A Geiger Mode Avalanche Photodiode
Fabricated in a Conventional CMOS Technology
A. Rochas, P.A. Besse, R.S. Popovic
Swiss Federal Institute of Technology, Lausanne
Alexis.Rochas@epfl.ch
Abstract
A Geiger mode avalanche photodiode with
outstanding characteristics is fully fabricated in a conventional CMOS process. The
80 µm2 active area photodiode for blue detection contains an efficient guard-ring
structure to prevent edge breakdown. The
selectivity for blue is obtained using a P+/
Nwell/Psubstrate dual junction structure.
The first junction close to the surface is
biased above the breakdown voltage, the second junction is short-circuited. A characterization using a passive quenching method is
done. A value of the dark count rate of only
220 c.p.s. is obtained for an excess bias voltage of -2.85V, at room temperature. The
maximum Geiger mode quantum efficiency
for this excess bias voltage is about 20% at
λ=460nm. The fabrication in a CMOS process opens the way to the co-integration of
passive quenching or active quenching depending on applications and read-out electronics.
1. Introduction
Silicon avalanche photodiodes biased
above breakdown voltage allow single photon detection. The so-called Geiger mode is
a very attractive working principle for highsensitivity, low-light level applications. We
recently demonstrated low-noise CMOS silicon avalanche photodiodes suitable for UVblue detection applications[1]. However,
such devices have to be combined with amplification electronics that deteriorate the
signal to noise ratio of the sensor. Therefore,
for many applications, like in flame detection, scintillation detection, or as night vision detectors, the photomultipliers tubes or
gas discharge tubes remain the most suitable
detectors. But they are large in size, easily
damaged, expensive, susceptible to magnetic fields, the linear range is limited and a
high voltage supply is required.
Kindt demonstrated [2] high quality pixels
for a Geiger mode avalanche silicon photodiode array but with a dedicated fabrication
process.
In this paper, we present a Geiger mode
avalanche photodiode (GAPD) fabricated in
a conventional CMOS technology that approachs the characteristics of a photomultiplier tube without its drawbacks. The
fabrication in a CMOS process offers additional advantages: possibility to integrate
passive or active quenching and read-out
electronics on the same chip, low-cost fabri-
cation, reproducibility of the dark count rate
and gain characteristics, and potentially very
low dark count rate due to the maturity of the
CMOS process.
periphery can be seen.
Similar design rules have been used to
fabricate 80µm2 active area photodiodes for
Geiger mode operation in an industrial
0.8µ m CMOS process.
2. Fabrication
We recently presented a new design idea
to fabricate avalanche photodiode in a conventional CMOS process[1]. The lateral diffusion of two nwell regions during the drivein step of the process is used to create a
guard ring at the edge of the p+ anode
[Fig.1]. Its lower doping concentration prevents an edge breakdown.
3. Geiger mode characterization
3.1. Experimental set-up for characterization
In the Geiger mode, a single photon can
initiate avalanche breakdown. A quenching
circuit is needed to generate single pulses by
ending the avalanche event. A load resistor
Rq is introduced between the voltage source
and the photodiode [Fig.3]. The bias voltage
corresponds to the sum of the breakdown
voltage Vbd and an excess voltage Ve. The
photocurrent induces a voltage drop over R q,
quenching the avalanche process. Nwell and
P substrate are shortcircuited to ground to ensure selectivity for blue radiations[1,9].
"
#
%
Figure 1. CMOS GAPD’s cross section
&
Fig.2 presents a full scan of a 80µm in
diameter photodiode with a 2µm blue LED
spot. The mean gain is plotted as a function
3
4
'
5
(
)
+
-
/
1
2
&
Ω
23
22
21
20
19
18
17
16
15
14
13
<G> 12
11
10
9
8
7
6
5
4
3
2
1
0
100
90
80
70
60
Y[microns]
50
40
30
20
10
90
100
80
70
60
50
40
30
20
0
10
0
X[microns]
Figure 2. Mean gain <G> as a function
of X position[µm] and Y position [µm]
of the spot position on the diode. The nonuniformity of the gain is less than 7% and no
detrimental prebreakdown at the junction
Figure 3. Experimental set-up
Geiger pulses generated across a 50 Ω resistor are amplified and counted using a programmable threshold level counter. After
quenching, the voltage across the GAPD
slowly increases from Vbd to V bd+V e. The
corresponding dead time is determined by
the value of Rq and the capacitance of the
GAPD (including parasitic capacitances).
3.2. Dead time/Afterpulsing/Dark count rate
Fig.4 shows a typical Geiger pulse observed across the 50Ω resistor without amplification. The bias voltage Vbd +V e is equal
to -21V with V bd = -19.65V. Using a 220kΩ
quenching resistor, the pulse width (fwhm)
is less than 20 ns.
Typical values of the dead time are about
2µs using a 220kΩ quenching resistor. Such
long dead times are inherent in passive
quenching. However, active quenching can
a carrier released during the dead time exhibits a smaller gain i.e. induces a smaller
pulse, a well adjustement of the counter
threshold allows the excluding of afterpulses
due to short lifetime traps.
Fig.5 presents the dark count rate as a
function of Ve , at room temperature. For a
80µm2 active area element, a dark count rate
of 220 c.p.s. is obtained, for Ve=-2.85V.
This value is in the range of commercialy
available photomultipliers[6].
3.3. Avalanche breakdown probability Pbd
Figure 4. Typical Geiger pulse shape
Dark count rate [c.p.s.]
be used[4] and integrated[5] to reduce the
dead time in the 50ns range. In applications
where a large dynamic range is not a requirement, such as flame detection, a long dead
time could be used to limit afterpulsing influences. Afterpulses are due to the release
of carriers trapped in the multiplication region during an avalanche process. In a first
study of the afterpulsing, we noticed that at
room temperature, most of the afterpulses
Breakdown initiation probabilities as a
function of V e in both cases of pure electrons
and pure holes injection [Fig.6] in the multiplication region have been calculated for our
P +/Nwell structure using the numerical calculation described by Oldham[7]. A one side
abrupt junction assumption was made for
this calculation. The ionization coefficients
published by Van Overstraeten and de
Man[8] for electrons α e and for holes α h
have been used.
1,0
0,8
Pbd
Rq=220k Ω
Ve =-1.35V
Pure electrons injection
0,6
λ=420nm
0,4
100
0,2
Pure holes injection
0,0
0
1
2
3
4
5
Ve [V]
10
0,0
0,5
1,0
1,5
Ve
2,0
2,5
3,0
[V]
Figure 5. Dark count rate[c.p.s.] as a
function of the excess voltageV e  [V]
were encountered during the first microsecond after a full discharge of the diode. Since
Figure 6. Calculated and measured at
420nm breakdown probabilities as a
function of  Ve [V]
The breakdown probability has also been
measured at λ=420nm as a function of Ve
[Fig.6]. The GAPD has been illuminated
with a monochromator using a 150W Xenon
lamp combined with a 1% transmission neu-
tral density filter. The breakdown initiation
probability Pbd is then given by:
P bd = α
--η
where η is the quantum efficiency of the
diode (η≈ 50% at 420nm, [1]) and α is the
Geiger mode quantum efficiency defines as
the probability that an incident photon will
be detected. One can see that Pbd increases
with V e , as well as dark count rate [Fig.5]
and afterpulsing probability. Since at
λ=420nm the light absorption occurs very
close to the surface (i.e. above the multiplication region), the illumination results in a
close to pure electrons injection.
3.4. Geiger mode quantum efficiency α
Fig.7 shows the Geiger mode quantum efficiency α of the GAPD for Ve = -0.85V,
-1.85V and -2.85V as a function of the wavelength, at room temperature.
Ve = -0.85V
Ve = -1.85V
Ve = -2.85V
-1
α
10
-2
10
10-3
300
400
500
600
700
800
900
Wavelength [nm]
Figure 7. GAPD’s α as a function
of λ for different values of Ve
Since P bd increases with V e [Fig.6], α also
increases with V e . A Geiger mode quantum
efficiency of 20% , close to state-of-the-art
photomultipliers[6],
is
obtained
for
Ve =-2.85V and λ=460nm.
The dual jonction structure P +/Nwell /
P substrate leads to a large selectivity for blue
wavelengths without any additional interference filter. If necessary, α can be greatly improved in the UV range by removing the
Si3N 4 layer deposited to passivate the circuit[9]. The maximum of α can also be
shifted towards lower wavelengths using advanced CMOS processes with ultra-shallow
junctions or towards higher wavelengths
using technologies with deeper junctions.
4. Conclusion
We demonstrate the outstanding potential
of a 80µm2 active area GAPD fabricated in a
conventional CMOS process. Using a passive quenching, a low value of the dark
count rate of 220 c.p.s. has been obtained for
an excess bias voltage of -2.85V, at room
temperature. The maximum Geiger mode
quantum efficiency for this excess bias voltage is 20% at λ=460nm. The fabrication in a
CMOS process opens the way to the co-integration of quenching and read-out electronics.
5. References
[1] A. Pauchard, A. Rochas, Z. Randjelovic, P.A. Besse and
R.S. Popovic, "Ultraviolet Avalanche Photodiode in CMOS
Technology", IEDM2000, Technical Digest, 709-712
[2] W.J. Kindt, "Geiger mode avalanche photodiode arrays
for spatially resolved single photon counting", PhD thesis, Delft
University Press, 1999
[3] R.G. Brown, K.D. Ridley, J.G. Rarity, "Characterization
of silicon avalanche photodiodes for photon correlation measurements. I: Passive quenching", Applied-Optics Vol.25, N°22,
15 Nov. 1986, 4122-4126
[4] R.G. Brown, R. Jones, J.G. Rarity, K.D. Ridley, "Characterization of silicon avalanche photodiodes for photon correlation measurements. II. Active quenching", Applied-Optics
Vol.26, N°12, 15 June 1987, 2383-2389
[5] F. Zappa et al., "An integrated active quenching circuit for
single-photon avalanche diodes", IEEE transactions on instrumentation and measurement, Vol.49, N°6, dec2000, 1167-1175
[6] http://www.electron-tubes.co.uk
[7] W.G.Oldham, R.R. Samuelson, P.Antognetti, "Triggering phenomena in avalanche diodes", IEEE Transactions on
Electron devices, Vol.19, N°9, 1972, 1056-1060
[8] R. Van-Overstraeten-R, H. De-Man, "Measurement of the
ionization rates in diffused silicon p-n junctions", Solid-StateElectronics, Vol.13, N°5, May 1970, 583-608
[9] A. Pauchard, P.A. Besse, R. Popovic, "A new silicon
blue/UV selective stripe-shaped photodiode," Sensors and Actuators A, 76 (1999), 172-177