diameter SPAD operating at room temperature. It is
clearly visible that the optimal value of V
exc
is around 4–5 V, where the NEP curves show a broad minimum. Conversely,
if the measurement is background or photon noise limited the optimal V
exc
should be sufficiently high to push the PDE
close to saturation.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
400 500 600 700 800 900 1000
Wavelength (nm)
Photon Detection Efficiency
10 V
7 V
5 V
Excess Bias Voltage
Fig. 5 Photon detection efficiency of a 200 µm SPAD device as a function of wavelength. Measurements were performed at
different excess bias voltages.
Proc. of SPIE Vol. 6900 69001D-6
123456789
Excess bias voltage (V)
Noise Equivalent Power (W Hz )
-1/2
10
-15
10
-16
10
-17
λ
= 850 nm
λ
= 550 nm
Fig. 6 NEP of a 50 µm diameter SPAD as a function of the excess bias voltage. Measurements were performed at room
temperature.
3.2. Dark count rate
Fig. 7 shows the DCR as a function of temperature for SPAD devices having different active area diameters. The
breakdown voltage varies between 33.8 V and 36 V over the temperature range from -50 °C to 20 °C. Measurements
were performed at a constant excess bias voltage of 5 V. The DCR decreases almost exponentially with temperature: at a
relatively low temperature of –25 °C (easily obtainable with a thermoelectric cooler) the typical DCR is 5, 50 and
1500c/s for the 50, 100 and 200 µm SPADs.
0.1
1
10
100
1000
10000
100000
-50 -40 -30 -20 -10 0 10 20
Temperature (°C)
Dark count rate (c/s)
200 µm
50 µm
100 µm
Fig. 7 Dark count rate as a function of temperature for SPAD devices having different active area diameters.
The advantage of an engineered electric field profile is clearly visible in Fig. 8, where the DCR of two 50 µm SPAD
devices made by using different processes are plotted. The DCR of the SPAD device with “standard” electric field [25]
Proc. of SPIE Vol. 6900 69001D-7
can be accurately fitted by taking into account band-to-band tunneling (dotted line) and field-enhanced Shockley-Read-
Hall generation (dashed line): the weights of these two contributions become equivalent at a temperature of about -5 °C
(corner temperature). Since the dependence of tunneling on temperature is relatively weak, just a minor advantage is
gained by cooling the device well below the corner temperature. Conversely, for the SPAD device with “engineered”
electric field no evidence of the onset of tunneling is observed down to -50 °C: cooling can thus be exploited for
obtaining a strong reduction of the DCR. Furthermore, the DCR of devices with engineered electric field profile is lower
also at room temperature, due to the reduced weight of the field-enhanced generation.
0.1
1
10
100
1000
10000
-80 -60 -40 -20 0 20
Temperature (°C)
Dark count rate (c/s)
SPAD with "standard" electric field
SPAD with "engineered" electric field
Fig. 8 Dark count rate as a function of temperature for 50 µm SPAD devices with “standard” and “engineered” electric field profile.
DCR uniformity was assessed by testing about 2000 SPAD devices all over the wafer at room temperature. The
percent point distribution function (i.e. the inverse of the cumulative distribution function) of the DCR is shown in
Fig. 9a for SPAD devices having different active area diameter. It must be noted that the distribution curves for the 100
and 200 µm devices are similar, whereas the DCR distribution for the 50 µm shows a relatively flat region (about 35% of
the 50 µm SPAD devices show a DCR between 1 and 2 kc/s) followed by a steep increase. Fig. 9b shows the DCR
distribution of the most recent SPAD generation. Most of the sources of metal contamination in the fabrication process
were identified and removed or mitigated. As a result, the DCR was reduced by a factor of 5 on the average. It must be
noted that 50% of the 50 µm SPAD devices show a DCR below 500 c/s.
0 20406080100
% of devices with DCR < N
DL
N
DL
: limit to DCR (c/s)
200 µm
100 µm
50 µm
10
10
10
10
10
2
3
4
5
6
Fig. 9 Percentage of SPAD devices (horizontal scale) found within a given limit of the individual dark counting rate (vertical
scale) measured at room temperature. a) previous SPAD generation; b) improved SPAD generation with reduced metal
contamination.
0 20406080100
% of devices with DCR < N
DL
N
DL
: limit to DCR (c/s)
200 µm
100 µm
50 µm
10
10
10
10
10
2
3
4
5
6
a) b)
Proc. of SPIE Vol. 6900 69001D-8
Afterpulsing Probability Density (ns 1)
0 0 0
& 61
0 0 0
C"
0
0
0
0
0
-I
)
01
0
C)
I')
-I
01
0
C)
-U
a
1*.
0)
C"
0
0
Ni
0
0
0
Data shown in Fig. 8 and 9a are consistent with the presence of two types of defects within the depletion region of
the junction [24]. Defects of the first type (type A) have an energy level close to the midgap and a density of about 4
10
8
cm
-3
. The probability of no occurrence of such defects is 40% for a 50 µm device, 3% for a 100 µm device and
practically negligible for a 200 µm device. This observation may justify the relatively flat region in the lower part of the
distribution curve in Fig. 9 for the 50 µm devices. Defects of the second type (type B) have a greater density (between
10
9
and 10
10
cm
-3
), so that they may occur with high probability also in smaller devices. However, their energy level is
displaced by about Eg/4 from the midgap. In general, good devices have no type A defects and their DCR decreases very
steeply by lowering the temperature. Devices with dominant type A defects show a DCR that decreases with a smaller
slope, whereas devices with a mix of non-dominant type A and type B defects show a DCR that decreases with
intermediate slope [24].
3.3. Afterpulsing
Afterpulsing measurements were performed at different temperatures by using the time-correlated carrier counting
(TCCC) technique [26]. TCCC essentially consists in: a) filling the deep levels with a pulsed stimulus; b) measuring the
time interval from the filling pulse to the detection of a released carrier; and c) repeating the procedure for collecting a
histogram of the carrier emissions versus time. Deep levels can be populated by current or light pulses. We adopted an
avalanche current pulse, which has the advantage of filling only the traps involved in the normal device operation. Note
that only electrons released in the p-side and holes released in the n-side cross the high-field region, thus being able to
trigger the avalanche. Moreover, during the avalanche, minority carriers are present only where impact ionizations occur.
It follows that TCCC with avalanche filling stimuli is inherently sensitive only to minority-carrier traps in the high-field
region.
Fig. 10 shows the probability density in time for the occurrence of an afterpulse after an initial avalanche pulse for a
200 µm SPAD device operated at 5 V excess bias voltage. An active quenching circuit (AQC) with a dead time of 80 ns
was used in the experiments. It is worth noting that the probability of having an afterpulse quickly decays, being
negligible after about 1 µs from the initial avalanche pulse. The total afterpulsing probability is typically 2% at room
temperature and it increases up to 6% at -25 °C. This increase is due to the dependence of the trap emission lifetime on
temperature and to the presence of a fixed dead time. At lower temperatures, the emission lifetime of a given trap gets
longer. Accordingly, the probability that a carrier is emitted after the dead time (thus being able to trigger an avalanche)
gets higher.
Fig. 10 Afterpulsing probability density for a 200 µm SPAD device, measured at different temperatures.
Total afterpulsing probabilities ranging from 0.5% to 2% were typically obtained with 50 µm and 100 µm devices
operated at -15 °C with an excess bias voltage of 5 V.
It must be pointed out that it is very difficult to extract significant information about the nature and the
concentration of deep levels from afterpulsing measurements. This is because: i) the electric field has a strong influence
on the emission lifetimes, ii) the extent to which lifetimes are affected depends on the energy and on the charge state of
the deep level, and, iii) the electric field is not uniform in the depletion region.
Proc. of SPIE Vol. 6900 69001D-9
3.4. Time resolution
The time resolution of a SPAD device is determined by the precision with which the arrival instant of the incident
photon on the photodetector is identified. We recently demonstrated [27] that remarkable timing performance is
achievable with large-area SPADs, provided that the avalanche current is sensed at very low level (about a hundred µA
level), when the multiplication process is still confined within a small area around the photon absorption point. In order
to perform a true low-level sensing of the avalanche current it is mandatory to preserve the shape of the first part of the
leading edge by minimizing any filtering action. To this end, we designed and patented [28] a special current pick-up
circuit that can be added to any of the quenching circuit configurations described in the technical and scientific literature.
This circuit employs AC coupling with very short time constant for extracting a signal having a rise-time comparable to
that of the avalanche current. The extracted signal has a very short duration, which makes it possible to maintain very
good timing performance up to very high counting rate.
Time resolution measurements were performed by using an ultra-fast diode (Antel MPL-820 laser module) emitting
15 ps FWHM optical pulses at 820 nm wavelength. Fig. 11 shows the time resolution FWHM of SPAD devices having
different diameters as a function of the threshold level of timing discriminator [27]. Measurements were performed at
5 V excess bias voltage. It is confirmed that a time resolution of 35 ps or slightly less can be achieved regardless of the
device diameter, provided that the threshold is reduced below 10 mV. Fig. 12 shows the time resolution FWHM of a
50 µm-diameter SPAD device operated at different excess bias voltages as a function of the threshold level of timing
discriminator. These measurements clearly show that the time resolution is almost insensitive to the excess bias voltage
provided that the threshold is reduced below 10 mV.
0
50
100
150
200
250
300
350
0 100 200 300 400 500
Threshold voltage (mV)
Time resolution FWHM (ps)
100 µm
50 µm
20 µm
Fig. 11 Time resolution FWHM of SPAD devices having different diameters as a function of the threshold voltage of timing
discriminator. SPADs were operating at 5 V excess bias voltage.
Even more interestingly, we performed the same measurements on 50 µm-diameter SPAD devices belonging to
three different generations. The devices mainly differ in their peak electric field, which decreases by going from
generation #1 to #3. By looking at Fig. 13, it can be concluded that the lower the electric field the steeper is the increase
of the time resolution FWHM with the threshold voltage. Nevertheless, comparable time resolutions FWHM can be
attained by lowering the threshold voltage below 10 mV.
In summary, the use of a low current threshold breaks the complex trade-off between time resolution, active area, and
electric field, thus enabling the fabrication of large-area SPAD devices with excellent DCR and timing performance.
Proc. of SPIE Vol. 6900 69001D-10
0
50
100
150
200
250
300
350
400
0 100 200 300 400 500
Threshold voltage (mV)
Time resolution FWHM (ps)
V
exc
= 5 V
V
exc
= 7 V
V
exc
= 10 V
Fig. 12 Time resolution FWHM of 50µm-diameter SPAD operated at different excess bias voltages as a function of the threshold
voltage of timing discriminator.
Finally, Fig. 14 shows the time response of a 200 µm SPAD detector to a picosecond laser pulse at 820 nm
wavelength. The curve shows a prompt peak with a full-width at half maximum of 35 ps and a clean exponential
diffusion tail with a time constant of 280 ps.
20
40
60
80
100
120
140
160
180
200
220
0 100 200 300 400 500
Threshold voltage (mV)
Time resolution FWHM (ps)
Run #1
Run #2
Run #3
Fig. 13 Time resolution FWHM of 50µm SPAD devices belonging to different generations as a function of the threshold voltage
of timing discriminator.
Proc. of SPIE Vol. 6900 69001D-11
1
10
100
1000
10000
100000
11.5 12.0 12.5 13.0 13.5 14.0 14.5
Time (ns)
Counts
FWHM = 35 ps
Fig. 14 Time resolution curve of a 200 µm SPAD detector. The curve shows a prompt peak with a full-width at half maximum of
35ps and a clean exponential diffusion tail with a time constant of 240ps.
4. CONCLUSIONS
SPAD devices fabricated in planar epitaxial technology offer the typical advantages of microelectronic devices and
provide remarkable performance in single photon counting (SPC) and time-correlated single photon-counting (TCSPC).
Dedicated silicon fabrication technologies provide to the SPAD device designer the quality and flexibility necessary for
attaining further improvements requested by the users. In particular, there is realistic prospect of commercial
development of SPADs with diameter larger than 100 µm, high photon detection efficiency in visible range and excellent
timing resolution. To achieve this objective it is necessary to address both device design and technology issues and
circuit design issues. Our approach to such challenge was a combination of different strategies that is ultra-clean
fabrication process, specific gettering procedures, electric field profile engineering, and low-level detection of the
avalanche current leading edge. By adopting these strategies we fabricated SPAD devices with 200µm active area
diameter. At moderately low temperature (-25 °C with Peltier cooler) these devices have a typical DCR of 1500 c/s and it
is not difficult to select devices with less than 1000 c/s. A fairly low total afterpulsing probability of 2% was measured
with a dead-time of 80 ns. The photon detection efficiency peaks at 52% around 550 nm and stays above 30% over all
the visible range. By using our patented pulse pick-up circuit for processing the avalanche current, a photon timing
resolution
≤ 35 ps FWHM was obtained at room temperature, not dependent on the size of the SPAD active area.
These unprecedented characteristics make these SPAD devices a real alternative to MCP-PMTs in demanding TCSPC
applications.
5. ACKNOWLEDGEMENTS
This work was supported by European Commission, Sixth Framework Programme, Information Society Technologies
(NANOSPAD project) and the Italian Ministry of University and Research (MIUR-FIRB project n. RBNE01SLRJ;
MIUR-PRIN project n. 2005027857).
REFERENCES
1. J.R. Lakowicz, “Principles of Fluorescence Spectroscopy”, 3
rd
edition, Springer, Berlin (2006).
2. V. O'Connor and D. Phillips, "Time-correlated Single Photon Counting", Academic Press, London (1984).
Proc. of SPIE Vol. 6900 69001D-12
3. Becker, W., “Advanced Time-Correlated Single Photon Counting Techniques”, Springer, Berlin (2005).
4. R3809U MCP-PMT Data Sheet, Hamamatsu Photonics. Available online at: www.hamamatsu.com.
5. P.P.Webb, R. J. McIntyre, J.Conradi , "Properties of Avalanche Photodiodes," RCA Review, 35, 234-278 (1974).
6. H. Dautet, P. Deschampes, B. Dion, A.D. MacGregor, D. MacSween, R.J. McIntyre, C. Trottier, and P. Webb,
"Photon Counting techniques with silicon avalanche photodiodes," Appl.Opt., 32, 3894-3900 (1993).
7. S.Cova, A.Longoni, and A.Andreoni, "Towards picosecond resolution with single-photon avalanche diodes",
Rev. Sci. Instrum. 52, 408-412 (1981)
8. S. Cova, M. Ghioni, A. Lacaita, C. Samori, and F. Zappa, “Avalanche photodiodes and quenching circuits for
single-photon detection,” Appl.Opt., vol. 35, pp. 1956–1963 (1996).
9. A. Spinelli, A. L. Lacaita, “Physics and Numerical Simulation of Single Photon Avalanche Diodes", IEEE Trans.
Electron Devices, 44, 1931-1943 (1997).
10. SPCM-AQ Data Sheet, PerkinElmer Optoelectronics. Available online at: http://opto.perkinelmer.com
.
11. K. König, , I. Riemann, “High resolution optical tomography of human skin with subcellular resolution and
picosecond time resolution” J. Biom. Opt. 8, 432-439, 2003.
12. H. Yang, G. Luo, P. Karnchanaphanurach, T.-M. Louie, I. Rech, S. Cova, L. Xun, and X. S. Xie, “Protein
conformational dynamics probed by single-molecule electron transfer,” Science, 302, 262–266, (2003).
13. A. Rochas, M. Gani, B. Furrer, P. A. Besse, R. S. Popovic, G. Ribordy and N. Gisin, “Single photon detector
fabricated in a complementary metal–oxide–semiconductor high-voltage technology”, Rev. Sci. Instrum. 74, 3263
(2003).
14. F. Zappa, S. Tisa, A. Gulinatti, A. Gallivanoni, and S. Cova, ``Complete single-photon counting and timing
module in a microchip,'' Opt. Lett. 30, 1327 (2005).
15. S. M. Sze, “Physics of Semiconductor Devices”, pp. 520–527, Wiley, New York (1981).
16. Hurkx, G.A.M.; de Graaff, H.C.; Kloosterman, W.J.; Knuvers, M.P.G., “A new analytical diode model including
tunneling and avalanche breakdown”, IEEE Trans. On Elec. Dev., 39, 2090 - 2098 (1992).
17. Hurkx, G.A.M.; Klaassen, D.B.M.; Knuvers, M.P.G., “A new recombination model for device simulation
including tunneling”, IEEE Trans. On Elec. Dev., 39, 331 – 338 (1992)
18. G. Vincent, A. Chantre, and D. Bois, “Electric field effect on the thermal emission of traps in semiconductor
junctions,” J. App. Phys., 50, 5484-5487 (1979).
19. P. A. Martin, B. G. Streetman, and K. Hess,”Electric field enhanced emission from non-Coulombic traps in
semiconductors,” J. Appl. Phys., 52, 7409-7415 (1981).
20. K. Graff, “Metal Impurities in Silicon-device Fabrication”, 2
nd
edition, Springer-Verlag, Berlin (1995).
21. D. Schroder, “Semiconductor Material and Device Characterization”, 3
rd
edition, Wiley, New York (2006).
22. W.O. Oldham, R.R. Samuelson and P.Antognetti, "Triggering Phenomena in Avalanche Diodes," IEEE Trans.
Electron Devices, ED- 19, 1056-1060 (1972).
23. A. Lacaita, M. Ghioni, S. Cova, "Double epitaxy improves single-photon avalanche diode performance", Electron.
Lett., 25, 841-843 (1989).
24. M. Ghioni, A. Gulinatti, P. Maccagnani, I. Rech, S. Cova, “Planar silicon SPADs with 200-µm diameter and 35-ps
photon timing resolution”, Proceedings of SPIE -- Volume 6372, Advanced Photon Counting Techniques,
Wolfgang Becker Editor, 63720R, 0277-786X/06/$15 (2006).
25. A. Gulinatti, I. Rech, P. Maccagnani, M. Ghioni, S. Cova, “Large-area avalanche diodes for picosecond time-
correlated photon counting,” Proceedings of 35th European Solid-State Device Research Conference, ESSDERC
2005, 12-16 Sept. 2005, 355-358 (2005).
26. S.Cova, A.Lacaita and G.Ripamonti, "Trapping phenomena in avalanche photodiodes on nanosecond scale," IEEE
Electron.Dev.Lett., 12, 685-687 (1991).
27. A. Gulinatti, P. Maccagnani, I. Rech, M. Ghioni, and S. Cova, ``35 ps time resolution at room temperature with
large area single photon avalanche diodes,'' Electron. Lett. 41, 272 (2005).
28. S. Cova, M. Ghioni, and F. Zappa, ``Circuit for high precision detection of the time of arrival of photons falling on
single photon avalanche diodes,'' U.S. Patent 6 384 663 B2, May 7, 2002.
Proc. of SPIE Vol. 6900 69001D-13