Research Article Vol. 28, No.

23 / 9 November 2020 / Optics Express 34772

Photo detection and modulation from 1,550 to

2,000 nm realized by a GeSn/Ge
multiple-quantum-well photodiode on a 300-mm
Si substrate
H AO Z HOU , 1 S HENGQIANG X U , 2 S HAOTENG W U , 1 Y I -C HIAU
H UANG , 3 P ENG Z HAO, 1 J INCHAO TONG , 1 B ONGKOWN S ON , 1
X IN G UO, 1 D AOHUA Z HANG , 1 X IAO G ONG , 2 AND C HUAN S ENG
TAN 1,*
1 School of Electrical and Electronic Engineering, Nanyang Technological University, 639798, Singapore
2 Department of Electrical and Computer Engineering, National University of Singapore, 117576, Singapore
3 Applied Materials Inc., Sunnyvale, CA 95054-3299, USA
* tancs@[Link]

Abstract: A GeSn/Ge multiple-quantum-well (MQW) p-i-n photodiode structure was proposed

for simultaneously realizing high detectivity photo detection with low dark current and effective
optical modulation based on the quantum confined Stark (QCSE) effect. The MQW stacks were
grown on a 300-mm Ge-buffered Si substrate using reduced pressure chemical vapor deposition
(RPCVD). GeSn/Ge MQW p-i-n photodiodes with varying mesa diameters were fabricated
and characterized. An ultralow dark current density of 16.3 mA/cm2 at -1 V was achieved as
expected due to the low threading dislocation density (TDD) in pseudomorphic GeSn layer.
Owing to the ultralow dark current density and high responsivity of 0.307 A/W, a high specific
detectivity of 1.37×1010 cm·Hz1/2 /W was accomplished at 1,550 nm, which is comparable with
commercial Ge and extended-InGaAs photodetectors. Meanwhile, the bias voltage-dependent
photo response was investigated from 1,700 to 2,200 nm. The extracted effective absorption
coefficient of GeSn/Ge MQW shows a QCSE behavior with electric field-dependent exciton
peaks from 0.688 to 0.690 eV. An absorption ratio of 1.81 under -2 V was achieved at 2 µm, which
shows early promise for effective optical modulation. The high frequency response was calculated
theoretically, and the predicted 3-dB bandwidth for the photodiode with a mesa diameter of 30
µm could reach 12 GHz at -2 V.

© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement

1. Introduction
Ge photodetectors are widely used in the state-of-art silicon photonics since the direct bandgap
of Ge is ∼0.8 eV which exactly corresponds to a pivotal telecommunication band: C-band. The
incorporation of another group IV material Sn is known to reduce the bandgap of Ge and the
detection range of GeSn photodetectors with a small Sn fraction can cover all telecommunication
bands from O-band to U-band (from 1,260 to 1,675 nm) [1–3]. Incorporating higher Sn
concentration into Ge can even push the upper limit to longer wavelength and make it capable
of photo detection at 2 µm [4]. Over the past decades, 2 µm band has received increasing
interests due to its potential as a new communication window with advancement in hollow-core
photonic-bandgap fibers (HC-PBGFs) and thulium-doped fiber amplifiers (TDFAs) [5–10].
GeSn-based photodetectors with enhanced and extended detectivity beyond 2 µm have been
demonstrated [11–15]. In addition, optical modulation on the basis of Franz-Keldysh effect
and lasing operation based on tuned direct bandgap in bulk GeSn have also been substantiated
experimentally [16–21]. Therefore, GeSn has the potential to offer a monolithic integration

#409944 [Link]
Journal © 2020 Received 10 Sep 2020; revised 21 Oct 2020; accepted 22 Oct 2020; published 2 Nov 2020
 Research Article Vol. 28, No. 23 / 9 November 2020 / Optics Express 34773

methodology for the entire suit of active photonics devices including laser, modulator, and
photodetectors.
GeSn photodetectors suffer from higher dark current which degrades their detectivity compared
with commercial Ge photodetectors. Due the lattice mismatch, GeSn with higher Sn fraction
directly grown on Ge-buffered Si substrate is afflicted with higher threading dislocation density
(TDD). The misfit dislocations not only act as generation/recombination centers for Shockley-
Read-Hall (SRH) leakage current, but also offer energy states in the bandgap for trap-assisted
tunneling (TAT) leakage current [22]. Taking the well-established knowledge from III-V quantum
structures, GeSn/Ge MQW structure using pseudomorphic GeSn as well region is able to decrease
the TDD significantly by avoiding strain relaxation. Moreover, MQW structure provides strong
confinement for excitons at room temperature and the corresponding QCSE effect has been
widely used for high speed and low power dissipation modulator. Although strong QCSE effect
has been reported in Ge/SiGe MQW [23,24], it remains a mystery in GeSn/Ge MQW system.
In this work, Ge0.92 Sn0.08 /Ge MQW p-i-n photodiodes were fabricated and characterized. For
the photodiode with mesa diameter of 250 µm, an ultralow dark current density of 16.3 mA/cm2
at -1 V was achieved due to the pseudomorphic GeSn/Ge MQW design. The photo detection
performance of the photodiode was examined from 1,550 to 2,000 nm. The fabricated p-i-n
photodiode shows a high-efficiency optical response with responsivity of 0.307 A/W at 1,550 nm.
Taking both dark current and responsivity into consideration, a high detectivity of 1.37×1010
cm·Hz1/2 /W was achieved which is comparable with commercial Ge photodetector. Meanwhile,
its potential for light modulation was experimentally evaluated through characterization of
spectral response under varying reverse bias voltages. Significant electric field-dependent photo
response was observed from 1,700 to 2,200 nm. The extracted effective absorption coefficient
spectral of the GeSn/Ge MQW region show a QCSE behavior. It is the first time that QCSE
effect was discovered in GeSn/Ge MQW system. Relatively high absorption ratio (∆α/α0 ) of
1.81 under reverse bias voltage of 2 V was achieved at 2 µm, showing its potential for effective
optical modulation.

2. Material characterizations and device fabrication

The GeSn/Ge MQW structure was grown on a 300 mm Si substrate using RPCVD. Before the
growth of device layer, a 1 µm strain-relaxed Ge layer was grown as a buffer layer to confine
the defects at Ge/Si interface. The n-type and p-type Ge contact layers were realized by in-situ
doping during the growth procedure in RPCVD. The intrinsic GeSn/Ge MQW region consists
of 15 periods of 20 nm Ge barrier layer and 7.5 nm Ge0.92 Sn0.08 well layer. The thin GeSn
well layer was designed not only to reduce the TDD by preventing strain relaxation, but also to
provide better confinement for the excitons. The material quality after growth was characterized
by atomic force microscope (AFM) and cross-sectional transmission electron microscopy (TEM)
in Fig. 1. The root-mean-square (RMS) roughness obtained from a scanning area of 50×50 µm2
is 1.04 nm and the achieved roughness is uniform across the entire 300 mm wafer surface. From
the cross sectional TEM image, uniform and abrupt interfaces between GeSn and Ge layers are
observed which manifest that the as-grown material is of high quality.
X-ray diffraction (XRD) characterization was used to evaluate the strain in the GeSn/Ge MQW
region. Figure 2(a) shows the high-resolution XRD rocking curves of the as-grown sample at
(004) orientation. The main peaks (from right to left) correspond to Si substrate and relaxed Ge
buffer layer respectively. The satellite peaks on the left of Ge peak are uniform and well-defined,
revealing a uniform GeSn/Ge MQW structure with high crystalline quality. The asymmetric
(115) reciprocal space mapping (RSM) of the GeSn/Ge MQW structure is shown in Fig. 2(b).
The diffraction peaks of the GeSn/Ge MQW region have same reciprocal lattice vector Qx as the
strain-relaxed Ge buffer layer, demonstrating that the GeSn layers in the MQW structure are fully
strained.
 Research Article Vol. 28, No. 23 / 9 November 2020 / Optics Express 34774

Fig. 1. (a) 3D AFM image of as-grown sample across a scanning area of 50 × 50 µm2 . (b)
Cross-sectional TEM image of the GeSn/Ge MQW n-i-p structure.

Fig. 2. (a) High resolution XRD rocking curve of GeSn/Ge MQW sample at (004)
orientation. (b) Asymmetric (115) reciprocal space mapping of the as-grown wafer.

GeSn/Ge MQW p-i-n photodiode with double-mesa structure was fabricated on the as-grown
sample. Figure 3(a) shows the three-dimensional (3D) schematic diagram of the designed
GeSn/Ge MQW photodiode structure. In the fabrication procedure, top mesas with various
diameters were firstly patterned and formed by chlorine-based reactive ion etching (RIE) to
expose bottom p+ -Ge region. Bottom mesa was implemented using similar process. This step is
for the electrode pads to be formed on the intrinsic Ge buffer layer rather than the heavily-doped
p+ -Ge, favorable for high frequency performance. After the formation of double-mesa structure,
a ∼400 nm SiO2 was deposited as passivation and isolation layer by plasma enhanced chemical
vapor deposition (PECVD) at temperature of 250°C. The contact window was patterned and
opened by a combination of fluorine-based dry etching and diluted buffered oxide etch (BOE)
wet etching for a decent etching surface. Finally, metal stack of Ti (20 nm)/TiN (50 nm)/Al
(300 nm) was patterned and formed by sputtering and lift-off process. The entire process was
performed under the temperature of 250°C to avoid Ge-Sn inter-diffusion in the MQW region,
which may affect the electrical and optical properties of the devices. The top-view scanning
electron microscope (SEM) image of the fabricated device with mesa diameter of 250 µm is
shown in Fig. 3(b). The clean surface with well-defined double-mesa structure and standard
ground-signal-ground (GSG) electrode pads can be observed clearly.
 Research Article Vol. 28, No. 23 / 9 November 2020 / Optics Express 34775

Fig. 3. (a) 3D schematic diagram of GeSn/Ge MQW p-i-n photodiode with double-mesa
structure. (b) Top-view SEM image of fabricated GeSn/Ge MQW photodiode.

3. Characterizations of photodiode as photodetector

Dark current is one of the pivotal metrics of photodiodes. As shown in Fig. 4(a), dark current
was measured from -1 to 1 V on photodiodes with varying mesa diameters. For the photodiode
with mesa diameter of 60 µm, high on/off ratio around 5×104 was achieved, manifesting excellent
rectifying behavior. Low dark current density around 16.3 mA/cm2 was achieved on a photodiode
with mesa diameter of 250 µm. As benchmarked in Fig. 5 [1,3,25–39], such value is among
the lowest in GeSn photodiodes. The dark current density usually increases with Sn fraction
since higher Sn composition means higher lattice mismatch. To the best of our knowledge, the
achieved dark current density is the lowest for GeSn photodiode with Sn composition larger than
4%. The low dark current density is attributed to the low TDD in GeSn/Ge MQW because the
GeSn layer is pseudomorphic without strain relaxation. In order to investigate the attribution of
dark current density, surface leakage and bulk leakage current densities were extracted according

Fig. 4. (a) Dark current characteristics of the GeSn/Ge MQW photodiodes with varying
mesa diameters. (b) Linear fitting of dark current density versus 4/D for GeSn/Ge MQW
photodiodes allows for extraction of bulk and surface leakage densities. (c) Surface leakage
percentages of fabricated photodiodes with varying mesa diameters.
 Research Article Vol. 28, No. 23 / 9 November 2020 / Optics Express 34776

to the following equation:

4
Jdark = Jbulk + Jsurf (1)
D
where D is the mesa diameter and 4/D is known as perimeter-to-area ratio. The extracted
bulk leakage and surface leakage current densities from linear interpolation in Fig. 4(b) are
15.98 mA/cm2 and 2.13 µA/cm, respectively. The surface leakage mainly originates from minority
carriers and current flows on the sidewall, while the bulk leakage is related to material defects.
The percentages of surface leakage were plotted in Fig. 4(c). The photodiode with smaller mesa
diameter has higher surface leakage percentage due to increased perimeter-to-area ratio. Owing
to effective passivation methods, surface leakage percentages for all photodiodes are smaller than
10%. Even though the MQW structure is beneficial for low TDD, bulk leakage still dominates
dark current. The bulk leakage probably derives from the migration of threading dislocation at
Ge/Si interface. For group IV material, lattice mismatch is always an intractable issue and the
degraded performance as compared with III-V material is mainly rooted on it.

Fig. 5. Benchmarking of the dark current density for GeSn photodiodes grown on Si
substrate at bias voltage of -1 V.

The dark current characteristic of the photodiode can be expressed using the Shockley equation:

q(V − Rs Inet ) V
I = I0 exp[ ]+ (2)
nkT Rsh
where I 0 is reverse saturation current, n is ideality factor, Rs and Rsh are series resistance and
shunt resistance respectively. Near 0 V, the equation can be simplified as I = I0 +V/Rsh and Rsh can
be extracted by dV/dI. The shunt resistances at room temperature (T=300 K) for photodiode with
mesa diameters of 60 and 250 µm are 111 and 82.2 kΩ, respectively. The high shunt resistances
achieved are consistent with low surface leakage density and further indicate the effective surface
passivation. After obtaining shunt resistance, the ideality factor n and series resistance Rs can be
extracted using the transformed expression: dV/dInet =(nkT/q)Inet −1 +Rs where the net current
flow Inet can be calculated using Inet = I − V/Rsh . As shown in Fig. 6, the extracted ideality
factors for both photodiodes are around 1.29 which is close to an ideal pn junction. The extracted
series resistances are 26.4 and 13.4 Ω for photodiode with mesa diameters of 60 and 250 µm,
respectively. The low series resistance is beneficial for high speed operation due to the reduced
resistance-capacitance (RC) delay.
 Research Article Vol. 28, No. 23 / 9 November 2020 / Optics Express 34777

Fig. 6. Linear fitting of Inet −1 versus dV/dInet allows for the extraction of series resistance
Rs and ideality factor n for GeSn/Ge MQW photodiodes with mesa diameter of 60 and 250
µm.

Photo current was characterized at two essential communication bands: 1,550 and 2,000 nm,
respectively. Before the measurement, the propagation loss in the fibers and the coupling loss
in the connectors were calibrated. As shown in Fig. 7(a), the calibrated optical power varies
from 0.73 to 5.14 mW at 1,550 nm. The flat response curves under reverse bias suggest the
effective collection of photon-generated carriers. The relationship between optical power and
photo current at -1 V was plotted in Fig. 8(a). Relatively high responsivity around 0.307 A/W
was achieved by linear interpolation. Similarly, the optical response of the photodiode at 2 µm
was measured and extrapolated in Fig. 7(b) and Fig. 8(b). The slight deviation in the linear
interpolation probably attributes to the increased loss between input fiber and photodiode surface.

Fig. 7. Current-voltage (I-V) characteristics of the GeSn/Ge MQW photodiode at incident

wavelength of (a) 1,550 nm with varying optical power from 0.73 to 5.14 mW and (b) 2 µm
with varying optical power from 1.19 to 23.21 mW.
 Research Article Vol. 28, No. 23 / 9 November 2020 / Optics Express 34778

At 2 µm, the extracted responsivity is 0.007 A/W. The optical performance manifests that the
GeSn/Ge MQW photodiode is capable of photo detection from 1,550 to 2,000 nm.

Fig. 8. Linear fitting of photocurrent versus optical power at (a) 1,550 nm and (b) 2 µm.

The relatively low responsivity at 2 µm is reasonable because the total thickness of absorptive
GeSn layers in the MQW region is only 112.5 nm. Moreover, the bandgap of GeSn layer in MQW
structure increases due to the compressive strain and quantum confined effect. The band structure
of the GeSn/Ge MQW was calculated using single band approximation. In the calculation,
the Ge layer was treated as fully relaxed and the in-plain and normal strain in GeSn layer is
around -0.0116 and 0.0088. The energy shifts introduced by strain were modeled by deformation
potential theory. Taking the energy shift caused by strain into consideration, the band offsets for
Γ valley, L valley, heavy hole and light hole bands were calculated, respectively. The eigenvalues
and wave functions were obtained by solving one dimensional time independent Schrödinger’s
equation:
h2 ∂ 2 φn
− 2 ∗ + V(z)φn = En φn (3)
8π m ∂z2
where the index n represents the nth band state in the quantum well, V(z) is the band offset
obtained in the previous calculation and m* is the effective carrier mass for different energy
bands. The calculation results were plotted in Fig. 9. The direct bandgap (EΓ1 -Ehh1 ) in the GeSn
well region is around 0.66 eV which corresponds to a wavelength of 1,880 nm. It explains the
relatively low responsivity at 2 µm since weak indirect bandgap absorption is dominant at this
wavelength.
The specific detectivity D* was calculated to compare the performance of GeSn/Ge MQW
photodiode with commercial Ge and extended-InGaAs photodetectors operating at similar
wavelength range in Fig. 10. Due to the low dark current in pseudomorphic GeSn/Ge MQW
region, the shot noise current achieved in this work is around 2.15×10−13 A·Hz−1/2 when the
input optical power is close to 0. It should be noted that all the measurements performed in this
work were at room temperature (T=300 K). The calculated specific detectivity D* at 1,550 nm is
1.37×1010 cm·Hz1/2 /W which is two orders higher than bulk Ge0.9 Sn0.1 photodetector in Ref.
[31]. The achieved specific detectivity at C-band makes the GeSn/Ge MQW photodiode in this
work competitive as compared with commercial Ge and extended-InGaAs photodetectors. At
2 µm, the specific detectivity is 3.20×108 cm·Hz1/2 /W due to the relatively low responsivity as
 Research Article Vol. 28, No. 23 / 9 November 2020 / Optics Express 34779

Fig. 9. Band structure between Ge barrier layer and Ge0.92 Sn0.08 well layer calculated
using single band model. Ground states in each bands were represented using dashed lines.

explained before. The relatively low detectivity at 2 µm can be optimized by increasing the Sn
concentration to reduce the bandgap of GeSn or increasing the thickness of GeSn well region to
weaken the quantum confined effect. It is desirable to realize high detectivity across the entire
range from 1,550 to 2,000 nm using the optimized GeSn/Ge MQW structure in the future.

Fig. 10. Comparison of specific detectivity D* of fabricated GeSn/Ge p-i-n photodiode

with commercial extended-InGaAs photodetector, commercial Ge photodetector and bulk
Ge0.9 Sn0.1 photodetector [31] from 1,500 to 2,000 nm.

4. Characterizations of photodiode as modulator

To investigate the modulation effect in the GeSn/Ge MQW photodiode, the spectral response was
obtained from 1,700 to 2,200 nm at varying reverse bias voltages in Fig. 11. Since the calculated
direct bandgap of the GeSn well region is 0.66 eV (1,880 nm), the measured wavelength range has
 Research Article Vol. 28, No. 23 / 9 November 2020 / Optics Express 34780

covered the direct bandgap. It can be observed that the spectral response is bias voltage-dependent.
At a reverse bias voltage of 2 V, the achieved responsivities are the highest across the entire
measurement range. In terms of reverse bias voltages of 0.5, 1, 1.5 V, there is a cross-over region
between 1,820 to 1,860 nm where the responsivity does not increase with bias voltage. In order
to understand the physical mechanism behind the spectral response, absorption coefficient of
GeSn layer in the MQW region was extracted according to the following equation:
eλ −αn−Ge tn−Ge
Ropt = (1 − R) e (1 − e−αGe tGe e−αGeSn tGeSn ) (4)
hc
where Ropt is responsivity, R is reflection, and αGe and αGeSn are the absorption coefficient of
Ge and GeSn respectively. The absorption coefficient for n+ doped Ge region is assumed as
same as intrinsic Ge region and αGe was obtained using ellipsometer on a bulk Ge wafer. The
assumption is reasonable since the absorption coefficient of Ge is very small (< 100/cm) at this
wavelength range. The reflection spectrum was measured using microscope Fourier-transform
infrared spectroscopy (FTIR) in Fig. 12. Standard gold film was used to calibrate the reflection
in the measurement. It can be observed that resonance behavior exists in the reflection spectrum.
The resonance peaks are attributed to air-Ge-Si cavity structure since the refractive index of
Ge0.92 Sn0.08 is close to Ge. According to the position of resonance peaks, the total thickness
of Ge layer (device layer + Ge buffered layer) is estimated to be 2.48 µm which is close to the
proposed thickness in the material growth.

Fig. 11. Spectral response of GeSn/Ge MQW photodiode from 1,700 to 2,200 nm at reverse
bias voltage of 0, 0.5, 1, 1.5 and 2 V respectively.

The absorption coefficient spectral of GeSn layer in the MQW region were extracted from
0.57 to 0.73 eV at varying voltages in Fig. 13(a). It can be observed that the absorption spectral
can be divided into three regions: indirect bandgap absorption, Urbach tail and direct bandgap
absorption. Direct bandgap and indirect bandgap absorption are two typical inter-band transition
in semiconductors. Urbach tail is related to the transition between band tail states such as doping
and defects energy states which are exactly above the valence band or below the conduction band
[40]. The absorption from Urbach tail could be dominant when the Eg indirect < hν < Eg direct
since the indirect bandgap absorption is very weak. However, it is noted that the direct bandgap
absorption region is not as smooth as expected. Small peak was observed in the direct bandgap
absorption regions at all reverse bias voltages. In terms of direct bandgap transition, it satisfies
 Research Article Vol. 28, No. 23 / 9 November 2020 / Optics Express 34781

Fig. 12. Micro-FTIR reflection spectrum of fabricated GeSn/Ge MQW photodiode from
1,700 to 2,200 nm.

the following equation near the band edge:

(αhν)2 = A(hν − Eg ) (5)
where A is a constant and hν is the energy of incident photons. Using this equation, the direct
bandgap of GeSn layer was extrapolated at 0 V in Fig. 13(b). The extracted direct bandgap in the
quantum well is around 0.693 eV which is close to the calculated value of 0.660 eV before.

Fig. 13. (a) Extracted absorption coefficient spectral of GeSn layer in the MQW region at
varying reverse bias voltage. (b) Linear fitting of (αhν)2 versus photon energy hν allows for
extraction of direct bandgap of GeSn well layer at 0 V.

In order to investigate the mechanism behind the small peak in the direct bandgap absorption
region, GeSn/Ge MQW layer was treated as a whole and the photon absorption was separated
into two parts: absorption within and outside the depletion region. Therefore, the effective
absorption coefficient of GeSn/Ge MQW layer can be extrapolated using the following equation:
eλ −αn−Ge tn−Ge
Ropt = (1 − R) e (1 − e−αv td e−α0 (ti −td ) ) (6)
hc
 Research Article Vol. 28, No. 23 / 9 November 2020 / Optics Express 34782

where td is the thickness of depletion region, ti is the total thickness of the intrinsic MQW region
and α0 and αv are effective absorption coefficients of GeSn/Ge MQW at 0 V and any other reverse
bias V. The thickness of depletion region can be obtained through C-V measurement by d=ε i A/Cj
because the photodiode can be regarded as a parallel capacitor. The total capacitance C obtained
from C-V measurement is composed of parasitic capacitance Cp and junction capacitance Cj .
Since the junction capacitance is proportional to the mesa area, the parasitic capacitance Cp =4.69
pF can be extracted from the linear interpolation in Fig. 14.

Fig. 14. Linear fitting of total capacitance versus mesa area allows for extraction of parasitic
capacitance Cp .

After obtaining junction capacitance and thickness of depletion region, the effective absorption
coefficient of GeSn/Ge MQW was calculated and plotted in Fig. 15. It is noteworthy that small
peaks close to the direct bandgap of GeSn well layer can be observed in the absorption spectral.
The small peaks probably originate from the exciton absorption in the MQW structure due to
QCSE effect. The binding energy of exciton in the quantum well can be described using the
equation:
e4 m e mh 1
n
Eexciton = 2 2( ) (7)
8ε h m e + m h (n − 1/2)
2

where me and mh are effective masses of electron and hole respectively. For the excitons confined
at Γ valley and heavy hole band, the parameters used here are consistent with band structure
calculation: me =0.0316 m0 and mh =0.223 m0 . The calculated binding energy is around 5.63 meV
which roughly satisfies the energy difference Eexciton = Eqw_direct - Ebinding . Moreover, the peak
position shifts to lower photon energy with increasing reverse bias voltage which is also one of
the characteristics of QCSE effect. In the GeSn/Ge MQW structure, the well region is 7.5 nm
which provides relatively sufficient confinement for the exciton. However, the exciton absorption
is weak in this work due to the following two possible reasons. One of the reasons is that the
material defects in the GeSn/Ge MQW region is not as low as expected. Even though the GeSn
layer is pseudomorphic, the defects at Ge/Si interface can still propagate into the device region
which is also the cause of large percentage of bulk leakage as discussed before. Such problem can
be solved by direct wafer bonding (DWB) process after the material growth [41]. The defective
Ge buffer layer can be etched away by transferring the sample on an insulator platform. The
other reason could be related to the measurement limitation. The resolution of optical filter is
10 nm and the applied reverse bias voltage is relatively low. Since the total thickness of the MQW
region is 412.5 nm, the electric field density in the MQW region is not sufficiently strong to bring
significant shift of the exciton peak. In addition, all the measurements were performed at room
temperature. The thermal perturbation at room temperature (kT = 25.8 meV) is destructive for
 Research Article Vol. 28, No. 23 / 9 November 2020 / Optics Express 34783

the excitons due to the low binding energy. Even though the exciton peak in this work is not
remarkable, it is the first time that QCSE effect was discovered in GeSn/Ge system. With further
optimization in the material growth or switching the barrier material to SiGeSn, more prominent
QCSE effect and high efficiency modulation based on exciton absorption can be realized.

Fig. 15. Extracted effective absorption coefficient of GeSn/Ge MQW region at varying
reverse bias voltage from 0.57 to 0.73 eV.

To achieve effective optical modulation, the deviation of absorption coefficient for a given
reverse bias voltage is pivotal. The change of absorption coefficient can be evaluated by absorption
ratio which is defined as ∆α/α0 . Here, the effective absorption coefficient of the GeSn/Ge MQW
was used in the calculation. As shown in Fig. 16, the absorption ratios at all bias voltages decrease
first and then increase. When the reverse bias voltage is lower than 2 V, the absorption ratio turns

Fig. 16. Change of absorption ratio of GeSn/Ge MQW region at varying reverse bias from
1,700 to 2,000 nm.
 Research Article Vol. 28, No. 23 / 9 November 2020 / Optics Express 34784

into negative around 1,830 nm and then returns to positive value at longer wavelength. At voltage
of -2 V, the absorption ratio remains positive across the entire range from 1,700 to 2,000 nm and
the maximum value is around 1.81 at 2 µm, showing promise for optical modulation at 2 µm
band.
The 3-dB bandwidth of GeSn/Ge MQW photodiode was calculated theoretically in Fig. 17. In
the calculation, both transit-limited bandwidth and RC delay-limited bandwidth were taken into
consideration. By assuming the velocity of electron in depletion region has reached the saturation
velocity vsat =6×106 cm/s, the transit-limited bandwidth can be calculated using fT = 0.45vsat /di
[42]. The RC delay limited bandwidth is dominant for device with large surface area and it can be
expressed by fRC =1/(2π(Rs +Rl )(Cj +Cp )). The load resistance Rl was estimated to be 50 Ω and
the values of series resistance Rs , junction capacitance Cj and parasitic capacitance Cp had been
obtained in the former discussion. As shown in Fig. 17, the theoretical 3-dB bandwidth for the
photodiode with mesa diameter of 30 µm can reach 12 GHz at -2 V. The theoretical calculation
could overestimate the 3-dB bandwidth of the photodiode since the carrier trapping effect in the
MQW was not taken into consideration. However, the carrier trapping effect will be alleviated at
higher bias voltage because of narrower barrier. Therefore, the GeSn/Ge photodiode fabricated
in this work is believed to be capable of high frequency operation.

Fig. 17. Theoretically calculated 3-dB bandwidth of fabricated GeSn/Ge MQW photodiode
as a function of mesa diameter. The blue dashed line represents transit-limited bandwidth.

5. Conclusion
The photo detection and modulation performance of GeSn/Ge MQW p-i-n photodiode was
investigated. The GeSn/Ge MQW p-i-n structure was grown on a 300-mm Ge-buffered Si
substrate by RPCVD which is promising for large-scale production. High material quality with
uniform and pseudomorphic GeSn well layers was verified by TEM and XRD characterization.
As a photodetector, the fabricated GeSn/Ge MQW photodiode shows an ultralow dark current of
16.3 mA/cm2 at -1 V and a high responsivity of 0.307 A/W at 1,550 nm. The achieved specific
detectivity at C-band (centered at 1,550 nm) is 1.37×1010 cm·Hz1/2 /W which is comparable
with commercial Ge and extended-InGaAs photodetectors. The potential of GeSn/Ge MQW
photodiode as optical modulator was also investigated from 1,700 to 2,200 nm. QCSE effect
was observed in the effective absorption coefficient spectra of the GeSn/Ge MQW region. The
exciton peak shifts from 0.690 to 0.688 eV when the reverse bias voltage increases from 0 to 2 V.
The achieved absorption ratio under -2 V at 2 µm is 1.81, manifesting the fabricated GeSn/Ge
MQW can be used as a modulator operating at 2 µm. This work paves a way for the application
 Research Article Vol. 28, No. 23 / 9 November 2020 / Optics Express 34785

of GeSn/Ge MQW photodiodes as photo detectors and optical modulators in the optoelectronic
integrated circuits.

Funding
National Research Foundation Singapore (NRF–CRP19–2017–01); Ministry of Education -
Singapore (R-263-000-D45-112); Ministry of Education - Singapore (R-263-000-C58-133).

Acknowledgments
The authors acknowledge Dr. Chong Gang Yih and Ms. Ngo Ling Ling in Nanyang Technological
University for the help in sputtering and PECVD.

Disclosures
The authors declare that there are no conflicts of interest related to this article.

References
1. J. Mathews, R. Roucka, J. Xie, S.-Q. Yu, J. Menéndez, and J. Kouvetakis, “Extended performance GeSn/Si (100)
p-i-n photodetectors for full spectral range telecommunication applications,” Appl. Phys. Lett. 95(13), 133506 (2009).
2. J. Werner, M. Oehme, M. Schmid, M. Kaschel, A. Schirmer, E. Kasper, and J. Schulze, “Germanium-tin pin
photodetectors integrated on silicon grown by molecular beam epitaxy,” Appl. Phys. Lett. 98(6), 061108 (2011).
3. S. Su, B. Cheng, C. Xue, W. Wang, Q. Cao, H. Xue, W. Hu, G. Zhang, Y. Zuo, and Q. Wang, “GeSn pin photodetector
for all telecommunication bands detection,” Opt. Express 19(7), 6400–6405 (2011).
4. H. Tran, T. Pham, J. Margetis, Y. Zhou, W. Dou, P. C. Grant, J. M. Grant, S. Alkabi, W. Du, and G. Sun, “Study of
High Performance GeSn Photodetectors with Cutoff Wavelength Up to 3.7 µm for Low-Cost Infrared Imaging,” in
2019 Conference on Lasers and Electro-Optics (CLEO), (IEEE, 2019), 1–2.
5. H. Zhang, N. Kavanagh, Z. Li, J. Zhao, N. Ye, Y. Chen, N. Wheeler, J. Wooler, J. Hayes, and S. Sandoghchi, “100
Gbit/s WDM transmission at 2 µm: transmission studies in both low-loss hollow core photonic bandgap fiber and
solid core fiber,” Opt. Express 23(4), 4946–4951 (2015).
6. P. Roberts, F. Couny, H. Sabert, B. Mangan, D. Williams, L. Farr, M. Mason, A. Tomlinson, T. Birks, and J. Knight,
“Ultimate low loss of hollow-core photonic crystal fibres,” Opt. Express 13(1), 236–244 (2005).
7. Z. Li, A. Heidt, J. Daniel, Y. Jung, S. Alam, and D. J. Richardson, “Thulium-doped fiber amplifier for optical
communications at 2 µm,” Opt. Express 21(8), 9289–9297 (2013).
8. Z. Li, A. Heidt, N. Simakov, Y. Jung, J. Daniel, S. Alam, and D. Richardson, “Diode-pumped wideband thulium-doped
fiber amplifiers for optical communications in the 1800–2050nm window,” Opt. Express 21(22), 26450–26455
(2013).
9. H. Zhang, Z. Li, N. Kavanagh, J. Zhao, N. Ye, Y. Chen, N. Wheeler, J. Wooler, J. Hayes, and S. Sandoghchi, “81 Gb/s
WDM transmission at 2µm over 1.15 km of low-loss hollow core photonic bandgap fiber,” in 2014 The European
Conference on Optical Communication (ECOC), (IEEE, 2014), 1–3.
10. D. J. Richardson, “Filling the light pipe,” Science 330(6002), 327–328 (2010).
11. S. Xu, K. Han, Y.-C. Huang, K. H. Lee, Y. Kang, S. Masudy-Panah, Y. Wu, D. Lei, Y. Zhao, and H. Wang, “Integrating
GeSn photodiode on a 200 mm Ge-on-insulator photonics platform with Ge CMOS devices for advanced OEIC
operating at 2 µm band,” Opt. Express 27(19), 26924–26939 (2019).
12. H. Zhou, S. Xu, Y. Lin, Y.-C. Huang, B. Son, W. Li, X. Guo, L. Liu, K. H. Lee, and X. Gong, “High-efficiency photo
detection at 2 µm realized by GeSn/Ge multiple-quantum-well photodetectors with photon-trapping microstructure,”
in CLEO: Science and Innovations, (Optical Society of America, 2020), STh4L. 1.
13. H. Tran, T. Pham, W. Du, Y. Zhang, P. C. Grant, J. M. Grant, G. Sun, R. A. Soref, J. Margetis, and J. Tolle, “High
performance Ge0. 89Sn0. 11 photodiodes for low-cost shortwave infrared imaging,” J. Appl. Phys. 124(1), 013101
(2018).
14. H. Tran, T. Pham, J. Margetis, Y. Zhou, W. Dou, P. C. Grant, J. M. Grant, S. Al-Kabi, G. Sun, and R. A. Soref,
“Si-based GeSn photodetectors toward mid-infrared imaging applications,” ACS Photonics 6(11), 2807–2815 (2019).
15. H. Zhou, S. Xu, Y. Lin, Y.-C. Huang, B. Son, Q. Chen, X. Guo, K. H. Lee, S. C.-K. Goh, and X. Gong, “High-efficiency
GeSn/Ge multiple-quantum-well photodetectors with photon-trapping microstructures operating at 2 µm,” Opt.
Express 28(7), 10280–10293 (2020).
16. S. Wirths, R. Geiger, N. Von Den Driesch, G. Mussler, T. Stoica, S. Mantl, Z. Ikonic, M. Luysberg, S. Chiussi, and J.
Hartmann, “Lasing in direct-bandgap GeSn alloy grown on Si,” Nat. Photonics 9(2), 88–92 (2015).
17. M. Oehme, K. Kostecki, M. Schmid, M. Kaschel, M. Gollhofer, K. Ye, D. Widmann, R. Koerner, S. Bechler, and E.
Kasper, “Franz-Keldysh effect in GeSn pin photodetectors,” Appl. Phys. Lett. 104(16), 161115 (2014).
18. S. Bechler, M. Oehme, O. Latzel, M. Schmid, K. Kostecki, R. Koerner, M. Gollhofer, E. Kasper, and J. Schulze,
“Franz-Keldysh Effect in GeSn Detectors,” ECS Trans. 64(6), 383–390 (2014).
 Research Article Vol. 28, No. 23 / 9 November 2020 / Optics Express 34786

19. A. Elbaz, D. Buca, N. von den Driesch, K. Pantzas, G. Patriarche, N. Zerounian, E. Herth, X. Checoury, S. Sauvage,
and I. Sagnes, “Ultra-low-threshold continuous-wave and pulsed lasing in tensile-strained GeSn alloys,” Nat. Photonics
14(6), 375–382 (2020).
20. D. Stange, N. Von Den Driesch, D. Rainko, C. Schulte-Braucks, S. Wirths, G. Mussler, A. Tiedemann, T. Stoica, J.
Hartmann, and Z. Ikonic, “Study of GeSn based heterostructures: towards optimized group IV MQW LEDs,” Opt.
Express 24(2), 1358–1367 (2016).
21. Y. Zhou, Y. Miao, S. Ojo, H. Tran, G. Abernathy, J. M. Grant, S. Amoah, G. Salamo, W. Du, and J. Liu, “Electrically
injected GeSn lasers on Si operating up to 100 K,” arXiv preprint arXiv:2004.09402 (2020).
22. B. Son, Y. Lin, K. H. Lee, Q. Chen, and C. S. Tan, “Dark current analysis of germanium-on-insulator vertical pin
photodetectors with varying threading dislocation density,” J. Appl. Phys. 127(20), 203105 (2020).
23. Y.-H. Kuo, Y. K. Lee, Y. Ge, S. Ren, J. E. Roth, T. I. Kamins, D. A. Miller, and J. S. Harris, “Strong quantum-confined
Stark effect in germanium quantum-well structures on silicon,” Nature 437(7063), 1334–1336 (2005).
24. Y.-H. Kuo, Y. K. Lee, Y. Ge, S. Ren, J. E. Roth, T. I. Kamins, D. A. Miller, and J. S. Harris, “Quantum-confined
Stark effect in Ge/SiGe quantum wells on Si for optical modulators,” IEEE J. Sel. Top. Quantum Electron. 12(6),
1503–1513 (2006).
25. S. Xu, Y.-C. Huang, K. H. Lee, W. Wang, Y. Dong, D. Lei, S. Masudy-Panah, C. S. Tan, X. Gong, and Y.-C. Yeo,
“GeSn lateral pin photodetector on insulating substrate,” Opt. Express 26(13), 17312–17321 (2018).
26. H. Tseng, H. Li, V. Mashanov, Y. Yang, H. Cheng, G. Chang, R. Soref, and G. Sun, “GeSn-based pin photodiodes
with strained active layer on a Si wafer,” Appl. Phys. Lett. 103(23), 231907 (2013).
27. M. Oehme, K. Kostecki, K. Ye, S. Bechler, K. Ulbricht, M. Schmid, M. Kaschel, M. Gollhofer, R. Körner, and W.
Zhang, “GeSn-on-Si normal incidence photodetectors with bandwidths more than 40 GHz,” Opt. Express 22(1),
839–846 (2014).
28. M. Oehme, D. Widmann, K. Kostecki, P. Zaumseil, B. Schwartz, M. Gollhofer, R. Koerner, S. Bechler, M. Kittler,
and E. Kasper, “GeSn/Ge multiquantum well photodetectors on Si substrates,” Opt. Lett. 39(16), 4711–4714 (2014).
29. Y.-H. Peng, H. Cheng, V. I. Mashanov, and G.-E. Chang, “GeSn pin waveguide photodetectors on silicon substrates,”
Appl. Phys. Lett. 105(23), 231109 (2014).
30. Y. Dong, W. Wang, D. Lei, X. Gong, Q. Zhou, S. Y. Lee, W. K. Loke, S.-F. Yoon, E. S. Tok, and G. Liang,
“Suppression of dark current in germanium-tin on silicon pin photodiode by a silicon surface passivation technique,”
Opt. Express 23(14), 18611–18619 (2015).
31. T. Pham, W. Du, H. Tran, J. Margetis, J. Tolle, G. Sun, R. A. Soref, H. A. Naseem, B. Li, and S.-Q. Yu, “Systematic
study of Si-based GeSn photodiodes with 2.6 µm detector cutoff for short-wave infrared detection,” Opt. Express
24(5), 4519–4531 (2016).
32. Y.-H. Huang, G.-E. Chang, H. Li, and H. Cheng, “Sn-based waveguide pin photodetector with strained GeSn/Ge
multiple-quantum-well active layer,” Opt. Lett. 42(9), 1652–1655 (2017).
33. M. Morea, C. E. Brendel, K. Zang, J. Suh, C. S. Fenrich, Y.-C. Huang, H. Chung, Y. Huo, T. I. Kamins, and K. C.
Saraswat, “Passivation of multiple-quantum-well Ge0. 97Sn0. 03/Ge pin photodetectors,” Appl. Phys. Lett. 110(9),
091109 (2017).
34. Y. Dong, W. Wang, S. Xu, D. Lei, X. Gong, X. Guo, H. Wang, S.-Y. Lee, W.-K. Loke, and S.-F. Yoon, “Two-micron-
wavelength germanium-tin photodiodes with low dark current and gigahertz bandwidth,” Opt. Express 25(14),
15818–15827 (2017).
35. J. Mathews, R. Roucka, C. Weng, R. Beeler, J. Tolle, J. Menéndéz, and J. Kouvetakis, “Near IR photodiodes with
tunable absorption edge based on Ge1-y Sny alloys integrated on silicon,” ECS Trans. 33(6), 765–773 (2019).
36. M. Oehme, M. Schmid, M. Kaschel, M. Gollhofer, D. Widmann, E. Kasper, and J. Schulze, “GeSn pin detectors
integrated on Si with up to 4% Sn,” Appl. Phys. Lett. 101(14), 141110 (2012).
37. W. Wang, S. Vajandar, S. L. Lim, Y. Dong, V. R. D’Costa, T. Osipowicz, E. S. Tok, and Y.-C. Yeo, “In-situ
gallium-doping for forming p+ germanium-tin and application in germanium-tin pin photodetector,” J. Appl. Phys.
119(15), 155704 (2016).
38. H. Cong, C. Xue, J. Zheng, F. Yang, K. Yu, Z. Liu, X. Zhang, B. Cheng, and Q. Wang, “Silicon based GeSn pin
photodetector for SWIR detection,” IEEE Photonics J. 8(5), 1–6 (2016).
39. S. Xu, W. Wang, Y.-C. Huang, Y. Dong, S. Masudy-Panah, H. Wang, X. Gong, and Y.-C. Yeo, “High-speed photo
detection at two-micron-wavelength: technology enablement by GeSn/Ge multiple-quantum-well photodiode on
300 mm Si substrate,” Opt. Express 27(4), 5798–5813 (2019).
40. H. Tran, W. Du, S. A. Ghetmiri, A. Mosleh, G. Sun, R. A. Soref, J. Margetis, J. Tolle, B. Li, and H. A. Naseem,
“Systematic study of Ge1− xSnx absorption coefficient and refractive index for the device applications of Si-based
optoelectronics,” J. Appl. Phys. 119(10), 103106 (2016).
41. K. H. Lee, S. Bao, Y. Wang, E. A. Fitzgerald, and C. Seng Tan, “Suppression of interfacial voids formation during
silane (SiH4)-based silicon oxide bonding with a thin silicon nitride capping layer,” J. Appl. Phys. 123(1), 015302
(2018).
42. B. Son, Y. Lin, K. H. Lee, Y. Wang, S. Wu, and C. S. Tan, “High speed and ultra-low dark current Ge vertical pin
photodetectors on an oxygen-annealed Ge-on-insulator platform with GeO x surface passivation,” Opt. Express
28(16), 23978–23990 (2020).