TMP DEA2

Photonics 2014, 1, 162-197; doi:10.
3390/photonics1030162
OPEN ACCESS
photonics
ISSN 2304-6732
www.mdpi.com/journal/photonics
Review
Monolithically Integrated Ge-on-Si Active Photonics

Jifeng Liu
Thayer School of Engineering, Dartmouth College, 14 Engineering Drive, Hanover, NH 03755, USA;
E-Mail: [email protected]; Tel.: +1-603-646-9885; Fax: +1-603-646-8778
Received: 27 May 2014; in revised form: 19 June 2014 / Accepted: 19 June 2014 /
Published: 2 July 2014
Abstract: Monolithically integrated, active photonic devices on Si are key components in

Si-based large-scale electronic-photonic integration for future generations of high-performance,
low-power computation and communication systems. Ge has become an interesting
candidate for active photonic devices in Si photonics due to its pseudo-direct gap behavior
and compatibility with Si complementary metal oxide semiconductor (CMOS) processing.
In this paper, we present a review of the recent progress in Ge-on-Si active photonics
materials and devices for photon detection, modulation, and generation. We first discuss the
band engineering of Ge using tensile strain, n-type doping, Sn alloying, and separate
confinement of Γ vs. L electrons in quantum well (QW) structures to transform the material
towards a direct band gap semiconductor for enhancing optoelectronic properties. We then
give a brief overview of epitaxial Ge-on-Si materials growth, followed by a summary of
recent investigations towards low-temperature, direct growth of high crystallinity Ge and
GeSn alloys on dielectric layers for 3D photonic integration. Finally, we review the most
recent studies on waveguide-integrated Ge-on-Si photodetectors (PDs), electroabsorption
modulators (EAMs), and laser diodes (LDs), and suggest possible future research directions
for large-scale monolithic electronic-photonic integrated circuits on a Si platform.
Keywords: Ge; GeSi; GeSn; tensile strain; band engineering; integrated silicon photonics;
optoelectronics; photodetectors; modulators; lasers
1. Introduction
In our information age, information and communication technology (ICT) already consumes ~5% of
electricity worldwide in 2012 [1]. Considering that the data traffic has grown 30× from 2010–2015 [2],
this number will increase to >10% by 2020 in developed countries [3] and become a significant part of
Photonics 2014, 1 163
energy consumption in the 21st century. More than half of the energy is consumed by interconnects
that transfer information from one place to another, rather than computation itself. Electronic-photonic
integrated circuits (EPICs) on Si have become an increasingly appealing solution to this issue [4,5].
They combine the merits of photons in energy-efficient, high bandwidth data transmission with the
merits of electrons in high-speed data processing on a single microchip, providing the most efficient
synergy for Green ICT [6].
A significant challenge for high-volume, large-scale EPICs on Si is active photonic devices for light
generation, modulation and detection. These devices are conventionally based on direct band gap III-V
semiconductors to benefit from the efficient direct gap transitions. However, epitaxial growth of
high-quality III-V semiconductors on Si typically requires a thick buffer layer due to the lattice
mismatch, making it difficult to integrate with complementary metal oxide semiconductor (CMOS)
circuits (Figure 1a). Hybrid integration of III-V devices on Si by chip bonding has achieved significant
progress in recent years, yet scaling to high-volume, wafer-scale processing is limited by the
difficulties in incorporating III-V device fabrication into a standard CMOS processing flow.
Figure 1. Schematics of hybrid photonic integration vs. monolithic Si photonic integration.

Hybrid integration with III-V semiconductors includes (a) III-V epitaxy on Si, and
(b) chip-bonding of III-V on Si. Monolithic Si photonic integration includes (c) Planar
integration of Si photonics with complementary metal oxide semiconductor (CMOS)
circuits on the same layer, and (d) 3D photonic integration with the photonic interconnect
layer positioned well above the CMOS layer. In (c) and (d), waveguide-coupled Ge photo
detectors are shown as examples of CMOS-compatible, monolithic photonic devices. The
color bar on the left shows the processing temperature in different layers.
Active photonic devices based on Si itself have also been extensively studied in the past decade.
While Si electro-optical modulators based on carrier-induced refractive index changes have achieved
great success [7–9], the trade-off between operation wavelength range and energy consumption/thermal
stability is still hard to overcome [9,10]. Conventional Mach-Zehnder interferometer (MZI) structures
tend to have large footprints with high power consumption, while energy-efficient micro-resonator
modulators have limited operation wavelength range and tend to suffer from resonant peak shifts due
to thermo-optical effects. Si nanostructures have been investigated for light emission [11,12], and
defects states have been engineered to extend the absorption range of Si for photodetection at
telecommunication wavelengths [13,14]. However, the difficulty in electrical injection limits the
efficiency of Si nanostructure light emitting diodes (LEDs). While high electroluminescence efficiency
has been observed in Si nanocrystal-organic hybrid systems at very low injection levels; at elevated
injected current density >1 mA/cm2 the efficiency drops dramatically [15]. The spectrally extended Si
photodetectors (PDs) using engineered defect states suffer from a low absorption coefficient and a
trade-off in quantum efficiency particularly at CMOS-compatible driving voltages (<5 V).
To address the limitations of III-V semiconductors and stand-alone Si, since the late 1990s Ge has
become an interesting candidate for active photonic devices on Si platforms due to its pseudo-direct
gap behavior [16] and compatibility with Si CMOS technology [17]. Utilizing the efficient direct gap
transition of Ge and band engineering techniques to further reduce the energy difference between the
direct and indirect band gaps, high-performance Ge-on-Si PDs, electroabsorption modulators (EAMs)
and prototype laser diodes (LDs) have been demonstrated in recent years. While epitaxial Ge-on-Si
materials and devices comprise the main approach so far (Figure 1c), recent research is also exploring
the possibility of monolithic 3D integration using back-end-of-line (BEOL) CMOS processing (Figure 1d).
In this paper, we review recent progress in band engineering, materials growth and the design/performance
of integrated Ge active photonic devices on Si. We also propose some possible directions for future
investigations. The results demonstrate that Ge-on-Si is an ideal solution for monolithic, large-scale
electronic-photonic integration.
2. Band Structure and Band Engineering
The band structure of bulk Ge is schematically shown in Figure 2a. Although Ge is considered an
indirect gap semiconductor, the energy difference between its direct gap at the Γ valley and indirect
gap at the L valleys is only 136 meV [16], or ~5 kBT at room temperature (kB is the Boltzmann
constant). The direct band gap of 0.8 eV corresponds to a wavelength of 1550 nm, the most technically
important wavelength in optical communications. Active Ge photonic devices on Si are fundamentally
based on this direct gap transition. In this section, we will discuss several band-engineering approaches
that transform Ge towards a direct gap semiconductor.
2.1. Tensile Strain
The energy difference between the direct and indirect gaps of Ge can be further reduced by tensile
strain, as schematically shown in Figure 2b [18,19]. With biaxial tensile stress, both direct and indirect
gaps shrink, but the direct gap shrinks faster. Therefore, Ge transforms from an indirect towards a
direct gap semiconductor with the increase of tensile strain. Furthermore, the top of the valence band is
determined by the light hole band under biaxial tensile stress. The small effective mass of the light
hole band reduces the density of states in the valence band, which in turn decreases the threshold for
optical transparency and lasing. Theoretically, Ge can be transformed into a direct gap material with
~2% biaxial tensile strain [19–21] or >4% uniaxial tensile strain along <100> direction [22]. The band
gap decreases to ~0.5 eV correspondingly. Experimentally, a thermally induced tensile strain
of 0.2%–0.3% can be obtained in the Ge thin films utilizing the difference in thermal expansion
coefficients between Ge and Si [23–25]. This level of tensile strain reduces the difference between the
direct and indirect band gaps of Ge from 136 to ~100 meV, and redshifts the direct band gap from
0.8–0.76 eV. Several approaches have been investigated to further enhance the tensile strain in
epitaxial Ge. These include growths on relaxed GeSn buffer layers [26,27], utilizing silicide [28] and
nitride stressor layers [29,30], and coupling micro-electromechanical systems (MEMS) with suspended
thin Ge layers. Jain et al. [31] have demonstrated a prototype MEMS device with up to 1.4% uniaxial
and 1% biaxial tensile strain introduced to the Ge layer by a Si3N4 stressor. Compared to bulk Ge,
the photoluminescence (PL) peak redshifts to λ = 1860 nm (hν = 0.64 eV) and the integrated intensity
increased by 260 times. More recently, Süess et al. [32] achieved 3.1% uniaxial tensile strain in a
suspended Ge beam by patterning and under-etching 0.15% tensile strained Ge thin films grown on Si
and silicon-on-insulator (SOI) substrates. This approach utilizes the fact that, at a constant force in the
plane of the Ge films, the stress is inversely proportional to the cross-sectional area of the suspended
structures. Since the suspended beam can be patterned into a very small cross section, the stress in this
region is greatly enhanced compared to the blanket film regime. The direct band gap shrinks by 0.21 eV
to hν = 0.59 eV (λ = 2100 nm) in this case, and the PL intensity increases by 25 times. Further increase
in strain is only limited by the mechanical strength of Ge itself. Indeed, very lately the same group has
further achieved 4% uniaxial strain [33], while Sukhdeo et al. have reported 5.7% tensile strain along
<100> direction [34]. In addition, Capellini et al. have reported up to ~0.9% equivalent biaxial tensile
strain in a SiN(stressor)/Ge/SOI(001) stack using 130 nm bipolar-CMOS (Bi-CMOS) processing [35].
Figure 2. (a) Schematic band structure of bulk Ge, showing a 136 meV difference between
the direct gap and the indirect gap. (b) the difference between the direct and the indirect
gaps can be decreased by tensile strain, and (c) the rest of the difference between direct
and indirect gaps in tensile strained Ge can be compensated by filling electrons into the L
valleys via n-type doping. Because the energy states below the direct Γ valley in the
conduction band are fully occupied by extrinsic electrons from n-type doping, injected
electrons are forced into the direct Γ valley and recombine with holes, resulting in efficient
direct gap light emission [18].
(a) E (b) E (c) E

Injected
Conduction band
electrons
Γ
L Γ Γ L
L
0.800 eV
0.664 eV
hv electrons from
photons n-type doping
k k k
<111> <111> <111>
Injected
heavy hole band holes
Light hole band
bulk Ge tensile strained intrinsic Ge tensile strained n+ Ge

2.2. N-Type Doping
To obtain optical gain from the direct gap transition of Ge while maintaining the emission wavelength
in the C and L bands (1520–1630 nm), n-type doping has been combined with 0.2%–0.3% tensile strain
in order to compensate the energy difference between direct and indirect gaps (Figure 2c) [18,19].
In this case, the required n-type doping level is in the order of 1019–1020 cm−3. For larger tensile strain
the required n-doping level can be reduced as the energy difference between direct and indirect gaps
decreases. Since the lower energy states in indirect L valleys are already occupied by electrons from
n-type doping, upon carrier injection some injected electrons are forced to occupy the direct Γ valley
and recombine with holes radiatively via efficient direct transitions. Furthermore, the injected electrons
in the Γ valley are depleted much faster than those in the L valleys because the radiative recombination
rate of the direct transition is 4–5 orders of magnitude higher than that of the indirect transitions [36].
In order to maintain the quasi-equilibrium of electrons in the conduction band, the electrons initially
injected into in the L valleys will populate the Γ valley following inter-valley scattering, as shown by
the horizontal dashed arrow in Figure 2c. This process results in further radiative recombination via
efficient direct transitions. In terms of optical gain upon carrier injection, a major difference from the
intrinsic Ge case is that the Fermi level is already raised close to the Γ valley by n-type doping (Figure 2c)
so that a much smaller amount of injected carriers is needed to achieve population inversion. As a result,
free hole absorption, which dominates free carrier absorption (FCA) in Ge [37–39], is significantly
reduced and net gain is enabled.
2.3. Sn Alloying
In addition to tensile strain and n-type doping, alloying with Sn has been developed as another
approach to convert Ge into a direct band gap semiconductor [40–43]. The diamond cubic phase of Sn
is known as α-Sn, and it has a negative direct band gap of −0.41 eV since the s-like Γ7− conduction band
edge is 0.41 eV lower than the p-like Γ8+ valence band edge. On the other hand, the indirect gap at L
valley in α-Sn is ~0.1 eV [40,44]. Therefore, with the increase of Sn composition the energy of the
direct Γ valley decreases faster than the indirect L valleys, leading to indirect-to-direct gap transition.
Based on photoreflectance and ellipsometry measurements of GeSn alloy thin films deposited on Si by
chemical vapor deposition (CVD), it has been suggested that the Sn composition needed for indirect to
direct gap transition should not exceed 11 at % for relaxed GeSn [44]. The corresponding direct band
gap is ~0.5 eV at indirect-to-direct gap transition. PL results from GeSn samples deposited on
GaAs/InGaAs substrates by molecular beam epitaxy (MBE) have suggested that this transition occurs
at ~7 at % Sn for relaxed GeSn [45]. Overall, we can state that 8–10 at % Sn alloying is needed to
convert GeSn into direct gap semiconductor. While the equilibrium solubility of Sn in bulk Ge is only
1 at %, CVD [44], MBE [45,46] and direct crystallization of amorphous GeSn (a-GeSn) thin films [47]
all have achieved substitutional Sn compositions of ~9–10 at %. The last approach can be applied to
any amorphous dielectric layers towards monolithic 3D photonic integration of active photonic devices
on Si [47]. This method will be discussed in more detail in Section 3.2. It is also noteworthy that up to
25 at % Sn alloying has been demonstrated recently using MBE at a very low growth temperature of
120 °C [48]. The thickness of the Ge0.75Sn0.25 thin film is limited to 50 nm before the epitaxy breaks down.
2.4. Cocktail Material Band Engineering Approaches for Optimal Ge-Based Gain Media
Each of the three band engineering methods mentioned above has its limitations when used alone.
For the tensile strain engineering, it is very difficult to achieve 2% biaxial tensile strain, or 4% uniaxial
tensile strain that is required to transform Ge completely into a direct gap material. For n-type doping,
achieving an active dopant concentration >5 × 1019 cm−3 turns out to be challenging. For Sn alloying,
achieving device-quality GeSn with >10 at % Sn concentration remains an unresolved problem.
Therefore, the most promising strategy is to find a sweet spot that incorporates intermediate levels
of tensile strain, Sn alloying, and/or n-type doping, instead of stretching the fabrication capabilities
for one of the methods. As an example, Figure 3a maps the energy difference between the indirect and
direct gaps (EgL − EgΓ) as a function of tensile strain and Sn composition. In this calculation, the band
gaps of relaxed Ge1−xSnx are expressed as:
E gi ( x )  (1  x ) E gi (Ge )  xE gi ( Sn )  bi x (1  x ); i  L ,  (1)
where E g (Ge )  0.80 eV , E gL (Ge )  0.664 eV , E g ( Sn )  0.41 eV , and E gL ( Sn )  0.10 eV . We use
the bowing parameter b  2.0 eV for the direct gap [44,45], and the upper limit of bL  1.23 eV [44]
for the indirect gap as a conservative estimate of the indirect-to-direct gap transition. From Equation (1),
the contribution of Sn alloying to the reduction in (EgL − EgΓ) is
 Sn ( E gL  E g )  x (1.42  0.77 x )( eV ) (2)
We then approximate the deformation potentials of Ge1−xSnx (x ≤ 0.1) using the corresponding
parameters of pure Ge to account for the strain effect [24,49]. The reduction in (EgL − EgΓ) contributed
by biaxial tensile strain, ε, is given by:
 strain ( E gL  E g )  6.7ε(eV) (3)
Figure 3. (a) Energy difference between the indirect and direct gap, (EgL − EgΓ), as a
function of biaxial tensile strain and Sn atomic percentage; (b) Direct band gap EgΓ as
a function of biaxial tensile strain and Sn atomic concentration.
As shown in Figure 3a,b, a small tensile strain of 0.3% and 8.6 at % Sn alloying can transform
Ge into a direct gap semiconductor (EgL − EgΓ = 0) with a band gap of ~0.5 eV. This calculated result is
in very good agreement with our recent experimental data on a 0.3% tensile strained Ge0.913Sn0.087
sample [47]. Further considering the thermal excitation energy at 300 K, the material can already be
regarded direct gap when EgΓ − EgL ~kBT = 0.026 eV. For 0.3% tensile strain, this can be achieved at
7 at % Sn alloying. If the tensile strain can be moderately engineered to 0.6%, only 5 at % Sn alloying
is needed to achieve the indirect-to-direct gap transition effectively. Both can be readily implemented
using existing strain and Sn alloying technologies. Even for EgΓ − EgL = 2kBT = 0.052 eV, the n-type
doping required to compensate for the energy difference between the direct and indirect gap is
significantly reduced to ~2 × 1019 cm−3, compared to 7.6 × 1019 cm−3 that we originally proposed
in [19]. The lower n-type doping not only helps to reduce the detrimental Auger recombination, but
also simplifies the materials growth process. In addition, it has been shown that such a pseudo-direct
gap configuration can increase the emission efficiency at high temperature and high injection levels
due to enhanced L-to-Γ valley excitation or scattering [50–54]. Thus, one can also optimize the relative
positions of L vs. Γ valley for high power on-chip laser sources, where high injection current and high
thermal stability up to 80 °C are required in EPICs. All these results indicate that it is promising to
achieve high-quality direct gap Ge-based material on Si using a combination of tensile strain, Sn
alloying, and n-type doping engineering. The small band gap of tensile strained GeSn shown near the
top right corner of Figure 3b also indicates that it has significant potential applications in
Si-based mid infrared (MIR) integrated photonics.
2.5. Separate Confinement of Γ. vs. L Electrons for Direct-Gap Ge Quantum Well (QW) Structures
In addition to directly engineering the Ge material itself, we proposed that the band structure of
Ge could also be engineered towards direct gap semiconductor by carefully designing the QW device
structures [55]. QWs are commonly applied in LDs to reduce the threshold current density and
increases quantum efficiencies. Theoretical investigations on direct gap Si/Ge superlattices and QWs
have started since 1980s [56] and have continued through recent years [57,58]. However, there has
been no experimental demonstration of direct gap SiGe QW structures so far. A couple of challenges
have to be addressed. Firstly, in SiGe/Ge QWs the quantum confinement actually increases the energy
difference between the direct and indirect gaps due to the much smaller effective mass and much
stronger confinement of Γ electrons, as shown in Figure 4a. Secondly, the epitaxial growth of Ge on
SiGe typically results in compressive strain in Ge due to the lattice mismatch, which also increases the
energy difference between Γ and L valleys.
In order to improve the Ge QW structures, we proposed a “separate electron barrier” approach [55],
where the barrier material provides strong confinement in the L valleys, but weak confinement in the Γ
valley. This idea can be implemented by carefully choosing the barrier material that provides a large
band-offset in L valleys while small offset in Γ valley with Ge (Figure 4b). Such band alignment can
make Ge “direct band gap” by raising the L valleys above the Γ valley due to the stronger quantum
confinement of electrons in the L valleys. As an example, the band diagram of an AlGaAs/Ge/AlGaAs
QW structure is shown in Figure 4b, where the Ge QW is converted into a direct gap semiconductor
due to the stronger quantum confinement of electrons in the indirect L valleys provided by the AlGaAs
barriers. One could also tweak the design to engineer the L valley slightly below the Γ valley for high
lasing performance at elevated temperatures. There are several advantages of this approach: (1) The
AlGaAs barrier layer is not an active material but just a barrier layer electrode, so it can be deposited at
low temperature to avoid diffusion of III and V elements and achieve CMOS compatibility. (2) Off-cut
wafers may not be needed either, since antiphase boundaries will not significantly affect the performance
in this case. (3) Pseudo-direct gap Ge lasers offer much better thermal stability, and in this case one can
engineer the relative positions of L and Γ valleys by QW design for optimal performance. In III-V
lasers, quantum dot (QD) lasers also show good thermal stability, but the wavelength of these QDs
is far from 1550 nm and they have not been integrated on Si so far. Therefore, this new Ge QW design
could offer a simple solution to high-performance monolithic lasers on Si at 1500–1700 nm. In the
case of tensile strained Ge or GeSn active gain media, the energy gap reduction due to tensile strain
and/or Sn alloying can also be partially compensated by the quantum confinement effect. This
compensation helps to avoid potential issues related to the dark current in small band gap semiconductors.
Figure 4. (a) Ge0.85Si0.15/Ge/Ge0.85Si0.15 QW band structure assuming 0.25% tensile strain

in an 11.8 nm thick Ge QW. (b) Band structures of an AlGaAs/Ge/AlGaAs QW for
separate electron confinement in L and Γ valleys assuming no strain (due to nearly perfect
lattice matching). The thickness of Ge is 3.3 nm. For different valleys, dashed lines show
the band edges and solid lines show the first energy level in the wells. Barrier material is
chosen so that there is little quantum confinement in Γ valley but significant confinement
in L valley. © 2013 IEEE. Reprinted, with permission, from [55].
(a) (b)
In summary of this section, our analyses show that it is highly feasible to achieve direct gap or
optimized pseudo-direct gap Ge-based active photonic devices on Si by leveraging tensile strain,
Sn alloying, n-type doping, as wells as separate confinement of L vs. Γ electrons in QW structures.
3. Ge and GeSn Growth on Si
To implement the band engineering approaches mentioned in Section 2, high quality Ge and GeSn
growth on Si is the first critical step. The greatest challenge for high quality Ge epitaxy on Si is the
4.2% lattice mismatch between these two materials. In this section, we will discuss both epitaxial
growth and recent investigations towards low-temperature growth of high crystallinity Ge and GeSn on
amorphous dielectric layers for 3D photonic integration.
3.1. Epitaxy
Epitaxy is the most common approach to grow Ge on Si. To achieve device quality epitaxial Ge on
Si without using very thick SiGe buffer layers, a two-step direct Ge growth technique has been applied
to prevent Stranski–Krastanov growth [59,60] and subsequent annealing has been developed to
decrease the threading dislocation (TD) density below 107 cm−2 [61]. In the initial growth step, a thin
epitaxial Ge buffer layer of 30–60 nm is directly grown on Si at 320–360 °C. At such a low growth
temperature islanding of Ge associated with Stranski–Krastanov growth is kinetically suppressed due
to the low surface diffusivity of Ge. In the second growth step, the growth temperature is increased to
>600 °C for higher growth rates and better crystal quality. A post-growth annealing at >750 °C can
reduce the TD density from 109 to 107 cm−2. It has also been shown that the TD density decreases with
the increase of Ge film thickness [62], similar to the case of GaAs epitaxy on Si [63]. Luan demonstrated
that the TD density could be reduced to 7.9 × 106 cm−2 for a 3.9 μm-thick Ge layer [62]. This
phenomenon can be well explained by a dislocation coalescence/annihilation model during the epitaxial
growth or annealing [63,64]. More recently, a low threading dislocation density of 1.3 × 106 cm−2 has
been demonstrated in blanket Ge-on-Si films by growing a 4.5 μm-thick Ge layer and etching it back
to 1.8 μm [65]. In selectively grown Ge mesas with lateral dimensions ~10 μm, the average TD density
can also be reduced to low 106 cm−2 even for a relatively thin film of 1 μm [61,62] because these
dislocations can glide to the edge of the mesas and annihilate. This selective growth method is more
efficient than the growth of thick blanket Ge layers and more compatible with photonic device
integration. Details about the two-step growth and dislocation reduction methods can be found in
some recent reviews [66–68]. In particular, in-situ n-type doping methods have been reviewed
in [66,67]. In CVD process, a phosphorous doping level up to 5 × 1019 cm−3 has been achieved by
depositing a stack of delta-doped Ge layers at low temperatures on top of regularly in-situ doped Ge
(with n = 1–2 × 1019 cm−3) followed by drive-in annealing [69,70]. These materials enabled the first
demonstration of band-engineered Ge-on-Si LDs. To accommodate the thermal budget for epitaxial
growth in device integration, the Ge epitaxy is commonly performed after the gate formation but
before the metallization process (see the processing temperature color bar in Figure 1).
Epitaxy of GeSn has been implemented by both MBE and CVD. The development of SnD4
precursor was the key factor that enabled Sn alloying in CVD growth [71]. The deposition temperature
is typically between 200 and 350 °C, and higher Sn content typically requires lower growth temperature
to prevent Sn segregation. Compared to Ge, GeSn tends to relax more easily on Si substrate due to
an even larger lattice mismatch. Similar to epitaxial Ge, annealing has been shown to improve the
optoelectronic properties of GeSn [45,72] by reducing the defect density. Rapid thermal annealing (RTA)
is preferred in this case to remove the defects without causing Sn segregation. Since room-temperature
light emission has been demonstrated from both MBE and CVD growth samples, there does not seem
to be significant difference in material quality between these two epitaxial growth methods.
In recent years, the rapid melting growth (RMG) or liquid phase epitaxy (LPE) of Ge has been
developed as an alternative to MBE and CVD epitaxy [73–75]. A typical Ge RMG/LPE structure is
schematically shown in Figure 5. An amorphous Ge (a-Ge) thin film is deposited on a dielectric layer,
with a small window in contact directly with the single crystalline Si substrate or a patterned Si grain
to serve as the seedling region for the single crystal Ge growth upon melting and recrystallization.
The a-Ge deposition does not require MBE or CVD epitaxy tools, so it can be implemented by less
costly, higher throughput methods such as plasma enhanced CVD (PECVD). In some cases, a very
thin SiGe buffer layer is deposited first to reduce the columnar growth and the surface roughness [75].
A dielectric capping layer such as SiO2 is often deposited on top of the a-Ge before the RMG process.
Since a high temperature >937 °C is required to melt Ge, RMG has to be incorporated into the
front-end-of-line (FEOL) processing during device integration, along with the RTA for activating
implanted dopants in the source/drain regions [75]. Upon melting at ~1000 °C, the Ge crystal grows
laterally, seeded by the region in contact with single crystal Si. The key to this approach is to achieve
a fast lateral growth rate to prevent random nucleation of Ge grains on the dielectric layers. Experimental
results have consistently shown that the single crystal regime can sustain for tens to hundreds of
microns, large enough for integrated Ge photonic and electronic device fabrication. Since the defects
due to Si/Ge lattice mismatch is confined to the seed region, nearly defect-free single crystalline Ge
strips can be obtained away from this seeding region. Indeed, a high hole mobility of >1000 cm2 V−1 s−1
has been reported [74], indicating a high material quality. Recently, a thermally induced tensile
strain of 0.4% has been reported from Ge-on-insulator grown by RMG [76] due to the high growth
temperature and reduced stress relaxation upon rapid cooling after RTA. One of the challenges for
photonic applications, though, is that the Ge region almost inevitably incorporates a small amount of Si
due to Ge-Si inter-diffusion upon melting. Incorporation of Si tends to blue-shift the direct band gap
and increases the difference between the direct and indirect gaps. For PDs, the blue-shift will affect the
optical absorption at longer wavelengths (λ > 1550 nm). For EAMs and LDs, the performance tends to
deteriorate due to the increase in the energy difference between the direct and indirect band gaps.
Figure 5. Schematics of a typical structure for Ge RMG/LPE growth. A cross-sectional

view is shown here.
3.2. Growth Methods towards 3D Photonic Integration
So far, most of the Ge photonic devices are grown on single-crystal Si (either SOI or Si substrate)
via heteroepitaxy, constituting a planar integration scheme (Figure 1c). However, heteroepitaxy of Ge
consumes device area on single crystal Si, thereby significantly sacrificing CMOS area for photonic
devices. The size of the photonic devices is limited by the diffraction as well as the optical path length
necessary for light-matter interaction. For example, a waveguide-integrated Ge PD is typically >0.1 μm2
in cross-section and from several microns to tens of microns in length. This footprint is more than
100× larger than CMOS transistors in 22 nm technology node. Furthermore, this planar integration
approach has to introduce some modifications to existing CMOS process flow in order to accommodate
the epitaxial growth as well as the vertical dimension of photonic devices (typically several hundred nm
thick). These modifications are hard to implement in <45 nm CMOS technology nodes. Therefore, for
large-scale EPICs it would be ideal to adopt a 3D configuration, moving the photonic circuits to the
metal/dielectric interconnect level above the CMOS layer using BEOL processing. For monolithic 3D
photonic integration, the major challenges for materials growth are: (1) the lack of single crystal
template for epitaxy makes it difficult to achieve high crystallinity; and (2) the processing temperature
has to be kept below ~500 °C to be compatible with the metal contacts. Although metal induced
crystallization (MIC) has been investigated to achieve low crystallization temperature [77–79], the
metals commonly used in MIC, including Ni, Au, Co, Ag, Pd and Al, either introduce deep-level
defect centers or exceedingly high doping concentrations that are detrimental to semiconductor device
performance [80]. On the other hand, direct growth of Ge or GeSn on amorphous dielectric layers also
offers a new opportunity in that the constraint of lattice mismatch in epitaxy is naturally removed.
Once implemented, it offers unprecedented flexibility in fabrication of 3D circuit structures.
To grow single crystalline Ge on amorphous dielectric layers, we have proposed and demonstrated
the concept of geometrically-confined growth of single crystalline Ge on amorphous Si at a low
growth temperature of 450 °C, as shown in Figure 6 [81]. The facets of the Ge material in Figure 6a
indicate the single-crystal nature of the grain. The observed boundary (also confirmed by electron
backscatter diffraction, EBSD) is a coherent interface with no dangling bonds to cause mid-gap states,
so it will not adversely affect the optoelectronic properties. To achieve geometrical confinement for Ge
growth, we first fabricate ~200 nm-wide amorphous Si (a-Si) wires on a SiO2 layer, then cover them
with another SiO2 layer and etch from the top to expose one end of the Si wires (Figure 6b). An undercut
etching of the a-Si wire using TMAH forms a hollow channel for geometrical confinement (Figure 6c).
During the ultrahigh vacuum CVD (UHVCVD) process, Ge only selectively deposits on Si but not on
SiO2. Therefore, in this geometric configuration Ge growth has to start from the a-Si seed with a
nanoscale cross-section of ~200 × 200 nm2 and propagate laterally along the channel. Multiple Ge
nuclei may start to form on the a-Si seed initially, leading to a growth competition of Ge grains inside
the channel. During this competition, the nucleus with the fastest lateral growth rate along the channel
will eventually dominate and block the growth of other grains if the aspect ratio of the channel, d/h, is
large enough. This nucleus typically has its <110> orientation aligned with the channel, the fastest
growth direction in Ge considering the effect of twinning. Therefore, the fastest-growing grain is
selected by the geometric confinement and leads to single crystal Ge growth outside the channel
(Figure 6d). Using this geometrically confined growth method, in principle the area of the single crystal
Ge region can be enlarged to any desirable size simply by increasing the growth time. However,
in reality the lateral growth rate is limited to a few μm/h in order to prevent random Ge nucleation on the
SiO2 mask layer. In addition, the shape of the Ge single crystal is hard to control exactly.
Considering that Sn and Ge form a eutectic system with a low eutectic temperature of 231 °C,
we expected that the GeSn alloy may crystallize at a lower temperature with a higher lateral growth
rate than Ge. As mentioned earlier, Sn alloying also enhances the optoelectronic properties by
transforming Ge towards a direct gap semiconductor. Therefore, recently we studied the crystallization
of GeSn on amorphous SiO2 layers towards monolithic 3D photonic integration [47]. Amorphous
GeSn (a-GeSn) films were evaporated on a layer of SiO2 followed by crystallization annealing.
The eutectic phase transition in Ge-Sn system greatly enhances the crystallization at <500 °C.
Pseudo-single-crystalline GeSn thin films with up to 8.7 at % Sn has been directly grown on a SiO2
layer at ~460 °C. As shown in Figure 7a,b, we clearly identified nucleation centers that are as far as
~0.1–1 mm away from one another, from which the grains grow radially. This striking observation is
highly unusual in the crystallization of amorphous semiconductor thin films, and it suggests an
extraordinarily fast lateral grain growth rate vs. a slow nucleation rate during the eutectically enhanced
crystallization process of GeSn. EBSD analysis shows that the average grain size is as large as
~10 μm, and the majority of grain boundaries are relatively benign twin boundaries (Figure 7c,d).
For an initial Sn composition of 9–13 at % in the as-deposited films, the crystallized GeSn film
exhibits a strong (111) texture, as shown by the inverse pole figure in the inset of Figure 7d and the
X-ray diffraction (XRD) data in Figure 7e. We also found that the excess Sn segregates on the surface
of the GeSn thin films along the small angle grain boundaries, and it can be selectively removed by a
37% HCl:H2O solution with a high Sn-to-GeSn (Ge-rich) etching selectivity >4800:1 [47]. After the
selective etching, β-Sn signals no longer exist in the XRD data. From XRD, micro-Raman, energy
dispersive X-ray spectroscopy (EDS) and EBSD analyses, it was found that the films exhibit ~0.3%
tensile strain and 8.7 at % substitutional Sn composition. Similar to the cases of MBE and CVD, the
substitutional Sn composition far exceeds the equilibrium solubility of Sn in Ge (~1 at %). Moreover,
the substitutional incorporation of 8.7 at % Sn into Ge and the 0.3% tensile strain transform the
material into a direct gap semiconductor. As mentioned in Section 2, the measured band gap of ~0.5
eV from optical transmittance spectroscopy (Figure 7f) is in good agreement with the theoretical
calculation in Figure 3. Further employing geometrically confined growth, it is promising to achieve
single crystalline GeSn on dielectric layers for high-performance active devices in 3D photonic
integrated circuits.
Figure 6. (a) Top view scanning electron microscopy (SEM) photo of a single crystal Ge
grain grown on a SiO2 layer using geometric confinement method. The fabrication steps for
the geometrically confined Ge growth are shown in (b–d). Reproduced from [81] with
permission from McComber, K.; Duan, X.; Liu, J.F.; Michel, J.; Kimerling, L.C.;
Advanced Functional Materials; published by John Wiley & Sons, 2012.
Figure 7. (a) Large-area SEM image of a Ge0.913Sn0.087 thin film crystallized from a-GeSn
with 9.5 at % Sn. The arrows point to six nucleation centers in a total area of ~1.6 mm2,
from which GeSn grains grow radially. The scale bar is 0.5 mm. (b) A zoomed-in image
of domain 2 in (a). The scale bar is 100 μm. (c) EBSD mapping of the crystallized GeSn
thin film, where different colors indicate different in-plane crystallographic orientations.
The scale bar is 5 μm. (d) Distribution of misorientation angles between neighboring
grains derived from the EBSD data. The inset shows the inverse pole figure obtained over a
0.2 × 0.2 mm2 area, indicating a strong (111) texture. (e) θ-2θ XRD data of the GeSn
sample vs. a pure Ge sample. The GeSn sample shows a strong (111) texture, confirming
the EBSD result. Note that the vertical axis is in log scale. (f) Transmittance spectrum of
the high crystallinity Ge0.913Sn0.087 thin film sample compared to a reference sample
without the GeSn thin film. The onset of strong absorption at ~λ = 2500 nm is consistent
with the calculated direct band gap of ~0.5 eV. Reproduced from [47] with permission
from the Optical Society of America under the Author’s Copyright Transfer Agreement.
4. Ge-on-Si Active Photonic Devices
Armed with high quality Ge materials on Si, high performance Ge-on-Si PDs and EAMs have been
demonstrated in recent years. Prototype Ge-on-Si LDs have also been achieved. In this section, we will
first discuss waveguide-coupling schemes for integrated Ge photonic devices on Si. We then discuss
each Ge-on-Si active photonic device individually.
4.1. Waveguide Integration Schemes
Waveguide-coupling with Ge active photonic devices is indispensable to construct integrated photonic

circuits. Butt-coupling and evanescent coupling are the most commonly used waveguide-coupling
schemes, as schematically shown in Figure 8.
Figure 8. Schematic structures of typical waveguide-integrated Ge active photonic devices.

(a) butt-coupled; (b) Evanescent coupling with waveguide at the bottom of the Ge photonic
device; (c) Evanescent coupling with waveguide on top of the Ge photonic device [18].
Butt-coupling launches optical power directly into the Ge active device region, thereby offering a
higher coupling efficiency compared to evanescent coupling in most cases. It is also straightforward to
couple the light into and out of the Ge or GeSi region for EAMs. Therefore, the same butt coupling
structure can be applied to all Ge active photonic devices for monolithic photonic data links [82,83].
The fabrication process of butt-coupled devices is typically more elaborate than evanescently coupled
ones, though, since Ge has to be selectively grown to fill pre-defined trenches in order to implement
butt-coupling structures [83–87]. Chemical mechanical polishing (CMP) is often required to flatten the
top of the Ge mesas since single crystalline selective growth usually results in faceting and
a non-planar surface. An example of such a fabrication process is schematically shown in Figure 9,
where cross-sectional SEM images of Ge overgrown out of the trench and after planarization are
also presented.
Figure 9. Schematic fabrication process of butt-coupled Ge PDs and EAMs on SOI platform.
“a-Si” and “c-Si” refer to amorphous Si and crystalline Si, respectively. Trench-filling selective
growth is applied to achieve these structures. Cross-sectional SEM pictures of trench-filled
Ge selective growth and the Ge region after CMP are also shown. © 2006 IEEE. Reprinted,
with permission, from [88].
Several types of butt-coupling structures have demonstrated high coupling efficiencies. For example,
the Ge PD/EAM has been designed and fabricated as a segment of the waveguide, with its dimensions
optimized for maximum modal overlap with that of the input and output Si waveguides [83,84].
Tapered mode converters have been applied to enhance butt-coupling efficiency [86]. To enhance the
tolerance to fabrication errors, especially lateral misalignment, in some cases a wide patch of Ge is
used to capture the optical power coupled and scattered into the Ge region [85]. This kind of optical
design is somewhat similar to multi-mode interference (MMI) couplers. It could be extended to EAMs
if output coupling is also designed carefully to distribute the modulated light into different channels.
The trade-off is that the Ge device area tends to be large.
Evanescent coupling uses the fact that light can be easily coupled evanescently from a lower index
material (i.e., waveguide) to a higher index material (e.g., Ge PDs) as long as the index difference is
small enough. The electromagnetic mode and hence the optical power is typically transferred gradually
to the high index material. The waveguide is usually placed either on bottom or on top of the Ge region
to simplify fabrication processes. In bottom-waveguide-coupled structures, Ge devices are typically
grown epitaxially on crystalline Si waveguides fabricated on SOI substrates. By comparison,
top-waveguide-coupled structures allow greater flexibility in choosing waveguide material since they
do not have to form a template for Ge epitaxy. The waveguide material can be amorphous, such as
amorphous Si (a-Si), SiON or SiN. These amorphous waveguides are applicable to photonic circuits at
the upper interconnection level for 3D photonic integration.
Ahn et al. [89,90] systematically studied evanescent coupling processes between the waveguide and
the Ge active region in the case of Ge PDs, which can be readily extended to other waveguide-coupled
Ge devices since the coupling only depends on the refractive index profile and the geometry. The
analysis identified three mechanisms: (i) Direct coupling and scattering into the Ge region due to mode
matching and abrupt effective index change (Figure 10a); (ii) Evanescent coupling at a constant rate
(Figure 10b); and (iii) Backscattering of residual optical power in the waveguide into Ge at the end
of waveguide/Ge coupling region due to the abrupt change in effective index and geometry. Most
waveguide-coupled Ge PDs are designed to absorb all input photons, so Mechanism (iii) is negligible
in most cases since there is hardly any residual optical power in the waveguide at the end of the
coupling region. However, this factor needs to be taken into account for EAMs since optical output
from the Ge active region is also important. Mechanism (i) is more efficient than Mechanism (ii) in
coupling optical power into the Ge region, so it is the preferred mechanism for waveguide-coupled Ge
PDs. On the other hand, Mechanism (ii) is similar to directional couplers and can be adapted to
waveguide-coupled GeSi EAMs. In this case, the light oscillates between the input waveguide and the
EAM active region for optical modulation before it exits the output port of the waveguide [91]. Ahn et al.
also investigated two major factors that affect coupling efficiency for vertical evanescent coupling
in Ge PDs: (1) The refractive index difference between Ge and the core of the waveguide materials,
nGe-ncore; and (2) The geometrical factor of the waveguide. Generally speaking, low index difference
between the waveguide and the Ge region, small cross-sectional dimensions of the waveguides, and
low height-to-width aspect ratio of waveguides lead to a Mechanism (i) dominated regime and greatly
enhanced coupling efficiency into Ge. Lower index differences reduce the impedance mismatch between
the waveguide and the Ge region, thereby facilitating coupling. Smaller waveguide cross-sectional
dimensions or aspect ratio help to increase optical confinement in the Ge region for more efficient optical
absorption. In some device structures, thick Si waveguides is preferred for optimal performance, such as
waveguide-coupled avalanche photodiodes (APDs). To enhance evanescent coupling efficiency between
thick Si waveguides and Ge APDs, we have designed a step coupling structure to create mirror imaging
modes at the Ge/Si interface using a vertical MMI mechanism [92]. The coupling length has been
effectively reduced by 5× for 70% absorption, and 3× for 90% absorption.
Figure 10. 3D finite difference time domain (FDTD) simulation of photon propagation in
bottom-waveguide-coupling structures (a) 3.0 μm vs. (b) 0.4 μm thick Ge active region is
placed on 0.2 μm thick Si channel waveguide. Ge absorption coefficient of αGe = 4000 cm−1
at λ = 1550 nm was assumed for simulation, i.e., considering 0.22% tensile strain in Ge.
As shown in the figures, when the thickness of the Ge region decreases relative to the Si
waveguide, the coupling efficiency deceases due to a shift in coupling mechanism from
direct coupling and scattering dominated regime (Mechanism i) to evanescent coupling at a
constant rate (Mechanism ii). The unit of the axes is micron. © 2010 IEEE. Reprinted, with
permission, from [89].
4.2. Ge-on-Si PDs
Ge-on-Si PDs are the most mature integrated Ge photonic devices. Most of them adopt a p-i-n diode
structure, with the intrinsic layer being Ge and at least one of the p/n electrodes being heavily doped Si
substrate [93–95]. The key to high performance devices is to optimize the built-in field in the intrinsic
Ge region so that carriers can be collected fast enough before recombining at any defect centers. For
high quality intrinsic epitaxial Ge layers, full responsivity can be achieved even at 0 bias (photovoltaic
mode) due to this large built-in electric field [93]. Free-space coupled Ge-on-Si p-i-n photodiodes have
reached a bandwidth of 49 GHz at −2 V reverse bias [96]. The dark current density has been reduced
from ~10 mA/cm2 to ~1 mA/cm2 at −1 V over the past decade mainly by improving the surface
passivation to decrease peripheral leakage for devices with small dimensions ≤100 μm2 [97]. Note that
reducing the device area also decreases RC delay and increases the bandwidth of the device. As
mentioned in Section 3.1, in these small area devices, the TD density is much lower than in blanket
films, so the dark current is dominated by generation-recombination at the Ge/passivation interface.
When the passivation interface is properly doped or treated, the depletion region is pushed away from
the passivation interface, thereby reducing the generation/recombination current. Based on the same
consideration, the low-quality Ge buffer layer mentioned in Section 3.1 is heavily doped either by
drive-in diffusion from the Si substrate during the post-growth annealing [62] or via in-situ doping [98].
The Ge buffer layer now becomes part of the electrode so that the incorporated misfit dislocation
network does not contribute significantly to generation/recombination. By now, the performance of
free-space Ge-on-Si p-i-n diodes is comparable to their III-V counterparts at 850 and 1310 nm. The
responsivity at 1550 nm is usually a bit lower than InGaAs detectors due to a lower absorption
coefficient. This is mainly because the direct band gap of 0.2% tensile strained Ge (0.77 eV) is larger
than InGaAs (≤0.75 eV). With the recent progress in band engineering methods described in Section 2,
it is possible to further improve the responsivity at 1550 nm for free-space Ge-on-Si PDs, e.g., by
increasing tensile strain or incorporating a small amount of Sn. Indeed, recently a 40 GHz Ge0.98Sn0.02
free-space p-i-n photodiode has been demonstrated [99] with the responsivity at 1550 nm 2× higher
than a pure Ge photodiode of the same absorber layer thickness. In 2011, commercial optical receivers
based on free-space Ge-on-Si PDs monolithically integrated with Si trans-impedance amplifier (TIA)
circuits were reported by Analog Devices, Inc. [100]. Data rate up to 43 Gb/s was demonstrated from
vertically illuminated Ge-on-Si receivers recently [101]. Free-space Ge-on-Si APDs, which
synergistically integrate the high absorption of Ge layers at telecommunication wavelengths with the
outstanding low-noise avalanche gain in Si, have demonstrated performance superior to their III-V
counterparts [102] and are being commercialized for 10G-PON applications [103].
To achieve photonic circuits on Si, waveguide-integrated Ge-on-Si PDs have been developing
rapidly in recent years. An additional benefit for device performance is that they have no trade-off
between bandwidth and quantum efficiency as do free-space PDs [88,104,105]. This is because in
waveguide-integrated PDs, the optical absorption path is in the longitudinal direction along the
waveguide, which can be as long as needed to fully absorb the input optical power. On the other hand,
carrier collection can be implemented in a transverse direction perpendicular to the light propagation
direction, with a dimension in the order of several hundred nm to minimize carrier transit time and
increase bandwidth. One can optimize the PD design for maximum bandwidth while maintaining
a high quantum efficiency >90%. Since the first reports on waveguide-integrated Ge-on-Si PDs, the
performance of these devices has been rapidly improving. Table 1 summarizes the reported performance
of waveguide-coupled Ge-on-Si PDs since 2009 (the data between 2006 and 2009 have been summarized
in [68]). The responsivity is typically ~1 A/W at λ = 1550 nm, or >80% quantum efficiency. The
bandwidth has generally exceeded 40 GHz since 2009 [75,85,106]. A 60 GHz evanescently coupled
Ge p-i-n diode has been demonstrated very recently by coupling the detector with a carefully designed
inductor-capacitor circuit utilizing inductive gain peaking [107]. An estimated bandwidth of 120 GHz
has been reported in [108] with the directly measured result exceeding 67 GHz. The maximum
bandwidth-efficiency product has exceeded 30 GHz [86,108]. In selectively grown butt-coupled Ge
photodetectors with sidewall passivation by PECVD SiO2, a low dark current density of 0.7 mA/cm2 at
−1 V has been achieved with a very small absolute dark current of 0.2 nA [84]. This absolute dark
current is orders of magnitude lower than typical free-space diodes. However, in some literature a large
dark current density was reported [86,108]. Considering the high responsivity and bandwidth of these
devices, it is likely that Ge material itself is still of high quality, and the leakage is mainly due to
peripheral passivation issues. In terms of system integration, Kotura, Inc. recently reported terabits/s
wavelength division multiplexing (WDM) photonic receiver chips based on 40 waveguide-coupled
Ge-on-Si PDs monolithically integrated with a 40-channel WDM echelle grating [109]. For Ge-on-Si
APDs utilizing avalanche gain in the Si layer, waveguide integration can at least double the primary
responsivity compared to free-spaced ones, which further improves their sensitivity [92,110]. Photocurrent
gain from the Ge region itself has also been utilized to enhance the sensitivity of waveguide-coupled
PDs [111].
Table 1. Summary of waveguide-coupled Ge-on-Si PD performance. All devices are

fabricated by epitaxial growth on Si and performance data are reported for −1 V reverse
bias, unless otherwise stated.
Responsivity Dark Current
3dB Bandwidth (GHz) Absolute Dark Device
(A/W) @ 1550 nm Density Reference
2
Current (μA)@ −1 V Design
Max 0 bias Max 0 bias (mA/cm )
butt
1.1 - 37@−3 V 17.5 1.6 × 104 1.3 [86]
p-i-n
butt
1@-4 V 0.2 42@−4 V 12 60 0.018 [85]
p-i-n
Bottom
>1.1 @ <1540 nm - 50@−5 V - 8900@ −5 V a 4@−5 V [106] b
msm
0.42 0 40@−2.5 V - - 90 bottom msm [75] c
bottom
0.95 - 36 - 29 0.0046 [112]
p-i-n
>67;
0.8 0.78 40 Gb/s 8 × 104 4 butt [108]
estimated 120
bottom
0.75@−2 V - 60@−2 V - 3750 3@−2 V [107]
p-i-n
a
Calculated and/or read from corresponding figures of the reference. b Ge layer fabricated by wafer bonding.
c
Ge layer fabricated by rapid melt growth (RMG).
In recent years, bandwidths at 0 bias (photovoltaic mode) have also been improved significantly
especially for waveguide-integrated p-i-n PDs. This is because the thickness or width of the intrinsic
Ge region can be reduced to obtain a stronger built-in electric field at 0 bias without affecting the
optical absorption path length in the longitudinal direction. Full responsivity and up to 40 Gb/s data
rate have been demonstrated at 0 bias [108]. Operations under photovoltaic mode at 0 bias are
especially beneficial to achieve high energy-efficiency in large-scale EPICs since the PDs practically
consume no power at all. One could even consider harvesting electrical power directly from the
incident laser by operating the Ge PDs under photovoltaic mode. For example, the photovoltage could
be used to drive electronic circuits without using power consuming TIA circuits. The energy
conversion efficiency can also be much higher than solar photovoltaic cells, since the optical power
from the laser is at a single wavelength. Using the demonstrated dark current density of 1 mA/cm2 and
responsivity of 1.1 A/W at 1550 nm, under 100 nW optical input an 0.5 × 10 μm2 waveguide-coupled
Ge PD would offer an open circuit voltage of ~0.2 V. Considering that the driving voltage of CMOS in
22 nm technology node has been decreased to ~0.8 V, 4 Ge PDs in series are enough to provide the
driving voltage for CMOS circuitry, potentially eliminating the need for TIAs and minimizing the
energy consumption in EPICs.
4.3. Ge and GeSi EAMs
Similar to III-V semiconductors, the direct gap absorption in Ge can be effectively modified
by an applied electric field, known as electroabsorption (EA) effect. EA effect in semiconductors
include Franz-Keldysh effect (FKE) in bulk materials and quantum-confined Stark effect (QCSE) in
QW structures.
Figure 11 schematically shows the optical absorption process and absorption spectra in a semiconductor
material with and without an electric field. Without the field, photons with energy less than the band
gap are not absorbed, while those with energy above the band gap are (Figure 11a). The absorption
coefficient spectrum follows the well-known square-root relation for direct gap transitions (Figure 11b,
dashed line). With an applied field, the absorption process can be considered as band-to-band
tunneling under photon excitation. In this case, even a photon with hν < EgΓ can excite an electron
from the valence band to the conduction band through a cooperative tunneling process across a
triangular energy barrier (Figure 11c). Therefore, the absorption coefficient at hv < EgΓ increases
compared to the case without the electric field. On the other hand, for hv slightly greater than EgΓ the
absorption curve shows oscillating behavior near the band edge, known as Franz-Keldysh oscillations
(Figure 11b, solid line). This is due to the fact that, for excited electrons with energy slightly greater
than the triangular barrier, the probability to cross the barrier and reach its final state on the tilted
conduction band edge is determined by the interference between the electron wavefunction and the
barrier. This phenomenon is well known in quantum tunneling and it is similar to the Fabry-Perot
oscillations in optics, both resulting from the nature of waves. While the picture presented above is
simplified, it does capture most of the physics in FKE [113].
Figure 11. (a) Schematics of band-to-band optical absorption in a bulk semiconductor

material without an applied electric field. (b) Absorption spectra of the material with
and without an electric field. (c) Schematics of absorption process in a bulk semiconductor
material under an applied electric field (Franz-Keldysh effect).
In QWs, the binding energy of excitons is notably larger than in bulk materials due to the quantum
confinement, so exciton absorption peaks are clearly visible even at room temperature. Therefore,
in addition to band-to-band transitions, the significant exciton absorption peak will notably enhance
the EA effect compared to bulk materials. A rigorous, quantitative analysis of FKE and QCSE can be
found in [114]. It is noteworthy that the stronger the optical absorption (without electric field), the
stronger the EA effect since both are proportional to the transition matrix element between the valence
band and the conduction band. As a result, only direct gap transitions show significant EA effect.
The figure-of-merit (FOM) for EAM materials and devices is the absorption contrast, Δα/α, i.e.,
relative change in absorption coefficient when an electric field is applied. As shown in Figure 11b, this
is maximized at photon energies slightly smaller than the direct band gap. While FKE in bulk Ge only
shows a Δα/α ~ 1 mainly limited by background absorption associated with indirect gap transition,
0.2% tensile-strained Ge demonstrates a significantly better absorption contrast of Δα/α ~ 4 at λ ~ 1630 nm
since the energy difference between direct and indirect gaps is decreased [115]. On the other
hand, strong QCSE from the direct gap transition of Ge has been demonstrated in strain-balanced
Ge/Si0.15Ge0.85 multiple QWs grown on a relaxed Si0.1Ge0.9 buffer layer on Si [116]. The composition
of Si0.15Ge0.85 was specifically chosen to achieve Type I band alignment for the L-valleys, although
it is not completely clear whether this is absolutely necessary to achieve significant QCSE from
the direct gap transition. In III-V materials, typically QCSE is much stronger than FKE due to the
additional contribution from excitons induced by quantum confinement. In the case of Ge, QCSE
indeed shows a much larger Δα compared to FKE. However, the compressive strain in SiGe/Ge QWs
as well as the quantum confinement effect itself increases the energy difference between the direct and
indirect gaps of Ge QWs, as discussed in Section 2.5. Consequently, the background absorption
from the indirect gap transition of Ge QWs also increases significantly. Overall, QCSE also provides a
Δα/α ~ 4 at an optimal wavelength of λ ~ 1460 nm [116], similar to tensile-strain enhanced FKE in Ge
at 1630 nm. Considering that the absorption contrasts are comparable while a single tensile-strained Ge
epitaxial layer is much easier to grow than multiple Ge QWs, for integration with CMOS process it is
more convenient to use tensile-strained enhanced FKE for waveguide-coupled Ge EAMs. Moreover,
the optical confinement in the GeSi FKE modulators is typically larger than that in the Ge QWs.
In order to shift the operation wavelength to λ ~ 1550 nm, a small amount of Si can be added to
tensile strained Ge to increase the direct band gap of FKE modulators (Figure 12). One should note
that the addition of Si tends to decrease Δα/α since it increases the difference between the direct and
indirect gaps. Nevertheless, a Δα/α of ~3 can still be achieve at λ ~ 1550 nm [82]. For SiGe/Ge QCSE
modulators with a direct band gap corresponding to λ ~ 1430 nm, one could heat up the device to
reduce the band gap, apply a larger electric field to extend the EA spectral regime, or utilize some
strain engineering approaches (see Section 2.1) to convert the compressive strain into tensile strain.
Strain engineering is the most preferred method since it avoids additional power consumption and
helps to reduce the energy difference between the direct and indirect gaps in Ge QWs, thereby
increasing Δα/α. On the other hand, broad-band QCSE effect has been demonstrated recently by
improving the QW design and controlling the DC reverse bias during operation [117]. Different from
pervious QW structures, the 0.18% thermally induced tensile strain in the SiGe buffer layer is taken into
account to design the QW stack. The Si composition in the buffer layer is increased from 10%–12% to
further reduce the lattice mismatch with the Si substrate, and the QW stack is designed to have an
average lattice constant identical to the tensile strained Si0.12Ge0.88 buffer layer. The Si composition in
the barrier layers is increased from 15%–19% correspondingly. This design enabled thin buffer layers
and Ge QW stacks such that the same electric field can be achieved at a lower reverse bias. By
measuring the responsivity spectra vs. reverse bias of a vertically incident Ge QW p-i-n photodiode,
it was found that the absorption edge can be shifted beyond 1550 nm at −3.3 V reverse bias. A Δα/α of
~3.5 can be achieved at a voltage swing of Vpp = 1 V in S band and C band by increasing the reverse
DC offset from −0.4–−3 V. Based on the materials performance, it was predicted that a 200 μm-long
waveguide-coupled Ge QW device can achieve 6.5 dB extinction ratio/4 dB insertion loss and 15 dB
extinction ratio/7 dB insertion loss at 1550 and 1490 nm, respectively.
Figure 12. (a) Absorption contrast, Δα/α, as a function of wavelength for GeSi alloys with
different Si compositions (b) Δα/α at λ = 1550 nm as a function of Si composition. We
assumed an applied electric field of 100 kV/cm and a built-in field of 10 kV/cm at 0 V in
the intrinsic GeSi region. Panel (b) is reproduced from [82] with permission from the
Optical Society of America under the Author’s Copyright Transfer Agreement.
Table 2 summarizes the performance of Ge EAMs reported in recent years. Compared to Si MZI
modulators based on free-carrier induced refractive index changes, GeSi EAMs are based on an
ultrafast (<1 ps) and highly efficient field-induced change in absorption near the direct band edge,
which enables a very compact device size and an ultralow capacitance. These features are especially
advantageous for short-range applications such as photonic links in data centers, while phase-based
modulators such as Si MZIs are more favored for coherent transceivers in metro and long-haul
applications [118]. Compared to Si microring modulators, GeSi EAMs have a broader range of operation
wavelengths and their performance is much less susceptible to temperature variations on chip.
The first waveguide-integrated GeSi EAM was demonstrated using tensile-strain enhanced FKE and a
butt-coupling scheme [83]. The material was alloyed with 0.75% Si for operation at 1550 nm wavelength,
according to the design in [82]. An extinction ratio of 8 dB at λ = 1550 nm was achieved with an
insertion loss of 3.7 dB and a peak-to-peak voltage swing of Vpp = 3 V (from −4 to –−7 V). A 30 GHz
waveguide-integrated Ge FKE modulator optimized for L-band operation (1610–1640 nm) [119] and a
40 GHz GeSi (0.7% Si) FKE modulator [120] optimized for 1550 nm operation have been
demonstrated by Kotura, Inc. Very recently, a nine-channel GeSi EAM array integrated with 40 nm
technology node digital CMOS driver and wavelength multiplexer was reported, marking the entry of
waveguide-coupled GeSi EAMs into system level integration. [121] Single channel operation up to
25 Gb/s and nine-channel operation at 10.25 Gb/s has been achieved at a Vpp of 2 V. Evanescently
coupled Ge FKE modulators have also been reported [91].
Table 2. Summary of Ge and GeSi EAM performance. The data are reported for 1550 nm operation, and the voltage swing is from 0 to −Vpp
(reverse bias) unless otherwise stated.
Average Dynamic
Optimal Wavelength Active Device
Reference Coupling Method Extinction Ratio (dB) Insertion Loss (dB) Bandwidth (GHz) Vpp (V) Energy/bit
Range (nm) Area (μm2)
1/4 CVpp2 (fJ/bit)
Ge and GeSi FKE Modulators:
[83] 1539–1553 butt 8 3.7 1.2 3 (−4 to −7 V) 0.6 × 50 25 a
[91] 1580–1610 evanescent 10@ 1600 nm 9.6 1.25 Gb/s 5 0.8 × 20 102–103
[119] 1610–1640 butt 6.3@ 1620 nm 3.6@ 1620 nm 30 4 1 × 45 ~70
[120] 1545–1581 butt 6 5 40.7 2.8 1 × 55 60
>40; >10 Gb/s for 50 for EAM;
[121] b 1525–1555 butt 3 3 2 -
9 channels 570 including driver
Ge QCSE Modulators:
[122] 1441–1461 Side-entry 7.3@ 1457 nm ~9@ 1457 nm - 10 450 × 450 -
[123] 1415–440 fiber coupled 7/10@ 1420 nm 3/7 dB@ 1420 nm - 6 L = 34/64 μm -
[124] 1425–1446 fiber coupled 9@ 1435 nm Vpp = 1 V 15@1435 nm, (−3 V) 23 1 (−3 to −4 V) 3 × 90 108
1450–1470@
[125] butt >3.2@ 1460 ~15 3.5 1 (−3 to −5 V) 0.8 × 10 0.75
−4 V DC bias c
[126] 1290–1310 fiber coupled 6@ 1293 nm 2.5 - 7 3 × 150 -
a 2
In [83] a conservative estimate of 50 fJ/bit was reported based on 1/2 CVpp , considering the most power-consuming scenario of flipping between “on” and “off” states in
every operation. b Integrated with 40 nm technology node CMOS driver and wavelength multiplexer. c The operation wavelength range could be redshifted up to 1550 nm
at a DC bias of –7.5 V and 1 V voltage swing judging from the photocurrent measurement. The direct modulation data is only available at 1460 nm with −4 V DC
bias, though.
The first Ge QCSE modulator was demonstrated using a side-entry approach [122]. The light was
coupled through the edge of a lightly doped Si wafer, then incident on a Ge QW mesa grown on top of
it. A 7.3 dB extinction ratio at 1457 nm was demonstrated with a Vpp of 10 V. Ge QCSE modulators
directly coupled to taper-lensed fibers were investigated in [123]. An extinction ratio greater than 7 dB
has been demonstrated at ~1420 nm with ~3 dB insertion loss for a 34 μm-long device. The same
group later achieved 23 GHz modulators with 9 dB extinction ratio and Vpp = 1 V (between −3 and −4 V
reverse bias) [124]. The insertion loss is partially due to the indirect gap absorption in the thick
Si0.1Ge0.9 buffer layer. The first waveguide-integrated Ge QCSE modulator was demonstrated on an
SOI substrate using a butt coupling approach. A 3.5 GHz bandwidth was demonstrated at 1460 nm
with an extinction ratio >3 dB, using an impressively compact device of 0.8 × 10 μm2 [125]. Although
the insertion loss is large due to the optical modal mismatch between the SOI waveguide and the Ge
QW active region, it could be improved by optimizing the device design and fabrication. The ultralow
capacitance of the device (0.75 fF) and low Vpp = 1 V are highly advantageous for energy-efficient
modulation. It is expected that the required DC voltage offset could also be reduced using the improved
QW design in [117]. Very recently, a Ge QCSE modulator with Si0.35Ge0.65 barrier layers working at
1300 nm has been reported, covering another important wavelength in optical communications [126].
In summary of this section, the results in Table 2 indicate that Ge and GeSi EAMs are ideal candidates
for high bandwidth, high energy-efficiency photonic modulators monolithically integrated on Si.
4.4. Ge Gain Media and Lasers
In 2007, we proposed combing tensile strain with n-type doping to compensate the energy difference
between the direct and indirect band gap of Ge, thereby achieving net optical gain for CMOS-compatible
LDs [19]. Optical gain [30,127] and lasing at room temperature have been demonstrated experimentally
under optical [128] and electrical pumping [129] since then. A recent review on this topic has been
presented in [66]. Therefore, in this section we will mainly discuss approaches to further improve the
performance of Ge-on-Si LDs based on the most recent results in this field.
The first electrically-pumped Ge-on-Si LD was based on a simple n+ Si/n+ Ge/p+ Si double
heterojunction structure [129]. Compared to optical pumping, the optical gain from the Ge active
region has to be increased in order to overcome the FCA losses in the electrodes. As such, the n-type
doping level in Ge is increased from 1 × 1019 cm−3 to >4 × 1019 cm−3 using delta-doped capping layers
as a dopant reservoir for drive-in diffusion (see Sections 2.2 and 3.1). Correspondingly, the
experimentally measured optical gain increases from 56 cm−1 to >500 cm−1, large enough to overcome
the FCA in heavily doped Si and metal electrodes for electrically pumped lasers. Figure 13a shows a
representative emission spectrum and L-I curve of an electrically pumped Ge-on-Si laser at room
temperature. In this case, the output optical power is ~1 mW at an emission wavelength of 1610 nm.
Depending on the thickness of the Ge layer and corresponding modal losses due to FCA in the
electrodes, the clamping conditions of lasing vary accordingly and lasing wavelengths in a broad spectrum
range of 1530–1660 nm has been observed. Such a broad gain spectrum is consistent with our
theoretical predictions [19], and it enables on-chip wavelength division multiplexing (WDM). The
observed maximum output power is 8 mW at room temperature. A slope efficiency of ηslope = 2% was
demonstrated in an un-optimized device structure. We estimated an internal loss of αi = 636 cm−1 due
to FCA (in both n+ Ge region and the Si electrodes) and a mirror loss of αm = 45 cm−1 [55]. Therefore,
the internal quantum efficiency (IQE) of this prototype Ge-on-Si diode is estimated to be
   m 
IQE   slope  i  ~ 30% (4)
 m 
The major limiting factor of IQE is likely to be Auger recombination, which tends to be dominated
by free holes in Ge. Therefore, reducing the required injected carrier density for transparency is the
key to further increasing the efficiency.
In analyzing the experimental data of the prototype Ge-on-Si LD, we noticed that the measured net
optical gain is significantly higher than our original theoretical prediction in [19], which assumed the
classical Drude model in extrapolating the FCA losses of n+ bulk Ge to λ = 1500–1700 nm. To
understand this discrepancy, we measured the absorption spectrum of 0.25% tensile-strained n+ Ge-on-Si
using Fourier transform infrared spectroscopy (FTIR). The absorption spectrum of 0.25% tensile
strained Ge with n = 1 × 1019 cm−3 is shown in Figure 13b after correction for the wavelength-dependent
backside roughness scattering and impurity infrared absorption of the Si substrate [130]. In Regime I,
the absorption drastically increases with wavelength at λ > 10 µm, indicating FCA. However,
experimentally measured FCA decreases much faster with wavelength than the λ2 Drude model at
λ < 15 µm, indicating the breakdown of Drude model in the mid- and near-infrared regime. Instead, the
experimental data is in good agreement with first principle calculations of the intra-L-valley (indirect
valley) absorption [131] that takes into account the band structure of tensile-strained Ge as well as
optical phonon, acoustic phonon and charged impurity scattering mechanisms. Therefore, we recalculated
intra-L-valley FCA at λ = 1550 nm using first principle approach and found that the Drude model
overestimated the intravalley FCA by an order of magnitude, as shown in Figure 13c [55]. At
n = 4 × 1019 cm−3 the FCA coefficient is 20–30 cm−1 in this wavelength range, compared to ~200 cm−1
from the extrapolation of the Drude model. This result agrees with both [131] and Haug’s quantum
mechanical analysis [132] showing that intravalley FCA at photon energies larger than the band gap is
in the order of 1 cm−1 for mid-1018 cm−3 n-type doping. It indicates that FCA losses are overestimated
in our previous theoretical analysis. On the other hand, Figure 13b shows that at λ < 11 μm, the
absorption starts to increase again. We proposed that this is due to the onset of intervalley scattering
absorption (IVSA) from the indirect L valleys to the direct Γ valley. The observed absorption edges in
Regime II for tensile strained Ge samples with different n-type doping levels are consistent with the
energy difference between the Γ valley and the Fermi level (Figure 13d, ΔE = EΓ − Ef). The shape of
the IVSA absorption curve is also similar to that of the n+ GaAs [133]. Note that the IVSA process in
n+ Ge promotes electrons from the indirect L valleys to the direct Г valley, which in turn enhances
light emission and optical gain from the direct gap transition of Ge. This process is exactly opposite to
the case of III-V semiconductors. The two factors discussed above explain why we observed a higher
net gain than the original theoretical modeling.
It is also interesting to note that IVSA has not been observed in previous literature on bulk n+ Ge
(0 strain) or slightly compressively strained (−0.07%) n+ Ge grown on GaAs reported recently [134].
Therefore, the 0.25% tensile strain in the n+ Ge films seems to enhance IVSA by decreasing the energy
difference between L and Г valleys and making more initial and final states accessible to phonon-assisted
IVSA. Indeed, pressure-dependent ab initio calculation shows that the contributions of various phonons
to the intervalley scattering are quite dependent on the energy separation between conduction band
valleys [135]. Further considering that the inherent optical gain from the direct gap transition of tensile
strained n+ Ge is 25× higher than the steady-state gain [136], enhancing L → Γ (indirect-to-direct)
intervalley scattering by tensile strain, Sn alloying and/or phonon engineering in nanostructures [137] may
help to significantly increase the optical gain of band engineered Ge LDs.
Figure 13. (a) Schematic cross-section, emission spectrum, and L-I curve of an electrically
pumped Ge-on-Si laser at room temperature. The width of the Ge waveguide is 1 μm and
the length is 300 μm. (b) Infrared absorption spectrum of 0.25% tensile strained Ge thin
film with n = 1 × 1019 cm−3. Regimes I–IV are respectively dominated by intra-L-valley
free carrier absorption, L → Г intervalley scattering absorption, indirect gap + intervalley
scattering absorption, and direct gap absorption. (c) Comparison of electron free carrier
absorption in tensile strained n+ Ge from first-principle model vs. the classical Drude
model. © 2013 IEEE. Reprinted, with permission, from [55]. (d) Theoretical and experimental
intervalley energy difference between L and Г valleys as a function of doping concentration.
The inset schematically illustrates the phenomenon of L → Г intervalley scattering absorption
in n+ Ge. Panels (b) and (d) are reproduced from [130] with permission from the Optical
Society of America under the Author’s Copyright Transfer Agreement.
Based on the results from the prototype device and the discussions in this paper, there are a number
of approaches to further enhance the performance of Ge LDs in future research. (1) Optimize the
balance among tensile strain, n-type doping, and Sn alloying. As mentioned in Section 2.4, each of
these methods has its own technical difficulties when used alone. However, one could utilize moderate
levels of tensile strain, n-type doping and Sn alloying to optimize optical gain and lasing without
inducing significant challenges in materials growth and fabrication. Considering the reports of 5.7%
uniaxial tensile strain [34], >4 × 1019 cm−3 n-type doping, and 9–10 at % Sn alloying achieved so far,
there is a lot of room available for such optimization. For 0.25% tensile strained Ge, our simulations
have shown that if the n-type doping is increased to >7 × 1019 cm−3 the threshold current density can
be decreased below 1 kA/cm2 at 1600–1700 nm, comparable to III-V double heterojunction lasers [55].
If tensile strain is increased or Sn is incorporated, the n-type doping level can be further decreased and
the performance improved. Interestingly, our optical gain simulations show that even for band-engineered
Ge right at the indirect-to-direct transition point (EgΓ = EgL), having an n-type doping in the order of
mid 1018 cm−3 still enhances the performance of the gain medium. Since the detrimental intra-L-valley
FCA of n+ Ge is an order lower than our original estimation, the increase of optical gain far exceeds
that of FCA losses when n-type doping is incorporated in this case. Moreover, lower injected carrier
density is needed to reach the same level of optical gain. Correspondingly, free hole absorption and
hole-hole-electron Auger recombination process—the most dominant loss mechanisms—are also
greatly reduced. All these factors can further reduce the threshold current density and increase lasing
efficiency. (2) Bufferless selective growth of high quality n+ Ge in narrow trenches ≤1 μm in width.
The highly dislocated low-temperature Ge buffer layer has been proved to be an undesirable sink for
the dopants in n+ Ge [63] and one of the limiting factors of carrier lifetime [138]. While the buffer
layer is necessary for blanket film growth, one can go without it for selective growths in narrow
trenches ≤1 μm in width [139,140] without affecting the growth profile or trench filling. This method
could improve the material quality at the Ge/Si interface and enhanced the carrier injection efficiency
in Ge LDs. (3) Incorporating tensile strain and/or Sn into Ge QWs. As discussed in Section 2.5, the
compressive strain in SiGe/Ge QWs as well as the quantum confinement effect itself increases the
energy difference between the direct and indirect gaps of Ge QWs. Therefore, under these conditions
SiGe/Ge QWs actually perform worse than bulk tensile-strained n+ Ge in optical gain and lasing [55].
With the recent progress in tensile strain engineering, we proposed that tensile strain has to be incorporated
into Ge QWs in order to counteract the detrimental effect of epitaxial compressive strain and the
adverse quantum confinement effect in SiGe/Ge QWs. Our modeling shows that 0.3% tensile-strained
Ge QWs with n > 4.5 × 1019 cm−3 can potentially achieve a threshold of ~500 A/cm2, comparable to
the performance of III-V LDs. The further incorporation of Sn alloying offers another handle to
optimize the gain and lasing performance from the QW structures. In this case, the energy gap
reduction due to tensile strain and Sn alloying can also be partially compensated by the quantum
confinement effect. This helps to avoid potential issues related to the leakage current and enhanced
Auger recombination in small band gap semiconductors. (4) Separate confinement of Γ vs. L electrons
in QW structures. As mentioned in Section 2.5, by adequately designing the band alignment between
the barrier and Ge QWs, one can enforce more quantum confinement for the L electrons (deeper well)
vs. the Γ electron (shallower well), thereby transforming Ge towards a direct gap semiconductor. This
approach overcomes the existing issue of SiGe/Ge quantum wells. One would prefer to use direct gap
barrier layers such as AlGaAs to implement such structures effectively. Although using III-V materials
for an active region is a significant challenge in CMOS compatible fabrications, using them as barrier
layers does not require such high material quality and may allow low-temperature, CMOS compatible
fabrications.
Another interesting topic to explore is the optimal relative positions of L vs. Γ valleys for on-chip
laser sources without active cooling. As mentioned in Section 2.4, a pseudo-direct gap configuration
can enhance the direct gap emission efficiency at high temperature and high injection levels due to
enhanced L-to-Γ valley excitation or scattering. Considering that the Auger recombination rate in
indirect valleys is orders of magnitude smaller than the direct valleys due to the momentum mismatch,
the same reason as its lower radiative recombination efficiency, it may be beneficial to “store” the
injected electrons in the L-valley and supply it to the direct Γ valley using the fast intervalley
scattering process (sub-ps) [135]. This approach helps to optimize the efficiency as well as the high
temperature performance of the band-engineered Ge LDs. Recent progress in Sn alloying and tensile
strain will allow us to systematically change the relative energy positions between L and Γ valleys for
this investigation. It is also important to study the Auger recombination process in these materials.
Spectrally resolved ultrafast pump-probe measurements will likely provide important fundamental
information on this aspect, especially distinguishing different processes during the relaxation of the
injected carriers.
5. Conclusions
We review the latest progress in monolithic Ge-on-Si active photonic materials and devices,
including band engineering, materials growth, device structures, and the current status of Ge PDs,
EAMs and LDs. Waveguide-integrated Ge PDs have achieved a performance comparable or even
better than their III-V counterparts. High performance of Ge PDs at 0 bias could enable photovoltaic
mode receivers with even higher energy efficiency. Ge EAMs have demonstrated ultralow power
consumption, high bandwidth, and a reasonably broad operation wavelength range for WDM
applications. Both Ge PDs and EAMs have entered system level applications, and they are being
commercialized by industry. Based on the results from the first prototype Ge-on-Si LDs and recent
investigations on band engineering, FCA and intervalley scattering, we find that it is highly feasible to
further improve the performance significantly by optimizing the synergy among tensile strain, Sn
alloying, n-type doping, and novel QW structures. Recent progress towards direct growth of single
crystal Ge and pseudo-single-crystalline, direct gap GeSn materials on dielectric layers also paves the
path towards monolithic 3D photonic integration. Clearly, these high-performance monolithic Ge-on-Si
active photonic devices will enable large-scale EPICs on Si.
Acknowledgments
The author’s work on waveguide-integrated GeSi EAMs and Ge PDs had been sponsored under the
Defense Advanced Research Projects Agency’s (DARPA) EPIC program supervised by Jagdeep Shah
in the Microsystems Technology Office (MTO) under Contract No. HR0011-05-C-0027 from 2003
to 2008. From 2006 to 2011, the authors’ research on Ge lasers was supported by U.S. Air Force
Office of Scientific Research’s (AFOSR) Multidisciplinary University Research Initiative (MURI) on
Silicon-Based Lasers supervised by Gernot Pomrenke. From 2010–2012, the research on

electrically-pumped Ge lasers was also supported by the Fully Laser Integrated Photonics (FLIP)
program under APIC Corporation, supervised by Raj Dutt, and sponsored by the Naval Air Warfare
Center-Aircraft Division (NAWC-AD) under OTA N00421-03-9-0002. The author’s current work on
low-temperature crystallization of pseudo-single-crystalline GeSn on dielectric layers for 3D photonic
integration has been sponsored by the National Science Foundation (NSF) under the grant number
DMR-1255066.
Conflicts of Interest
The author declares no conflict of interest.
References
1. Lannoo, B.; Lambert, S.; van Heddeghem, W.; Pickavet, M.; Kuipers, F.; Koutitas, G.; Niavis, H.;
Satsiou, A.; Beck, M.T.; Fischer, A.; et al. Overview of ICT Energy Consumption. Available
online: http://www.internet-science.eu/sites/internet-science.eu/files/biblio/EINS_D8%201_final.pdf
(accessed on 22 April 2014).
2. Klein, T.E. Sustainable ICT Networks: The GreenTouch Vision. Green Research at Alcatel-Lucent.
Available online: http://www.greentouch.org/uploads/documents/3%20Thierry%20Klein_EU%
20SEW%20-%20The%20GT%20Vision%20-%20v2.pdf (accessed on 22 April 2014).
3. Kimerling, L.C. Microphotonics: The Next Platform for the Information Age. 2011 MIT in Japan
Conference. Available online: http://ilp.mit.edu/media/conferences/2011-japan/Kimerling.pdf
(accessed on 22 April 2014).
4. Soref, R.A. The past, present and future of silicon photonics. IEEE J. Sel. Top. Quantum Electron.
2006, 12, 1678–1687.
5. Kirchain, R.; Kimerling, L.C. A roadmap for nanophotonics. Nat. Photonics 2007, 1, 303–305.
6. Benner, A. Optical interconnect opportunities in supercomputers and high-end computing.
In Proceedings of the 2012 Optical Fiber Communication Conference (OFC), Los Angeles, CA,
USA, 4–8 March 2012.
7. Liu, A.; Jones, R.; Liao, L.; Samara-Rubio, D.; Rubin, D.; Cohen, O.; Nicolaescu, R.; Paniccia, M.
A high-speed silicon optical modulator based on metal-oxide-semiconductor capacitor. Nature
2004, 427, 615–618.
8. Xu, Q.; Schmidt, B.; Pradhan, S.; Lipson, M. Micrometre-scale silicon electro-optic modulator.
Nature 2005, 435, 325–327.
9. Timurdogan, E.; Sorace-Agaskar, C.M.; Sun, J.; Hosseini, E.S.; Biberman, A.; Watts, M.R.
A One Femtojoule Athermal Silicon Modulator. arXiv:1312.2683. Available online:
http://arxiv.org/ftp/arxiv/papers/1312/1312.2683.pdf (accessed on 22 April 2014).
10. Lin, H.; Ogbuu, O.; Liu, J.F.; Zhang, L.; Michel, J.; Hu, J.J. Breaking the Energy-Bandwidth
Limit of Electrooptic Modulators: Theory and a Device Proposal. J. Light. Technol. 2013, 31,
4029–4036.
11. Pavesi, L.; dal Negro, L.; Mazzoleni, C.; Franzo, G.; Priolo, F. Optical gain in silicon
nanocrystals. Nature 2000, 408, 440–444.
12. Wilcoxon, J.P.; Samara, G.A. Tailorable, visible light emission from silicon nanocrystals.
Appl. Phys. Lett. 1999, 74, 3164.
13. Geis, M.W.; Spector, S.J.; Grein, M.E.; Yoon, J.U.; Lennon, D.M.; Lyszczarz, T.M. Silicon
waveguide infrared photodiodes with >35 GHz bandwidth and phototransistors with 50 AW-1
response. Opt. Express 2014, 22, 9150–9158.
14. Souhan, B.; Grote, R.R.; Driscoll, J.B.; Lu, M.; Stein, A.; Bakhru, H.; Osgood, R.M.
Metal-semiconductor-metal ion-implanted Si waveguide photodetectors for C-band operation.
Opt. Express 2009, 17, 51–9158.
15. Cheng, K.Y.; Anthony, R.; Kortshagen, U.R.; Holmes, R.J. High-efficiency silicon nanocrystal
light-emitting devices. Nano Lett. 2011, 11, 1952–1956.
16. Physics of Group IV Elements and III–V Compounds, Landolt-Börnstein: Numerical Data and
Functional Relationships in Science and Technology; Madelung, O., Ed.; Springer: Berlin,
Germany, 1982; Volume 17a.
17. Lee, M.J.; Fitzgerald, E.A. Strained Si, SiGe, and Ge channels for high-mobility
metal-oxide-semiconductor field-effect transistors. J. Appl. Phys. 2005, 97, 011101.
18. Liu, J.F.; Camacho-Aguilera, R.; Sun, X.; Wang, X.X.; Cai, Y.; Kimerling, L.C.; Michel, J.
Reprinted from Thin Solid Films, Ge-on-Si Optoelectronics; Elsevier: Amsterdam, The Netherlands,
2011; Volume 520, pp. 3354–3360.
19. Liu, J.F.; Sun, X.; Pan, D.; Wang, X.X.; Kimerling, L.C.; Koch, T.L.; Michel, J. Tensile-strained,
n-type Ge as a gain medium for monolithic laser integration on Si. Opt. Express 2007, 15,
11272–11277.
20. Soref, R.A.; Friedman, L. 1993b Direct gap Ge/GeSn/Si and GeSn/Ge/Si heterostructures.
Superlattice Microstruct. 1993, 14, 189–193.
21. El Kurdi, M.; Fishman, G.; Sauvage, S.; Boucaud, P. Band structure and optical gain of
tensile-strained germanium based on a 30 band k·p formalism. J. Appl. Phys. 2010, 107, 013710.
22. Aldaghri, O.; Ikonic, Z.; Kelsall, R.W. Optimum strain configurations for carrier injection in near
infrared Ge lasers. J. Appl. Phys. 2012, 111, 053106.
23. Ishikawa ,Y.; Wada, K.; Cannon, D.D.; Liu, J.F.; Luan, H.C.; Kimerling, L.C. Strain-induced
direct bandgap shrinkage in Ge grown on Si substrate. Appl. Phys. Lett. 2003, 82, 2044–2046.
24. Liu, J.F.; Cannon, D.D.; Wada, K.; Ishikawa, Y.; Danielson, D.T.; Jongthammanurak, S.;
Michel, J.; Kimerling, L.C. Deformation potential constants of biaxially tensile stressed Ge
epitaxial films on Si(100). Phys. Rev. B 2004, 70, 155309.
25. Hartmann, J.M.; Abbadie, A.; Papon, A.M.; Holliger, P.; Roland, G.; Billon, T.; Fédéli, J. M.;
Rouvière, M.; Vivien, L.; Laval, S. Reduced pressure-chemical vapor deposition of Ge thick
layers on Si(001) for 1.3–1.55 µm photodetection. J. Appl. Phys. 2004, 95, 5905–5913.
26. Fang, Y.Y.; Tolle, J.; Roucka, R.; Chizmeshiya, A.V.G.; Kouvetakis, J.; D’Costa, V.R.;
Menendez, J. Perfectly tetragonal, tensile-strained Ge on Ge1−ySny buffered Si(100). Appl. Phys. Lett.
2007, 90, 061915.
27. Takeuchi, S.; Shimura, Y.; Nakatsuka, O.; Zaima, S.; Ogawa, M.; Sakai, A. Growth of highly
strain-relaxed Ge1−xSnx/virtual Ge by a Sn precipitation controlled compositionally step-graded
method. Appl. Phys. Lett. 2008, 93, 231916.
28. Liu, J.F.; Cannon, D.D.; Wada, K.; Ishikawa, Y.; Danielson, D.T.; Jongthammanurak, S.; Michel,
J.; Kimerling, L.C. Silicidation-induced band gap shrinkage in Ge epitaxial films on Si. Appl.
Phys. Lett. 2004, 84, 660–662.
29. Kuroyanagi, R.; Ishikawa, Y.; Tsuchizawa, T.; Wada, K. Controlling strain in Ge on Si for EA
modulators. In Proceedings of the 8th IEEE International Conference on Group IV Photonics,
London, UK, 14–16 September 2011.
30. De Kersauson, M.; El Kurdi, M.; David, S.; Checoury, X.; Fishman, G.; Sauvage, S.; Jakomin,
R.; Beaudoin, G.; Sagnes, I.; Boucaud, P. Optical gain in single tensile-strained germanium
photonic wire. Opt. Express 2011, 19, 17925–17934.
31. Jain, J.R.; Hryciw, A.; Baer, T.M.; Miller, D.A.B.; Brongersma, M.L.; Howe, R.T.
A Micromachining-based technology for enhancing germanium light emission via tensile strain.
Nat. Photonics 2012, 6, 398–405.
32. Süess, M.J.; Geiger R.; Minamisawa, R.A.; Schiefler, G.; Frigerio, J.; Chrastina, D.; Isella, G.;
Spolenak, R.; Faist, J.; Sigg, H. Analysis of enhanced light emission from highly strained
germanium microbridges. Nat. Photonics 2013, 7, 466–472.
33. Geiger, R.; Süess, M.J.; Bonzon, C.; Spolenak, R.; Faist, J.; Sigg, H. Strained Ge microbridges to
obtain a direct bandgap laser. In Proceedings of the 7th International SiGe Technology and
Device Meeting, Singapore, Singapore, 2–4 June 2014.
34. Sukhdeo, D.S.; Nam, D.; Kang, J.H.; Brongersma, M.L.; Saraswat, K.C. Direct bandgap
germanium-on-silicon inferred from 5.7% <100> uniaxial tensile strain. Photonics Res. 2014, 2,
A8–A13.
35. Capellini, G.; Reich, C.; Guha, S.; Yamamoto, Y.; Lisker, M.; Virgilio, M.; Ghrib, A.; el Kurdi, M.;
Boucaud, P.; Tillack, B.; et al. Tensile Ge microstructures for lasing fabricatedby means of a
silicon complementary metal oxide-semiconductor process. Opt. Express 2014, 22, 399–410.
36. Haynes, J.R.; Nilsson, N.G. The direct radiative transitions in germanium and their use in the
analysis of lifetime. In Proceedings of the VIIth International Conference on Physics of
Semiconductors, Paris, France, 1964; pp. 21–31.
37. Spitzer, W.G.; Trumbore, F.A.; Logan, R.A. Properties of heavily doped n-type germanium.
J. Appl. Phys. 1961, 32, 1822–1830.
38. Newman, R.; Tyler, W.W. Effect of impurities on free-hole infrared absorption in p-type
germanium. Phys. Rev. 1957, 105, 885–886.
39. Carroll, L.; Friedli, P.; Neuenschwander, S.; Sigg, H.; Cecchi, S.; Isa, F.; Chrastina, D.; Isella, G;
Fedoryshyn, Y.; Faist, J. Direct-gap gain and optical absorption in germanium correlated to the
density of photoexcited carriers doping, and strain. Phys. Rev. Lett. 2012, 109, 057402.
40. Jenkins, J.W.; Dow, J.D. Electronic properties of metastable Ge1−xSnx alloys. Phys. Rev. B 1987,
36, 7994–8000.
41. He, G.; Atwater, H.A. Interband transitions in SnxGe1−x alloys. Phys. Rev. Lett. 1997, 79, 1937–1940.
42. Soref, R.; Kouvetakis, J.; Tolle, J.; Menendez, J.; D’Costa, V. Advances in SiGeSn technology. J.
Mater. Res. 2007, 22, 3281–3291.
43. Mathews, J.; Roucka, R.; Xie, J.Q.; Yu, S.Q.; Menéndez, J.; and Kouvetakis, J. Extended
performance GeSn/Si(100) p-i-n photodetectors for full spectral range telecommunication
applications. Appl. Phys. Lett. 2009, 95, 133506.
44. D’Costa, V.R.; Cook, C.S.; Birdwell, A.G.; Littler, C.L.; Canonico, M.; Zollner, S.; Kouvetakis, J.;
Menéndez, J. Optical critical points of thin-film Ge1−ySny alloys: A comparative
Ge1−ySny/Ge1−xSix study. Phys. Rev. B 2006, 73, 125207.
45. Chen, R.; Lin, H.; Huo, Y.; Hitzman, C.; Kamins, T.I.; Harris, J.S. Increased photoluminescence
of strain-reduced, high-Sn composition Ge1−xSnx alloys grown by molecular beam epitaxy.
Appl. Phys. Lett. 2011, 99, 181125.
46. Oehme, M.; Buca, D.; Kostecki, K.; Wirths, S.; Hollaender, B.; Kasper, E.; Schulze, J. Epitaxial
growth of highly compressively strained GeSn alloys up to 12.5% Sn. J. Cryst. Growth 2013,
384, 71–76.
47. Li, H.; Brouillet, J.; Salas, A.; Wang, X.X.; Liu, J.F. Low Temperature Growth of High
Crystallinity GeSn on Amorphous Layers for Advanced Optoelectronics. Opt. Mater. Express
2013, 3, 1385–1396.
48. Oehme, M.; Kostecki, K.; Schmid, M.; Oliveira, F.; Kasper, E.; Schulze, J. Epitaxial growth of
strained and unstrained GeSn alloys up to 25% Sn. Thin Solid Films 2014, 557, 169–172.
49. Van de Walle, C.G. Band lineups and deformation potentials in model-solid theory. Phys. Rev. B
1989, 39, 1871–1883.
50. Sun, X.; Liu, J.F.; Kimerling, L.C.; Michel, J. Direct gap photoluminescence of n-type
tensile-strained Ge-on-Si. Appl. Phys. Lett. 2009, 95, 011911.
51. Bessette, J.; Camacho-Aguilera, R.; Cai, Y.; Kimerling, L.C.; Michel, J. Optical characterization
of Ge-on-Si laser gain media. In Proceedings of the 8th IEEE International Conference on Group
IV Photonics, London, UK, 14–16 September 2011; paper # P1.25.
52. Sun, X.; Liu, J.F.; Kimerling, L.C.; Michel, J. Room-temperature direct bandgap electroluminesence
from Ge-on-Si light-emitting diodes. Opt. Lett. 2009, 34, 1198–1200.
53. Cheng, S.L.; Lu, J.; Shambat, G.; Yu, H.Y.; Saraswat, K.; Vuckovic, J.; Nishi, Y. Room
Temperature 1.6 μm electroluminescence from Ge light emitting Diode on Si Substrate.
Opt. Express 2009, 17, 10019–10024.
54. Arguirov, T.; Kittler, M.; Oehme, M.; Abrosimov, N.V.; Kasper, E.; Schulze, J. Room
temperature direct band-gap emission from an unstrained Ge p-i-n LED on Si. Solid. State. Phenom.
2011, 178–179, 25–30.
55. Cai, Y.; Han, Z.; Wang, X.X.; Camacho-Aguilera, R.E.; Kimerling, L.C.; Michel, J.; Liu, J.F.
Analysis of threshold current behavior for bulk and quantum well germanium laser structures.
IEEE J. Sel. Top. Quantum Electron. 2013, 19, 1901009.
56. Froyen, S.; Wood, D.M.; Zunger, A. Electronic-structure of ultrathin SinGen strained
superlattices—The possibility of direct band-gaps. Thin Solid Films 1989, 183, 33–48.
57. Kholod, A.N.; Ossicini, S.; Borisenko, V.E.; Arnaud d’Avitaya, F. True direct gap absorption in
germanium quantum films. Phys. Rev. B. 2002, 65, 115315.
58. Virgilio, M.; Grosso, G. Type-I alignment and direct fundamental gap in SiGe based
hetero-structures. J. Phys. C. Condens. Matter 2006, 18, 1021–1031.
59. Baribeau, J.M.; Jackman, T.E.; Houghton, D.C.; Maigné, P.; Denhoff, M.W. Growth and
Characterization of Si1−xGex and Ge Epilayers on (100) Si. J. Appl. Phys. 1988, 63, 5738–5746.
60. Colace, L.; Masini, G.; Galluzzi, F.; Assanto, G.; Capellini, G.; di Gaspare, L.; Evangelisti, F.
Ge/Si (001) photodetector for infrared light. Solid State Phenom. 1997, 54, 55–58.
61. Luan, H.C.; Lim, D.R.; Lee, K.K.; Chen, K.M.; Sandland, J.G.; Wada, K.; and Kimerling, L.C.
High-quality Ge epilayers on Si with low threading-dislocation densities. Appl. Phys. Lett. 1999,
75, 2909–2911.
62. Luan, H.-C. Ge Photodetectors for Si Microphotonics. Ph.D. Thesis, Massachusetts Institute of
Technology, Cambridge, MA, USA, 2001; pp. 74–76, 98–100.
63. Tachikawa, M.; Yamaguchi, M. Film thickness dependence of dislocation reduction in GaAs-on-Si
substrates. Appl. Phys. Lett. 1990, 54, 584–486.
64. Romanova, A.E.; Pompe, W.; Beltz, G.E.; Speck, J.S. An approach to threading dislocation
“reaction kinetics”. Appl. Phys. Lett. 1996, 69, 3342–3344.
65. Yamamoto, Y.; Kozlowski, G.; Zaumseil, P.; Tillack, B. Low threading dislocation Ge on Si by
combining deposition and etching. Thin Solid Films 2012, 520, 3216–3221.
66. Liu, J.F.; Kimerling, L.C.; Michel, J. Monolithic Ge-on-Si lasers for large-scale electronic-photonic
integration. Semicond. Sci. Technol. 2012, 27, 094006.
67. Liu, J.F.; Camacho-Aguilera, R.; Sun, X.; Wang, X.X.; Cai, Y.; Kimerling, L.C.; Michel, J.
Ge-on-Si Optoelectronics. Thin Solid Films 2012, 520, 3354–3360.
68. Michel, J.; Liu, J.F.; Kimerling, L.C. High-Performance Ge-on-Si Photodetectors. Nat. Photonics
2010, 4, 527–534.
69. Camacho-Aguilera, R.E.; Cai, Y.; Bessette, J.T.; Kimerling, L.C.; Michel, J. High active carrier
concentration in n-type, thin film Ge using delta-doping. Opt. Mater. Express 2012, 2, 1462–1469.
70. Cai, Y.; Camacho-Aguilera, R.E.; Bessette, J.T.; Kimerling, L.C.; Michel, J. High phosphorous
doped germanium: Dopant diffusion and modeling. J. Appl. Phys. 2012, 112, 034509.
71. Bauer, M.; Taraci, J.; Tolle, J.; Chizmeshya, A.V.G.; Zollner, S.; Smith, D. J.; Menendez, J.;
Hu, C.W.; Kouvetakis, J. Ge-Sn semiconductors for band-gap and lattice engineering.
Appl. Phys. Lett. 2002, 81, 2992–2994.
72. Mathews, J.; Beeler, R.T.; Tolle, J.; Xu, C.; Roucka, R.; Kouvetakis, J.; Menendez, J. Direct-gap
photoluminescence with tunable emission wavelength in Ge1−ySny alloys on silicon. Appl. Phys. Lett.
2010, 97, 221912.
73. Liu, Y.; Deal, M.D.; Plummer, J.D. High-quality single-crystal Ge on insulator by liquid-phase
epitaxy on Si substrate. Appl. Phys. Lett. 2004, 84, 2563–2565.
74. Miyao, M.; Toko, K.; Tanaka, T.; Sadohm, T. High-quality single-crystal Ge stripes on quartz
substrate by rapid-melting-growth. Appl. Phys. Lett. 2009, 95, 022115.
75. Assefa, S.; Xia, F.N.; Bedell, S.W.; Zhang, Y.; Topuria, T.; Rice, P.M.; Vlasov, Y.A.
CMOS-integrated high-speed MSM germanium waveguide photodetector. Opt. Express 2010, 18,
4986–4999.
76. Matsue, M.; Yasutake, Y.; Fukatsu, S.; Hosoi, T.; Shimura, T.; Watanabe, H. Strain-induced
direct band gap shrinkage in local Ge-on-insulator structures fabricated by lateral liquid-phase
epitaxy. Appl. Phys. Lett. 2014, 104, 031106.
77. Kanno, H.; Atsushi, K.; Sadoh, T.; Miyao, M. Ge-enhanced MILC velocity in a-Ge/a-Si/SiO2
layered structure. Mater. Sci. Semicond. Process. 2005, 8, 83–88.
78. Yu, C.H.; Yeh, P.H.; Cheng, S.L.; Chen, L.J.; Cheng, L.W. Metal-induced crystallization of
amorphous Si1−xGex by rapid thermal annealing. Thin Solid Films 2004, 469–470, 356–360.
79. Yang, C.M.; Atwater, H.A. Selective solid phase crystallization for control of grain size and
location in Ge thin films on silicon dioxide. Appl. Phys. Lett. 1996, 68, 3392–3394.
80. Schropp, R.E.I.; Carius, R.; Beaucarne, G. Amorphous silicon, microcrystalline silicon, and thin-
film polycrystalline silicon solar cells. MRS Bull. 2007, 32, 219–224.
81. McComber, K.; Duan, X.; Liu, J.F.; Michel, J.; Kimerling, L.C. Single-crystal germanium
growth on amorphous silicon. Adv. Funct. Mater. 2012, 22, 1048–1057.
82. Liu, J.F.; Dong, P.; Jongthammanurak, S.; Wada, K.; Kimerling, L.C.; Michel, J. Design of
monolithically integrated GeSi electroabsorption modulators and photodetectors on an SOI
platform. Opt. Express 2007, 15, 623–628.
83. Liu, J.F.; Beals, M.; Pomerene, A.; Bernardis, S.; Sun, R.; Cheng, J.; Kimerling, L.C.; Michel, J.
Waveguide integrated, ultra-low energy GeSi electro-absorption modulators. Nat. Photonics
2008, 2, 433–437.
84. Beals, M.; Michel, J.; Liu, J.F.; Ahn, D.H.; Sparacin, D.; Sun, R.; Hong, C.Y.; Kimerling, L.C.;
Pomerene, A.; Carothers, D.; et al. Process flow innovations for photonic device integration in
CMOS. Proc. Int. Soc. Opt. Eng. 2008, 6898, 689804.
85. Vivien, L.; Osmond, J.; Fédéli, J.M.; Delphine, M.M.; Crozat, P.; Damlencourt, J.F.; Cassan, E.;
Lecunff, Y.; Laval, S. 42 GHz p.i.n Germanium photodetector integrated in a silicon-on-insulator
waveguide. Opt. Express 2009, 17, 6252–6257.
86. Feng, D.; Liao, S.; Dong, P.; Feng, N.N.; Liang, H.; Zheng, D.W.; Kung, C.C.; Fong, J.; Shafiiha,
R.; Cunningham, J.; et al. High-speed Ge photodetector monolithically integrated with large
cross-section silicon-on-insulator waveguide. Appl. Phys. Lett. 2009, 95, 261105.
87. Ren, S.; Kamins, T.I.; Miller, D.A.B. Thin dielectric spacer for the monolithic integration of bulk
germanium or germanium quantum wells with silicon-on-insulator waveguides. IEEE. Photonics J.
2011, 3, 739–747.
88. Liu, J.F.; Pan, D.; Jongthammanurak, S.; Ahn, D.; Hong, C.Y.; Beals, M.; Kimerling, L.C.;
Michel, J.; Pomerene, A.T.; Hill, C.; et al. Waveguide-Integrated Ge p-i-n Photodetectors on SOI
Platform. In Proceedings of the 2006 IEEE International Conference on Group IV Photonics,
Ottawa, ON, Canada, 13–15 September 2006; paper # ThA2.
89. Ahn, D.; Kimerling, L.C.; Michel, J. Evanescent coupling device design for
waveguide-integrated group IV photodetectors. J. Light. Technol. 2010, 28, 3387–3394.
90. Ahn, D.H.; Kimerling, L.C.; Michel, J. Efficient evanescent wave coupling conditions for
waveguide-integrated thin-film Si/Ge photodetectors on silicon-on-insulator/germanium-on-insulator
substrates. J. Appl. Phys. 2011, 110, 083115.
91. Lim, A.E.-J.; Liow, T.Y.; Qing, F.; Duan, N.; Ding, L.; Yu, M.; Lo, G.-Q.; Kwong, D.L. Novel
evanescent-coupled germanium electro-absorption modulator featuring monolithic integration
with germanium p-i-n photodetector. Opt. Express 2011, 19, 5040–5046.
92. Wang, X.X.; Liu, J.F. Step-coupler for efficient waveguide coupling to Ge/Si avalanche
photodetectors. IEEE. Photonics Technol. Lett. 2011, 23, 146–148.
93. Liu, J.F.; Michel, J.; Giziewicz, W.; Pan, D.; Wada, K.; Cannon, D.D.; Kimerling, L.C.; Chen, J.;
Ilday, F.O.; Kaertner, F.X.; et al. High-performance, tensile-strained Ge p-i-n photodetectors on a
Si platform. Appl. Phys. Lett. 2005, 87, 103501.
94. Jutzi, M.; Berroth, M.; Wohl, G.; Oehme, M.; Kasper, E. Ge-on-Si vertical incidence photodiodes
with 39-G Hz bandwidth. IEEE Photonics Technol. Lett. 2005, 17, 1510–1512.
95. Morse, M.; Dosunmu, O.; Sarid, G.; Chetrit, Y. IEEE. Photonics Technol. Lett. 2006, 18,
2442–2444.
96. Klinger, S.; Berroth, M.; Kaschel, M.; Oehme, M.; Kasper, E. Ge-on-Si p-i-n photodiodes with a
3-dB bandwidth of 49 GHz. IEEE Photonics Technol. Lett. 2009, 21, 920–922.
97. DiLello, N.A.; Johnstone, D.K.; Hoyt, J.L. Characterization of dark current in Ge-on-Si
photodiodes. J. Appl. Phys. 2012, 112, 054506.
98. Colace, L.; Ferrara, P.; Assanto, G.; Fulgoni, D.; Nash, L. Low dark-current germanium-on
silicon near-infrared detectors. IEEE. Photonics Technol. Lett. 2007, 19, 1813–1815.
99. Oehme, M.; Kostecki, K.; Ye, K.; Bechler, S.; Ulbricht, K.; Schmid, M.; Kaschel, M.; Gollhofer, M.;
Körner, R.; Zhang, W.; et al. GeSn-on-Si normal incidence photodetectors with bandwidths more
than 40 GHz. Opt. Express 2014, 22, 839–846.
100. Deliwala, S. Integrated optical receivers. In Proceedings of the 2011 IEEE International
Conference on Group IV Photonics, London, UK, 14 September 2011.
101. Kim, I.G.; Jang, K.S.; Joo, J.; Kim, S.; Kim, S.; Choi, K.S.; Oh, J.H.; Kim, S.A.; Kim, G.
High-performance photoreceivers based on vertical-illumination type Ge-on-Si photodetectors
operating up to 43 Gb/s at λ ~ 1550 nm. Opt. Express 2013, 21, 30718.
102. Kang, Y.; Liu, H.D.; Morse, M.; Paniccia, M.J.; Zadka, M.; Litski, S.; Sarid, G.; Pauchard, A.;
Kuo, Y.H.; Chen, H.W.; et al. Monolithic germanium/silicon avalanche photodiodes with
340 GHz gain-bandwidth product. Nat. Photonics 2009, 3, 59–63.
103. Huang, M.; Cai, P.; Wang, L.; Shi, T.; Chen, W.; Li, S.; Hou, G.; Hong, C.-Y.; Dong, P.
Development of Si photonics technology: Ge/Si avalanche photodiode for PON applications.
In Proceedings of the 2014 Optical Fiber Communication Conference (OFC 2014),
San Francisco, CA, USA, 9–13 March 2014; paper # Tu2C2.
104. Ahn, D.; Hong, C.Y.; Liu, J.F.; Giziewicz, W.; Beals, M.; Kimerling, L.C.; Michel, J. High
performance, waveguide integrated Ge photodetectors. Opt. Express 2007, 15, 3916–3921.
105. Vivien, L.; Rouvière, M.; Fédéli, J.M.; Marris-Morini, D.E.; Damlencourt, J.F.; Mangeney, J.;
Crozat, P.; el Melhaoui, L.; Cassan, E.; le Roux, X.; et al. High speed and high responsivity
germanium photodetector integrated in a silicon-on-insulator microwaveguide. Opt. Express
2007, 15, 9843–9848.
106. Chen, L.; Preston, K.; Manipatruni, S.; Lipson, M. Integrated GHz silicon photonic interconnect
with micrometer-scale modulators and detectors. Opt. Express 2009, 18, 15248–15256.
107. Novack, A.; Gould, M.; Yang, Y.; Xuan, Z.; Streshinsky, M.; Liu, Y.; Capellini, G.; Lim, A.E.J.;
Lo, G.Q.; Baehr-Jones, T.; et al. Germanium photodetector with 60 GHz bandwidth using
inductive gain peaking. Opt. Express 2013, 21, 28387–28393.
108. Vivien, L.; Polzer, A.; Marris-Morini, D.; Osmond, J.; Hartmann, J.M.; Crozat, P.; Cassan, E.;
Kopp, C.; Zimmermann, H.; Fédéli, J.M. Zero-bias 40 Gbit/s germanium waveguide
photodetector on silicon. Opt. Express 2012, 20, 1096–1101.
109. Feng, D.D.; Qian, W.; Liang, H.; Luff, B.J.; Asghari, M. High-speed receiver technology on the
SOI platform. IEEE. J. Sel. Top. Quantum Electron. 2013, 19, 3800108.
110. Kang, Y.M.; Morse, M.; Paniccia, M.J.; Zadka, M.; Saad, Y.; Sarid, G.; Pauchard, A.;
Zaoui, W.S.; Chen, H.W.; Dai, D.; et al. Monolithic Ge/Si avalanche photodiodes. In
Proceedings of the 2009 IEEE International Conference on Group IV Photonic, San Francisco,
CA, USA, 9–11 September 2009; pp. 25–27.
111. Assefa, S.; Xia, F.; Vlasov, Y.A. Reinventing germanium avalanche photodetector for
nanophotonic on-chip optical interconnects. Nature 2010, 464, 80–85.
112. Liao, S.; Feng, N.-N.; Feng, D.; Dong, P.; Shafiiha, R.; Kung, C.-C.; Liang, H.; Qian, W.; Liu, Y.;
Fong, J.; et al. 36 GHz submicron silicon waveguide germanium photodetector. Opt. Express
2011, 19, 10967–10972.
113. Pollak, F.H. Study of semiconductor surfaces and interfaces using electromodulation.
Surf. Interface Anal. 2001, 31, 938–953.
114. Chuang, S.L. Physics of Optoelectronic Devices; Wiley: New York, NY, USA, 1995; Chapter 13.
115. Jongthammanurak, S.; Liu, J.F.; Wada, K.; Cannon, D.D.; Danielson, D.T.; Pan, D.;
Kimerling, L.C.; Michel, J. Large Electro-optic Effect in Tensile Strained Ge-on-Si Films.
Appl. Phys. Lett. 2006, 89, 161115.
116. Kuo, Y.H.; Lee, Y.K.; Ge, Y.; Ren, S.; Roth, J.E.; Kamins, T.I.; Miller, D.A.B.; Harris, J.S.
Strong quantum-confined Stark effect in germanium quantum-well structures on silicon. Nature
2005, 437, 1334–1336.
117. Edwards, E.H.; Lever, L.; Fei, E.T.; Kamins, T.I.; Ikonic, Z.; Harris, J.S.; Kelsall, R.W.;
Miller, D.A.B. Low-voltage broad-band electroabsorption from thin Ge/SiGe quantum wells
epitaxially grown on silicon. Opt. Express 2013, 21, 867–876.
118. Doerr, C.R.; Chen, L.Z.; Vermeulen, D.; Nielsen, T.; Azemati, S.; Stulz, S.; McBrien, G.; Xu, X.;
Mikkelsen, B.; Givehchi, M.; et al. Single-chip silicon photonics 100-Gb/s coherent transceiver.
In Proceedings of the 2014 Optical Fiber Communication Conference, San Francisco, CA, USA,
9–13 March 2014.
119. Feng, N.N.; Feng, D.Z.; Liao, S.; Wang, X.; Dong, P.; Liang, H.; Kung, C.C.; Qian, W.; Fong, J.;
Shafiiha, R.; et al. 30 GHz Ge electro-absorption modulator integrated with 3 μm silicon-on-insulator
waveguide. Opt. Express 2011, 19, 7062–7067.
120. Feng, D.Z.; Liao, S.; Liang, H.; Fong, J.; Bijlani, B.; Shafiiha, R.; Luff, B.J.; Luo, Y.;
Cunningham, J.E.; Krishnamoorthy, A.V.; et al. High speed GeSi electro-absorption modulator
at 1550 nm wavelength on SOI waveguide. Opt. Express 2012, 20, 22224–22232.
121. Krishnamoorthy, A.V.; Zheng, X.; Feng, D.; Lexau, J.; Buckwalter, J.F.; Thacker, H.D.; Liu, F.;
Luo, Y.; Chang, E.; Amberg, P.; et al. A low-power, high-speed, 9-channel germaniumsilicon
electro-absorption modulator array integrated with digital CMOS driver and wavelength
multiplexer. Opt. Express 2014, 22, 12289–12295.
122. Roth, J.E.; Fidaner, O.; Schaevitz, R.K.; Kuo, Y.H.; Kamins, T.I.; Harris, J.S.; Miller, D.A.B.
Optical modulator on silicon employing germanium quantum wells. Opt. Express 2007, 15,
5851–5859.
123. Chaisakul, P.; Marris-Morini, D.; Isella, G.; Chrastina, D.; le Roux, X.; Edmond, S.;
Coudevylle, J.R.; Cassan, E.; Vivien, L. Polarization dependence of quantum-confined Stark
effect in Ge/SiGe quantum well planar waveguides. Opt. Lett. 2011, 36, 1794–1796.
124. Chaisakul, P.; Marris-Morini, D.; Rouifed, M.-S.; Isella, G.; Chrastina, D.; Frigerio, J.;
le Roux, X.; Edmond, S.; Coudevylle, J.-R.; Vivien, L. 23 GHz Ge/SiGe multiple quantum well
electro-absorption modulator. Opt. Express 2012, 20, 3219–3224.
125. Ren, S.; Rong, Y.; Claussen, S.A.; Schaevitz, R.K.; Kamins, T.I.; Harris, J.S.; Miller, D.A.B.
Ge/SiGe quantum well waveguide modulator monolithically integrated with SOI waveguides.
IEEE Photonics Technol. Lett. 2012, 24, 461–463.
126. Rouifed, M.S.; Marris-Morini, D.; Chaisakul, P.; Frigerio, J.; Isella, G.; Chrastina, D.; Edmond, S.;
Roux, X.L.; Coudevylle, J.R.; Bouville, D.; et al. Advances toward Ge/SiGe quantum-well
waveguide modulators at 1.3 μm. IEEE. J. Sel. Top. Quantum Electron. 2014, 20, 3400207.
127. Liu, J.F.; Sun, X.C.; Kimerling, L.C.; Michel, J. Direct-gap optical gain of Ge on Si at room
temperature. Opt. Lett. 2009, 34, 1738–1740.
128. Liu, J.F.; Sun, X.; Camacho-Aguilera, R.; Kimerling, L.C.; Michel, J. Ge-on-Si laser operating at
room temperature. Opt. Lett. 2010, 35, 679–681.
129. Camacho-Aguilera, R.E.; Cai, Y.; Patel, N.; Bessette, J.T.; Romagnoli, M.; Kimerling, L.C.;
Michel, J. An electrically pumped germanium laser. Opt. Express 2012, 20, 11316–11320.
130. Wang, X.X.; Li, H.; Camacho-Aguilera, R.E.; Cai, Y.; Kimerling, L.C.; Michel, J.; Liu, J.F.
Infrared absorption of n-type tensile-strained Ge-on-Si. Opt. Lett. 2013, 38,652–654.
131. Tsai, C.-Y.; Tsai, C.-Y.; Chen, C.-H.; Sung, T.L.; Wu, T.-Y.; Shih, F.-P. Theoretical model for
intravalley and intervalley free-Carrier absorption in semiconductor lasers: Beyond the classical
Drude model. IEEE J. Quantun Electron. 1998, 34, 552–559.
132. Haug, A. Free carrier absorption in semiconductor lasers. Semicond. Sci. Technol. 1992, 7, 373–378.
133. Spitzer, W.G.; Whelan, J.M. Infrared absorption and electron effective mass in n-type gallium
arsenide. Phys. Rev. 1959, 114, 59–63.
134. Boucaud, P.; el Kurdi, M.; Ghrib, A.; Prost, M.; de Kersauson, M.; Sauvage, S.; Aniel, F.;
Checoury, X.; Beaudoin, G.; Largeau, L.; et al. Recent advances in germanium emission.
Photonics Res. 2013, 1, 102–109.
135. Tyulerev, V.G.; Obukhov, S.V.; Vast, M.; Sjakste, J. Ab initio calculation of electron-phonon
scattering time in germanium. Phys. Rev. B 2011, 84, 035201.
136. Wang, X.X.; Kimerling, L.C.; Michel, J.; Liu, J.F. Large inherent optical gain from the direct gap
transition of Ge thin films. Appl. Phys. Lett. 2013, 102, 131116.
137. Stroscio, M.A.; Dutta, M. Phonons in Nanostructures; Cambridge University Press: Cambridge,
UK, 2004.
138. Geiger, R.; Frigerio, J.; Süess, M.J.; Chrastina, D.; Isella, G.; Spolenak, R.; Faist, J.; Sigg, H.
Excess carrier lifetimes in Ge layers on Si. Appl. Phys. Lett. 2014, 104, 062106.
139. Langdo, T. Selective SiGe Nanostructures. Ph.D. Thesis, Massachusetts Institute of Technology,
Cambridge, MA, USA, 2001.
140. Camacho-Aguilera, R.E.; Cai, Y.; Kimerling, L.C.; Michel, J. Ge-on-Si bufferless epitaxial
growth for photonic devices. ECS Trans. 2012, 50, 469–473.
© 2014 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article
distributed under the terms and conditions of the Creative Commons Attribution license
(http://creativecommons.org/licenses/by/3.0/).

TMP DEA2

Uploaded by

Copyright:

Available Formats

TMP DEA2

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

TMP DEA2

Uploaded by

Copyright:

Available Formats

Photonics 2014, 1, 162-197; doi:10.

Monolithically Integrated Ge-on-Si Active Photonics

Abstract: Monolithically integrated, active photonic devices on Si are key components in

Figure 1. Schematics of hybrid photonic integration vs. monolithic Si photonic integration.

2. Band Structure and Band Engineering

2.1. Tensile Strain

(a) E (b) E (c) E

Light hole band

bulk Ge tensile strained intrinsic Ge tensile strained n+ Ge

2.2. N-Type Doping

Figure 4. (a) Ge0.85Si0.15/Ge/Ge0.85Si0.15 QW band structure assuming 0.25% tensile strain

3. Ge and GeSn Growth on Si

Figure 5. Schematics of a typical structure for Ge RMG/LPE growth. A cross-sectional

3.2. Growth Methods towards 3D Photonic Integration

4. Ge-on-Si Active Photonic Devices

4.1. Waveguide Integration Schemes

Waveguide-coupling with Ge active photonic devices is indispensable to construct integrated photonic

Figure 8. Schematic structures of typical waveguide-integrated Ge active photonic devices.

4.2. Ge-on-Si PDs

Table 1. Summary of waveguide-coupled Ge-on-Si PD performance. All devices are

4.3. Ge and GeSi EAMs

Figure 11. (a) Schematics of band-to-band optical absorption in a bulk semiconductor

4.4. Ge Gain Media and Lasers

Silicon-Based Lasers supervised by Gernot Pomrenke. From 2010–2012, the research on

The author declares no conflict of interest.

You might also like