1. Main text

There are 118 known atomic elements in the universe. Only a handful of them, in specific crystalline or non-crystalline arrangements, often in composite materials, can readily constitute a photonic platform with desired characteristics to be harnessed for our applications. Since many of these atomic elements are not compatible to form composite materials, artificial methods to compose hybrid structures have become increasingly popular, especially for new emerging applications in photonics, as photonic platforms. Integration of such hybrid structures or devices can lead to the realization of highly functional systems on a small platform, typically called integrated circuits. Electronic integrated platforms dating back to 1948 for the initial concept [6] and 1959 for the first hybrid integrated circuit realization by Kilby [7] have harnessed silicon, germanium, gallium arsenide, indium phosphide, gallium nitride, and many other semiconductor materials to exhibit remarkable performance together with other metal and insulator materials such as silicon dioxide, silicon nitride, hafnium oxide, copper, aluminum, gold, tungsten, titanium, etc. For photonics, the history of integration is relatively short compared to that of electronic integration. Stewart Miller’s publication in 1969 [8] is often considered the first conceptual proposal. The main photonic integrated circuit (PIC) platforms today include InP, GaAs, silica, and silicon as the base materials.

Not surprisingly, most materials (except for silica) chosen for photonic integration platforms are also common to electronic integration for several reasons. First, photonic integration often preferred to exploit a commercially available platform already established for electronic integration. Since the commercial electronic market is much larger than the photonic counterpart by nearly an order of magnitude, utilizing a common platform for both photonic and electronic ICs has become attractive from manufacturing ecosystem perspectives. Secondly, the photonic components are almost always accompanied by electronic components for control, signaling, buffering, and powering. Thirdly, semiconductors and the associated dielectrics are extremely attractive for photonic and electronic components such as lasers, detectors, waveguides, and transistors. As integration platforms, semiconductor platforms offered immense advantages, especially when complemented by dielectric materials. Initially, the above three reasons drove III-V platforms such as GaAs and InP compound semiconductors as attractive platforms for photonic and electronic integrated circuits; hence the monolithic III-V platforms were actively pursued in Optoelectronic Integrated Circuits (OEIC) in the 1980s [9–11]. Photonic integration desired much more than a simple photonic waveguiding capability in a PIC platform— it desired light generation and amplification, electro-optical modulation, optical-electrical detection, nonlinear optical frequency conversion, and RF-photonic signal processing. The III-V compound semiconductor platform offered practically all of the above functionalities as a PIC platform in addition to being an excellent electronic IC (EIC) platform. For instance, III-V Hetero-Bipolar Transistors’ (HBTs) collector-base junctions can also be used as photodetectors, electro-optical modulators, and waveguides, while emitter-base junctions can be modified to be light emitters or amplifiers. The research and development efforts for OEIC boomed in the late 1980s, but the seemingly promising OEIC expeditions never materialized as a commercially viable large-scale photonic-electronic VLSI platform. The reason, interestingly, is the same reason why III-V transistors (HEMTs, HBTs, FETs, etc.) failed to show competitive advantages against silicon CMOS transistors as enablers of a large-scale, high-yield, and commercially viable EIC platform. III-V materials lacked reliable and high-quality passivation methods. While there have been strong efforts to realize high-quality passivation materials for III-V CMOS devices, they were far inferior to the thermally oxidized silicon dioxide with extremely low defects and high density. Silicon photonics [12] exploiting silicon as the waveguide core and thermally-grown oxide as the under-cladding realized extremely low-loss and low-defect optical interfaces while offering strong confinement of optical modes enabling relatively compact low-loss bending of waveguides. Furthermore, the silicon photonic platform shares the same fabrication process steps in most parts as the silicon CMOS platform, leading to the commercial offering of silicon CMOS photonic electronic-photonic-integrated-circuits (EPIC) (e.g., GF9WG and GF45CLO from GlobalFoundries and SG25 EPIC from iHP). Compared to the III-V semiconductor platform, the silicon platform offers advantages such as the commercial availability of 300 mm wafers, the superior thermo-mechanical characteristics, the abundant material availability, and the well-established end-to-end ecosystem from design and MPW runs to testing and packaging. When considering large-scale integration, silicon photonics offer the superior integration density and yield compared to the III-V platform, thanks again to the high-quality, high-density, and low-defect passivation as well as the low-loss high-contrast optical confinement available in silicon photonics. The ‘photonic Moore’s law’ [13] applied to photonic integration extends its projection even more rapidly than the electronic Moore’s law. However, silicon photonic lacks some key properties desired in photonics: Pockel’s effects, classical optical gain, magneto-optical effects, second-order optical nonlinearities, and nonvolatile phase-change capabilities in silicon’s optically transparent spectral region (1.2–6.9 μm). One obvious solution to such deficiency is to heterogeneously integrate components using other materials on a preferred substrate, a silicon substrate. Such heterogeneous integration is widely used today because it combines various components already fabricated using mature fabrication methods optimized for diverse materials. On the other hand, hybrid photonic integration achieves much closer and higher-density integration by realizing optical waveguides consisting of heterogeneous materials to realize a combination of various optical properties from diverse materials.

The base platform for hybrid photonic integrated circuit platform should enable low-loss, compact, high-density, and high-yield PICs via low-cost manufacturing and packaging processes leveraging the existing ecosystem. From this perspective, silicon photonic platform(s) (including silicon nitride and related waveguide PICs on silicon) is clearly an attractive choice for wavelengths in the 1.2 - 2.6 μm wavelength range (limited by the transparent spectral ranges of silicon and silicon dioxide). At shorter wavelengths, silicon nitride/silicon dioxide PICs on silicon can extend the range to the 0.4-2.6 μm spectral band. Other waveguide platforms such as AlN/GaN on silicon can extend even further to ultraviolet. Various III-V semiconductors can cover the 0.3-20 μm despite the lower yield and other inferior characteristics compared to the Si/Si₃N₄/SiO₂ PIC platform on silicon substrates. To reiterate, silicon photonics with silicon nitride waveguides exploit the well-established CMOS ecosystem to realize electronic-photonic integrated circuit (EPIC) platforms. Then there is a long list of properties that you would like to add or alter in the new hybrid silicon PIC or EPIC platforms. They are (a) optical gain, (b) electro-optical modulation, (c) electro-absorption modulation, (d) photodetection, (e) non-volatile optical index variation, (f) optical non-reciprocity, (g) optical nonlinearity, (h) thermo-optical, (i) piezo-optical, and (j) quantum mechanical properties. For optical gain, hybrid III-V/Si waveguide platforms (AlInGaAs quantum well/Si [14], [15], InAs Quantum Dot/Si [16,17], AlInGaAs/Si quantum cascade lasers [18,19]) and hybrid erbium-doped Al₂O₃/Si₃N₄ waveguide [20] platforms have been successfully demonstrated. Depending on the designs and hybrid integration methods, the achievable gain magnitudes are lower than the gain achievable in the native material platforms, obviously due to the reduced confinement factor of the gain region in the hybrid gain waveguide. However, the hybrid integration method brings a new life to the silicon platform with embedded optical gain. For electro-optical Pockels effect, which does not exist in silicon, hybrid III-V/Si waveguide platforms (AlInGaAs quantum well/Si [21], InAs Quantum Dot/Si [22], and InGaAsP/Al₂O₃/Si MOS capacitor waveguide), hybrid Thin Film Lithium Niobate (TFLN) on insulator waveguide platforms ([23,24]), and more recent 2D material/Si hybrid waveguide modulators (Graphene on silicon modulators [25], WS2 on silicon modulators [26]) have been investigated. For electro-absorption modulation, Ge/Si hybrid waveguide modulators (quantum-confined Stark effects in Ge/Si multi-quantum-wells [27,28] and Frantz Keldysh effects in GeSi waveguides [29]) are among the popular ones. Many of the electro-optical modulators mentioned above, such as graphene/Si and III-V/Si hybrid platforms, exhibit both electro-optical and electro-absorption characteristics, governed by the well-known Kramers-Kronig relations. While these devices are typically used for high-speed data modulation, devices capable of changing and holding the optical refractive index’s real and/or imaginary parts of the optical refractive index can be very useful for photonic memory or photonic neural networks. Such photonic phase change materials in hybrid integrated platforms have recently emerged. Typically phase change materials such as Ge₂Sb₂Te₅(GST), VO₅, and Sb₂Se₃ have been deposited on silicon or silicon nitride waveguides to form hybrid waveguides with nonvolatile optical phase or amplitude variations [30–4]. For photodetection, germanium that can be epitaxially grown on silicon is most popular as hybrid integrated photodetectors for silicon PICs. However, depending on the designs, photodetection in III-V or II-VI materials can offer higher responsivity, higher speed, and lower noise than group IV materials due to their direct bandgap absorption and higher mobility. The successful hybrid InGaAs/Si photodetection with hetero-interfaces exploited photodetection in InGaAs and low-noise avalanche-multiplication in silicon [35]. Uni-traveling carrier (UTC) photodiodes [36] [37] can exploit high mobility of electrons to achieve high-power and high-speed photodetection based on hybrid Group-IV (e.g Ge/Si) as well as hybrid III-V/Si (e.g. InGaAs/InP/Si and InGaSbAs/Si [38]) structures as well as hybrid III-V/Si₃N₄ waveguide detectors [39] in both surface normal and waveguide geometries. Photodetection in hybrid quantum dot structures can provide flexible optimization of spectral characteristics. A review article on quantum dot infrared photodetectors [40] provides an excellent summary of technologies, including hybrid QD detectors on silicon and silicon nitride. The two-dimensional materials similar to those seen in hybrid electro-absorption modulators can also be used as photodetectors. Various hybrid graphene-silicon photodetectors have been demonstrated [41–45] with flexible wavelength coverage beyond the 1.55 μm [3] wavelength range. Successful demonstrations of optical non-reciprocity in Ce-YIG/Si optical Isolators [5], second harmonic generation in GaN/AlN on silicon [46], and piezo-optomechanical LiNbO3 on SOI [47] have also been achieved. The hybrid integrated photonic platforms mentioned above have temperature-dependent optical characteristics such that athermal operation of the resulting photonic integrated circuits becomes difficult. An athermal hybrid integrated photonic platform can be made possible by introducing materials with negative thermo-optical-coefficient (TOC) to compensate for the positive TOC of the other materials. For instance, polymers and TiO₂ have negative TOC with a comparable magnitude as TOC of silicon, and athermal waveguide ring modulators have been successfully demonstrated in both hybrid polymer-silicon waveguide [48] and hybrid TiO₂-silicon waveguide [1,49] platforms.

Realizing the above hybrid photonic integration requires fabrication techniques that can bring multiple dissimilar materials intimately together on the same platform. Hetero-epitaxy has been successful in some cases, such as in hybrid Ge-Si platforms where epitaxial growth and thermal annealing are repeated to overcome the lattice mismatch and to reduce defects [50]. In addition, hetero-epitaxy of GaP/Si has led to the realization of hybrid nonlinear optical frequency converters, and hetero-epitaxy of Ge/Si followed by heteroepitaxy of GaAs/Ge allowed realization of hetero-epitaxially grown InAs/GaAs quantum dot lasers on silicon. While hetero-epitaxy techniques offer robust and wafer-scale hybrid integration methods, only a small number of material combinations can achieve hybrid integration by heteroepitaxy due to the material compatibility of thermal, chemical, and mechanical characteristics. Hybrid integration by hydrophobic wafer-bonding [51,52], hydrophilic wafer-bonding [14], adhesive-bonding [53], transfer-printing [54], and non-epitaxial deposition methods including chemical-vapor-depositions (CVD), sputtering, and thermal evaporations offer much more flexibility in choices of the material combinations overcoming the different characteristics between the dissimilar materials. The hybrid integrated platforms based on these methods, such as pluggable silicon-photonic transceivers with hybrid-integrated InP/Si lasers [55], are already in commercial markets today, and we expect to see many more successful commercial transitions in the near future. This author's OPINION is that a desirable hybrid integrated photonic platform consists of (a) a base platform consisting of silicon photonic waveguides, silicon nitride waveguides, and monolithically integrated CMOS electronics, together with (b) hybrid-integrated photonic (and electronic) components realizing optical gain blocks such as hybrid InAs QD/Si gain elements, optical modulators such as hybrid InGaAsP/Al₂O₃/Si MOSCAP ring resonators, photoreceivers such as resonant hybrid InAlGaAs QW/Si APDs with integrated CMOS trans-impedance amplifiers, and nonvolatile photonic components such as hybrid Sb₂Se₃/Si Mach-Zehnder switches. As Fig. 1 illustrates, the hybrid integrated photonic platform based on silicon CMOS-photonics acquires additional diverse functionalities that silicon does not possess through the post-processing of heterogeneous materials on silicon. The silicon CMOS ecosystem has recently been updated to include silicon photonics for EPDA (electronic-photonic-design-automation), MPW, and TAPs (test and packaging). The hybrid-integrated silicon CMOS-photonic platform can also benefit from this ecosystem with additional functionality to facilitate rapid adoption and commercial deployment.

 

Fig. 1. Hybrid integrated photonic and electronic platform on silicon example including (a-b) hybrid TiO2/SOI athermal modulators and switches [1], (c) III-V/Si optical gain blocks [2], (d-e) graphene/Si plasmonic photodetectors [3], (f) nonvolatile optical switch [4], (g) Ce-YIG/Si optical Isolators [5], and (h) Si₃N₄/LiNbO₃/SiO₂ modulators, as well as CMOS application-specific integrated circuits (ASICs). The center figure shows an example of hybrid integrated photonic-electronic integrated circuits leveraging the CMOS ecosystem including electronic packaging.

Download Full Size | PDF