Highly Reliable, Cost-effective and Temperature-stable
Top-illuminated Avalanche Photodiode (APD) for
100G Inter-Datacenter ER4-Lite Applications
Jack Jia-Sheng Huang
1,2
, H. S. Chang
2
, Yu-Heng Jan
2,1
, H. S. Chen
2
, C. J. Ni
2
, E. Chou
2
,
S. K. Lee
2
and Jin-Wei Shi
3
1
Source Photonics, 8521 Fallbrook Avenue, Suite 200, West Hills, CA 91304, U.S.A.
2
Source Photonics, No.46, Park Avenue 2
nd
Rd., Science-Based Industrial Park, Hsinchu, Taiwan
3
Department of Electrical Engineering, National Central University, Zhongli, Taiwan
Keywords: Avalanche Photodiode, APD Photodetector, InGaAs/InAlAs APD, Top-Illumination APD, 100Gb/s Ethernet,
100G Datacenter, ER4, IEEE802.3.
Abstract: One of the key enablers for 100G ER4-Lite optical modules is 25G APD photodetector that can be employed
in 30-40km optical links for inter-datacenter applications. In this paper, we demonstrate that a cost-effective
top-illuminated InGaAs/InAlAs APD photodetector can be manufactured to meet stringent IEEE standard of
100G ER4-Lite. The 25G APD shows high bandwidth, high sensitivity with superb temperature stability of
breakdown voltage. The APD photodetector also possesses excellent durability against harsh optical and
electrical overload in both burst and continuous modes. Robust reliability performance based on aging
conditions of 85-175C has also been achieved with an activation energy of 1.18eV.
1 INTRODUCTION
Optical transmission for data center applications is
increasingly popular for future interconnection within
and between the cities (Bilal et al., 2013; Manzano et
al., 2013; Basa, 2016). The Institute of Electrical and
Electronics Engineers (IEEE) has developed various
802.3 standards for the 125m-40km optical links
listed in Table I. Since 2007, the technological
development on 40Gbit/s and 100Gbit/s applications
have further accelerated (Caruso, 2007). Historically,
the data center communications over 10km distance
represented the mainframe. However, ER4
applications have recently gained significant traction
due to its technological advantage in the 30-40km
optical reach.
One of the main challenges for ER4 ethernet is to
maintain undistorted optical signal over a long
transmission distance. One way to achieve the long-
distance transmission is to increase the output power
from the laser emitter. However, there are several
engineering tradeoffs associated with the higher
power design including reliability, design complexity
and processing cost (Connolly, 2002; Rinner et al.,
2003; Huang et al., 2012). In this paper, we
demonstrate an alternative low-cost ER4 solution by
employing high-sensitivity, high-speed top-
illuminated 25G APD photodetector on the receiver
side. Our data shows that it is feasible to attain low-
cost manufacturing of 25G APD with well-rounded
design-in aspects in bandwidth, sensitivity and
reliability. The top-illuminated APD (Huang et al.,
2016; Chen, et al., 2018) can bring advantages of low
cost and simple fabrication compared to the bottom-
illumiation design (Nada et al., 2014; Nada et al.,
2012). In addition, the 25G APD photodetector can
achieve superb durability against harsh electrical and
optical overload stress conditions as well as robust
reliability performance.
Table 1: IEEE 802.3 standards for 40G/100G optical
modules and fiber optics.
40G/100G Ethernet
Transmission
Distance
Optical fiber
SR4
125m
Multi-mode
fiber (MMF)
FR4 (CWDM4)
2km
Single-mode
fiber (SMF)
LR4
10km
SMF
ER4 Lite / ER4
30km / 40km
SMF
Huang, J., Chang, H., Jan, Y-H., Chen, H., Ni, C., Chou, E., Lee, S. and Shi, J-W.
Highly Reliable, Cost-effective and Temperature-stable Top-illuminated Avalanche Photodiode (APD) for 100G Inter-Datacenter ER4-Lite Applications.
DOI: 10.5220/0006510601190124
In Proceedings of the 6th International Conference on Photonics, Optics and Laser Technology (PHOTOPTICS 2018), pages 119-124
ISBN: 978-989-758-286-8
Copyright © 2018 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
119

2 EXPERIMENTAL
Figures 1 shows the conceptual schematics of the
top-illuminated APD device structure. The epitaxial
layers were grown by molecular beam epitaxy
(MBE). From top to bottom, it is composed of the p
+
-
InGaAs contact layer, p
+
-InP window layer, graded
InGaAs/InAlAs, InGaAs absorption layer, p-doped
InAlAs field control layer (~1x10
18
cm
-3
), intrinsic
InAlAs multiplication (M-) layer, and N
+
InAlAs/InP
contact layers. In order to shorten the avalanche delay
time, a thin M-layer (~90 nm) is chosen in our device
structure (Nada et al., 2015; Campbell et al., 2004).
The detail of our epi-layer structure can be referred to
our previous work (Huang et al., 2017; Chen et al.,
2018).
Figure 1: Schematics of top-illuminated mesa-type SACM
APD photodetector structure with coplanar N and P
electrodes.
We adopted the CH
4
/H
2
/Ar dry etching technique that
exhibited a slow etching rate for the In
0.52
Al
0.48
As
material in order to precisely control the depth of
mesa etch. The active diamter of the APD device was
about 20m. After the mesa etch and p- and n-metal
contacts, a thick (~5 m) benzocyclobutene (BCB)
film was used in the passivation process to reduce the
dark current and to minimize the parasitic
capacitance. An anti-reflection (AR) coating at 1.31
m wavelength was deposited on the surface of our
device to enhance its responsivity performance. A
co-planar stripe (CPS) was integrated with our device
for on-wafer high-speed measurement. The layout of
our CPS was suitable for traditional wire bonding
assembly of the top-illuminated APDs, bringing cost
advantage compared to the flip-chip package that was
used in some 25G bottom-illuminated APDs (Nada et
al., 2015; Nada et al., 2014).
3 RESULTS AND DISCUSSION
3.1 Avalanche Breakdown Voltage
Figure 2 shows the typical reverse current-voltage
(IV) curve of the 25G mesa-type SACM APD
photodetector. On the IV curve, there are two
transitions along the reverse voltage that occur at 3V
and 26V, respectively. The first transition at 3V
corresponds to the punch-through voltage (V
pt
) at
which electric field depletes both the InAlAs
multiplication and i-InGaAs absorption layers
(Huang et al., 2017). The second transition represents
the avalanche breakdown voltage (V
br
) that is
determined by the InAlAs multiplication layer. The
avalanche breakdown voltage taken at 10A is
estimated to be about 26V for this device. The typical
range of avalanche breakdown voltage is 24-26V.
Our design allows excellent avalanche gain while
keeping the tunneling and impact ionization away
from the InGaAs absorption layer. The dark current
refers to the reverse current at 0.9V
br
is estimated to
be about 13.4nA at 25C.
Figure 2: The reverse IV curve of the 25G APD
photodetector where the first transition is related to the
punch-through (V
pt
) and the second related to the avalanche
breakdown (V
br
).
3.2 Bandwidth
Figure 3 shows the bandwidth plot of the APD based
on small-signal modulation response at 25C. We
characterized the bandwidth curve at reverse voltage
of 17V and 21V. At the reverse bias of 17V, the
bandwidth taken by the 3dB roll-off reached 20GHz.
Such bandwidth was adequate for meeting the
requirement for 4x25G ER4 ethernet (Chen, et al.,
2018). The corresponding multiplication factor or
gain at -17V was about 2.2 (M=2.2). As the reverse
bias increased to 21V for high gain operation (M=5),
the 3dB bandwidth was maintained at around 15GHz.
PHOTOPTICS 2018 - 6th International Conference on Photonics, Optics and Laser Technology
120

Figure 3: Bandwidth plot of 25G APD photodetector
measured at 25C with the reverse bias at 17V to BW of
20GHz.
3.3 Sensitivity
Figure 4 shows the bit error rate (BER) of the 25G
APD photodetector as a function of input optical
power. At near-V
br
operation with gain of 6, the APD
achieved error-free for bit error rate (BER)<10
-12
,
with the sensitivity of -17dBm. Such sensitivity level
provided good margin for the photoreceiver detection
over the 40km transmission over fiber (Laird et al.,
2003; Alpert, 2015).
Figure 4: BER of 25G APD device showing that the
sensitivity of -17dBm can be achieved for BER of 10
-12
.
3.4 Temperature Stability
Temperature stability of the APD breakdown voltage
is also an important parameter for ER4 datacenter
applications. When the datacenter is running hot, the
stable temperature stability of the APD would allow
the gain to maintain constant when subject to
temperature fluctuations.
Figure 5: The normalized avalanche breakdown voltage of
25G APD as a function of temperature ranging from 25C
to 145C.
The avalanche breakdown voltage followed a
linear relationship with temperature (Tyagi, 1968), as
shown in Equation (1).
)(1)(/)(
00
TTTVTV
brbr
 
(1)
where V
br
(T)/V
br
(T
0
) is the normalized avalanche
breakdown voltage to the reference temperature T
0
and  is the normalized temperature coefficient.
Figure 5 shows the normalized avalanche breakdown
voltage as a function of temperature based on two
different APD wafers. The normalized temperature
coefficient () showed ultra-low value of about
5.1x10
-4
C
-1
, lower compared to the reported value of
7.2x10
-4
C
-1
(Tan et al., 2010).
3.5 Optical & Electrical Overload
Stress
In order to verify the photodetector’s robustness
against the simultaneous electrical and optical
stresses, the mesa-type APD devices were tested with
overload stress in burst and continuous modes, as
illustrated in Fig. 6. To determine the damage
threshold of overload, the optical stress was ramped
up from -4dBm to +4dBm.
Figure 6: Schematic of electrical and optical overload stress
applied to the APD for both burst and CW modes.
Highly Reliable, Cost-effective and Temperature-stable Top-illuminated Avalanche Photodiode (APD) for 100G Inter-Datacenter ER4-Lite
Applications
121

Table 2: Damage threshold of optical and electrical
overload stresses of APD for burst and CW modes.
Burst mode
CW mode
Damage
threshold
No failure up to
+4dBm.
No failure up to
+1dBm.
In the burst mode, the optical stress of 1% pulsed
duty cycle was applied to the APD for 60 seconds
while the device was electrically stressed at 2V below
V
br
. All APD devices survived with no failure after
being subjected to overload stress up to +4dBm, as
shown in Table 2. In the continuous mode, the
damage threshold was around +1dBm, well
exceeding the -6dBm requirement.
3.6 Reliability
To establish long-term reliability performance, the
APD devices were stressed at 85C under a reverse
current of 100A (Telcordia, 2004). Figure 7 shows
the relative change in avalanche breakdown voltage
as a function of aging time. The failure criterion was
defined as 1V change in V
br
. All APD receivers have
passed reliability test with excellent margin after
5100hr aging.
Figure 7: The relative breakdown voltage change as a
function of aging time based on the stress condition of
85C, 100A.
Due to the small degradation at regular aging
condition of 85C, the APD devices were also
stressed at highly elevated temperature. Figure 8
shows the failure time distributions of the APD from
the aging groups of 165C and 175C. The failure
time at 50% cumulative probability shows the
statistical mean-time-to-failure (MTTF). The MTTF
values for 165C and 175C are 896hrs and 446 hrs,
respectively.
Figure 8: The failure time distributions of APD
photodetectors based on the stress condition of 165C and
175C.
The device failure time (t
f
) follows the modified
Black’s equation (Black , 1969; Huang, 2005; Huang
et al., 2005) which provides a good empirical
description of device degradation over time as a
function of stress current and temperature as shown
in Equation (2).
)exp(
kT
E
I
A
t
a
N
f

(2)
In Equation (2), the first term represents the
current acceleration factor where A is a constant, I is
the stress current, and N is the current exponent; the
second term represents the temperature acceleration
where E
a
is the activation energy, k is the
Boltzmann’s constant, and T is the temperature.
For the sake of activation energy study, Equation
(2) can be rewritten in the form of natural logarithm
as shown in Equation (3) where the third term can
readily determine the activation energy.
kT
E
ILnNALnMTTFLn
a
• )()()(
(3)
Figure 9 shows the plot of Ln(MTTF) vs. 1/(kT)
where the slope is equal to Ea. The error bars
indicates the variation between devices. Based on the
experimental aging data of 165C and 175C, the Ea
of APD is estimated to be 1.18eV, in close agreement
with other reported values (Ishimura et al., 2007;
Watanabe et al., 1996).
PHOTOPTICS 2018 - 6th International Conference on Photonics, Optics and Laser Technology
122

Figure 9: The plot of Ln(MTTF) vs. 1/(kT) for the
determination of E
a
. The E
a
value was 1.18eV.
Table 3: Projected device lifetimes of mesa-type APD at
50C operating condition based on extrapolation from the
stress condition.
Aging stress
temperature
Failure time at
stress
temperature
Device
operating
lifetime at
50C
165C
896 hours
6912 years
175C
446 hours
6909 years
With the E
a
establishment, the failure times from
the aging test can be extrapolated to project the device
lifetimes at the operating condition by using Equation
(2). For the 50C operating condition, the device
lifetimes of the 25G APD are estimated to be close to
6900 years shown in Table 3, representing great
reliability margin for the 20-year stringent
requirement per Telcordia.
4 CONCLUSIONS
We have demonstrated low-cost manufacturing of
25G top-illuminated, high-sensitivity SACM APD
photodetector for the 100G ER4 Lite datacenter
applications. The high-speed, high-sensitivity APD
showed favorable V
br
(24-26V). The 3dB bandwidth
reached 20GHz at reverse bias of 17V, meeting the
requirement of 4x25G ER4 ethernet. The sensitivity
of -17dBm was achieved to enable the detection over
the error-free 40km transmission over fiber at
BER<10
-12
. The APD also showed excellent
temperature stability where the normalized
temperature coefficient () was about 5.1x10
-4
C
-1
,
superior to the industry benchmark.
The APD photodetector showed superb
durability against simultaneous electrical and optical
overload stress. In the burst mode, the APD showed
no failure up to +4dBm. In the CW mode, the damage
threshold was +1dBm, well exceeding the -6dBm
requirement. We have also achieved robust reliability
for the APD with little degradation after 5100hr aging
at 85C. Based on the aging data of 165C and 175C,
the actiation energy of the APD was determined to be
1.18eV. The projected device lifetime of the APD
was extrapolated to be about 6900 years at the
operating condition of 50C.
REFERENCES
Bilal, K., Khan, S.U. and Zomaya, A.Y., ”Green data.
center networks: challenges and opportunities”, 11th Int’l
Conf. on Frontiers of Information Technology (FIT),
p.229-234, (Islamabad, Pakistan, 2013).
Manzano, M., Bilal, K., Calle, E., and Khan, S. U., "On the
connectivity of data center networks," IEEE Comm.
Lett., vol. 17, no. 11, p. 2172-2175, (2013).
Basa, J., ”Source Photonics announces shipment of 10,000
single mode 100G QSFP28 modules”, Optical Fiber
Communications Conference (OFC, Anaheim, CA,
2016).
Caruso, J., ”Group pushes 100 Gigabit ethernet: the ”Road
to 100G’ alliance is born”, Network World, (June 21,
2007).
Connolly, M. J., “Semiconductor Optical Amplifiers”,
(Springer-Verlag, Boston, MA, 2002).
Rinner, F., Rogg, J., Kelemen, M., and Poprawe, R., “Facet
temperature reduction by a current blocking layer at the
front facets of high-power InGaAs/AlGaAs lasers”,
J. Appl. Phys., vol.93, no.3, p.1848-1850, (2003).
Huang, J.S., Isip, E., and Carson, R.F. “High-resolution
wavelength stability aging study of C-band 100mW
high-power DWDM laser modules”, IEEE Photon.
Conf. (Burlingame, CA, 2012), p.538-539.
Huang, M., Cai, P., Li, S., Wang, L., Su, T.-I., Zhao, L.,
Chen, W., Hong, C.-Y., and Pan, D., “Breakthrough of
25Gb/s Germanium on Silicon Avalanche Photodiode”,
Optical Fiber Communications Conf., OFC Technical
Digest, (Anaheim, CA, 2016), Paper Tu2D.2.
Chen, Y.H., Wun, J.M., Wu, S.L., Huang, J.S., Jan, Y.H.,
Chang, H.S., Chen, H.S., Ni, C.J., Chou, E. and Shi,
J.W., “Top-illuminated In
0.52
Ga
0.48
As based avalanche
photodiode with dual charge layers for high-speed and
low dark current performances”, IEEE J. Select. Topic.
Quantum Electronics, vol.24, no.2, p.3800208, (2018).
Nada, M., Muramoto, Y., Yokoyama, H., and Matsuzaki,
H., ”High-speed high-power-tolerant avalanche
photodiode for 100-Gb/s applications”, IEEE Photonic
Conf., (San Diego, CA, 2014), p.172-173.
Highly Reliable, Cost-effective and Temperature-stable Top-illuminated Avalanche Photodiode (APD) for 100G Inter-Datacenter ER4-Lite
Applications
123

Nada, M, Muramoto, Y., Yokoyama, H., Ishibashi, T., and
S. Kodama, “InAlAs APD with high multiplied
responsivity-bandwidth product (MR-bandwidth
product) of 168 A/W.GHz for 25 Gbit/s high-speed
operations”, Electron. Lett., vol. 48, no. 7, pp. 397–399,
(2012).
Nada, M., Yoshimatsu,T., Muramoto, Y., Yokoyama, H.
and Matsuzaki, H., “Design and Performance of High-
Speed Avalanche Photodiodes for 100-Gb/s Systems
and Beyond,” IEEE/OSA Journal of Lightwave
Technology, vol. 33, no. 5, pp. 984-990, (2015).
Campbell, J. C., Demiguel, S., Ma, F., Beck, A., Guo, X.,
Wang, S., Zheng, X., Li, X., Beck, J. D., Kinch, M. A.,
Huntington, A., Coldren, L. A., Decobert, J. and
Tscherptner, N., “Recent Advances in Avalanche
Photodiodes,” IEEE J. of Sel. Topics in Quantum
Electronics, vol. 10, pp. 777-787, (2004).
Huang, J.S., Chang, H.S., Jan., Y.H., Ni, C.J., Chen, H.S.,
and Chou, E., ”Temperature dependence study of mesa-
type InGaAs/InAlAs avalanche photodiode
characteristics”, Adv. Optoelectron., no. 2084621, p.1-
5, (2017).
Laird, J.S., Hirao, T., Onoda, S., Ohyama, H., and T.
Kamiya, T., “Heavy-ion induced single-event transients
in high-speed InP-InGaAs avalanche photodiodes”,
IEEE Tran. Nuclear Sci., Vol. 50, No. 6, p.2225-2232,
(2003).
Alpert, A., “High-speed jitter testing of XFP transceivers”,
Viavi White Paper, (2015).
Tyagi, M.S., “Zener and avalanche breakdown in silicon
alloyed p-n junctions” Solid State Electron., vol. 11,
p.99-128, (1968).
Tan, L.L. J., Ong, D.S.G., Ng, J.S., Tan, C.H., Jone, S.K.,
Qian Y., and David, J.P.R., “Temperature dependence
of avalanche breakdown in InP and InAlAs”, IEEE J.
Quantum. Electron., vol.46, no.8, p.1153-1157, (2010).
”Generic reliability assurance requirements for
optoelectronic devices used in telecommunication
equipment”, Telcordia, GR-468-CORE, (2004).
Black, J.R., ”Electromigration failure modes in aluminum
metallization for semiconductor devices”, Proc. IEEE,
vol.57, no.9, p.1587-1594, (1969).
Huang, J.S., ”Temperature and current dependences of
reliabliity degradation of buried heterostrcture
semiconductor lasers”, IEEE Tran. Device Mater.
Reliab. Vo.5, no.1, p.150-154, (2005).
Huang, J.S., Nguyen, T., Hsin, W., Aeby, I., Ceballo, R.,
and Krogen, J., ”Reliability of etch-mesa buried
heterostructure semiconductor lasers”, IEEE Tran.
Device Mater. Reliab. Vo.5, no.4, p.665-674, (2005).
Ishimura, E., Yagyu, E., Nakaji, M., Ihara, S., Yoshiara, K.,
Aoyagi, T., Tokuda, Y., and Ishikawa, T., ”Degradation
mode analysis on highly reliable guardring-free planar
InAlAs avalanche photodiodes”, J. Lightwave Tech.,
vol. 25, no. 12, p.3686-3693, (2007).
Watanabe, I., Tsuji, M., Hayashi, M., Makita K., and
Taguchi, K., ”Reliability of Mesa-Structure InAlGaAs-
InAlAs superlattice avalanche photodiodes”, IEEE
Photonics Tech. Lett., Vol. 8, No. 6, p.824-826, (1996).
PHOTOPTICS 2018 - 6th International Conference on Photonics, Optics and Laser Technology
124