Research Progress And Challenges Of Silicon-based Photonic Chips
Apr 22, 2024
Leave a message
Silicon-based photonic chips use photons as the information transmission medium, which has the advantages of high bandwidth, high speed, high integration, and compatibility with CMOS process, and has application value in many fields. A complete silicon-based photonic chip integrates a light source, optical waveguide, modulator, filter, detector and other devices, which can realize the generation, routing, modulation, processing and detection of light, and these functions together form an optical loop similar to an electronic integrated circuit, so as to realize the transmission, control and processing of information.
According to MEMS Consulting, the research team of Professor Yang Jianyi and Associate Researcher Wang Yuehai of the School of Information and Electronic Engineering of Zhejiang University introduced the different material platforms of silicon-based photonic chips, and reviewed their research progress and challenges in the fields of optical communication and optical interconnection, optical computing, biosensing, on-chip lidar and optical quantum, and finally summarized. The related research content was published in the journal "Semiconductor Optoelectronics" with the title of "Research Progress and Challenges of Silicon-based Photonic Chips".
A manufacturing platform for silicon-based photonic chips
Figure 1 shows the market value forecast of silicon-based optoelectronics in different application scenarios by market research institute Yole, and the market size of the entire silicon-based optoelectronics industry is estimated to reach $1.1 billion in 2026. There are a variety of manufacturing platforms for silicon-based photonic chips, and the following is a brief introduction to four commonly used manufacturing platforms: silicon on insulators (SOI), SiN, III.-V group (GaAs and InP), and lithium niobate thin films on silicon substrates.
Figure 1 Market Value Forecast of Silicon-based PhotonicsIn Different Application Scenarios
SOI Platform
SOI is composed only of silicon and silicon dioxide and is the basic material platform for silicon-based photonic chips. The transparent window of silicon material is 1270~1650 nm, so it is almost transparent to the light in the optical fiber communication band. To form an optical loop, a variety of passive and active components need to be integrated on the SOI platform.
Passive components do not require external electrical modulation, such as waveguides, microring resonators (MRRs), Mach-Zehnder interferometers (MZIs), gratings, etc., where waveguides are the basic devices. The core and cladding of the SOI rectangular waveguide are silicon and silicon dioxide, respectively, and the refractive indices of silicon and silica at 1550 nm wavelength are about 3.5 and 1.45, respectively. Active components include lasers, modulators, detectors, and more. In terms of light source, silicon is an indirect bandgap material with low luminous efficiency and is not suitable for light source, so it needs to be combined with other materials, such as rare earth doped light source, III.-V group light source, IV group light source, etc. SOI modulators generally use thermal modulation or carrier dispersion modulation. Figure 2 shows a sandwich-type germanium detector with a fin-transistor (FinFET)-like structure released by IHP in Germany in 2021, which reduces the intrinsic region width of germanium and the carrier overtime to achieve a bandwidth of up to 3 dB at 265 GHz at 1550 nm, surpassing all previous silicon-based integrated detectors, with a responsivity of 0.3 A/W and an operating dark current of 100~200 nA.
Fig.2. Cross-section of a Germanium detector with a FinFET structure
In addition, there are also abundant nonlinear effects in silicon, such as four-wave mixing, Kerr effect, carrier dispersion effect, etc., which can be used in optical frequency comb, quantum optics and other fields. However, when the pump power is high, two-photon absorption and free-carrier absorption effects occur in the silicon, resulting in additional nonlinear losses.
SiN Platform
SiN is used to isolate individual transistors in the CMOS process of traditional microelectronic chips, and is also used as a gate material for some kind of field-effect transistor, which can be used as a complementary platform for SOI. SiN's large transparent window and low transmission losses (<1 dB/m) from the visible band of 400 nm to the near-infrared band of 2350 nm enable MRRs with a figure of merit (Q) of up to one million. The core layer and cladding layer of SiN waveguides are SiN and SiO₂, respectively, and SiN materials also have good nonlinear effects, which are widely used in on-chip nonlinear studies. The recently developed SiN-on-SOI platform combines the advantages of both SiN and SOI platforms, and has application prospects in the fields of nonlinear optics, filters, low-loss waveguides, and integrated optical gyroscopes.
III.-V. Family Platforms
III.-V group materials were the main fabrication platforms for early optical communication chips. III.-V. compounds, especially gallium arsenide (GaAs), indium phosphide (InP), etc., are naturally direct bandgap materials, and their valence band top and conduction band bottom are in the same position in the wave vector k space, and the recombination of electrons and holes does not need to exchange momentum, has a high internal quantum efficiency, can emit light efficiently, and can be used as a gain material for laser sources. Heterogeneous integration of hybrid laser sources composed of III.-V. group materials on Si substrate is one of the ways to achieve light sources on silicon substrates, but heterogeneous integration increases the complexity of the manufacturing process. Figure 3 shows a 1550 nm electrically pumped quantum well laser directly epitaxial grown on a silicon substrate with a maximum continuous output power of 18 mW at room temperature in 2019 developed by Jonathan Klamkin's group at the University of California, Santa Barbara. Figure 4 shows a quantum dot tunable laser directly grown on silicon substrate realized by John E. Bowers' group at the University of California, Santa Barbara, in the same year, with an edge mode rejection ratio greater than 45 dB, a wavelength tunable range of 16 nm at room temperature, and an output power greater than 2.7 mW.
Fig.3 Schematic Diagram of a Quantum Well laser
Figure 4 Schematic Diagram of a Quantum Dot Tunable Laser
Applications and Challenges of Silicon-based Photonic Chips
Optical Communication and Optical Interconnection
At present, the main application scenario of silicon-based photonic chips is still optical communication. Silicon-based photonic chips have the advantages of high integration, good stability, low power consumption and good phase modulation characteristics, which are not only suitable for long-distance data transmission, but also very suitable for the requirements of short-distance and large-capacity data transmission within or between chips, and are ideal optical communication and optical interconnection platforms. Through the monolithic integration of microelectronic circuits, silicon-based photonic chips can achieve high-speed, high-bandwidth, low-power, low-latency on-chip interconnection, while reducing the number of devices on the chip, increasing the interconnection density, and breaking through the limitations of current microelectronic chips in data interconnection.
Silicon-based photoelectric transceiver chips have been widely researched and applied, and are of great significance in large-capacity data communication, and many advances have been made in recent years. Optical communications also require large-scale optical switching arrays. The number of communication connections in data centers is doubling every 2.5 years, leading to a dramatic increase in the need for interconnects, which can be met by optical co-packaging technology. In addition, photons have multiple dimensions of wavelength, polarization, mode, time and other resources can be used, multi-dimensional multiplexing technology has been widely used in optical fiber communication, combined with multi-dimensional photonic integrated chips are expected to solve the new capacity crisis of optical communication. In 2020, Yikai Su's research group at Shanghai Jiao Tong University proposed an on-chip multi-dimensional multiplexing/demultiplexing scheme for wavelength-mode-polarization signals using a backcoupler structure based on cascaded subwavelength gratings (SWGs), as shown in Figure 5.
Fig.5. Schematic Diagram of 8-channel Multi-dimensional Multiplexing Structure
Silicon-based optical communication and optical interconnection chips currently face three major challenges: (1) Silicon substrate on-chip light source. On-chip light sources can improve the integration and energy efficiency of optical interconnect networks, and the light source problem is a major challenge for the entire silicon-based optoelectronic technology. At present, the relatively mature silicon substrate on-chip light source is based on III.-V group materials, and III.-V group lasers are realized on silicon-based photonic chips through hybrid integration or heterogeneous integration. (2) The modulation bandwidth is limited due to the carrier dispersion effect. The use of new modulation mechanisms, such as lithium niobate, is expected to solve this problem. (3) Large-scale integration and reliable packaging. Co-packaging technology can now be used to improve large-scale integration capabilities.
Light Computing
In recent years, technologies such as artificial intelligence, neural networks, speech processing, and image recognition have developed rapidly, and large-capacity real-time data processing and analysis scenarios have generated a fierce demand for computing power. The current data processing relies on traditional microelectronic chips, although this chip processing and manufacturing technology is mature, due to structural defects, its bandwidth is small, the speed is slow, and the power consumption is large. Optical neural networks and high-performance computing based on silicon-based photonic chips are expected to solve this problem.
Artificial neural networks can be used as data processing tools for artificial intelligence, and their calculations are focused on a large number of matrix operations, while optical neural networks consume almost no energy in these operations. Optical neural networks mainly include feedforward neural networks (FNNs), recurrent neural networks (RNNs), and spiking neural networks (SNNs), which can be implemented using MZI or MRR.
Although silicon-based photonic chips have great application potential in computing and neural networks, although they are superior to traditional microelectronic chips in terms of computing speed and power consumption, they still have shortcomings in the all-optical implementation of neural network nonlinear activation functions, integration, and matching algorithms of photonic chips.
Biosensing
Biosensor is a device that can convert information about molecular structures such as proteins and nucleic acids into sound, light, electricity and other signals, and is widely used in biodiagnosis, drug discovery, life science and other research fields. Silicon-based optical biosensors use the interaction between biomolecules and the light field to change the phase, intensity, wavelength and other parameters of light, and use photoelectric conversion to convert the optical signal into an electrical signal, so as to obtain the structural information of biomolecules, which has the advantages of high sensitivity, strong anti-electromagnetic interference ability, convenient multi-functional integration and strong flexibility. According to the different sensing principles, silicon-based biosensors can be divided into biosensing based on refractive index changes, biosensing based on fluorescence technology, and biosensing based on Raman scattering. Biosensing based on refractive index change uses the evanescent wave of the waveguide, and the interaction between the evanescent wave and the solution to be measured will change the refractive index of the cladding, and then change the effective refractive index and phase of the light wave in the waveguide. MZI and MRR are used as examples to introduce the principle of this biosensing. MZI is an interferometric biosensor, one arm interacts with the solution when working, and the phase difference of the two arms changes with the solution to be measured, and the information of the substance to be measured can be obtained by detecting the transmission spectrum formed by the interference of the two arms at the output end.
Lidar
LiDAR is a technology that detects the orientation and speed of a target by emitting laser beams, and has important applications in autonomous driving, three-dimensional imaging and other fields. Traditional lidar adopts the method of mechanical steering, which has the disadvantages of complex structure, easy wear, and easy to be affected by temperature and vibration. Silicon-based LiDAR uses solid-state optical phased array (OPA) to overcome these problems. OPA is an important part of lidar on silicon substrates, responsible for the generation and emission of detection signals, and consists of four parts: laser source, beam splitter, phase shifter and emitter, which can be used in free space communication, detection, imaging, biosensing and other fields. The metrics of OPA mainly include field of view (FOV), beamwidth, sidelobe rejection, modulation speed, power consumption, etc. where the FOV determines the range of beamforming and steering, and the beamwidth measures the size of the OPA transmission or reception point.
.
Figure 6 Diagram of 1D OPA and 2D OPA
Optical Quantum
Optical quantum chips use wave guides to guide photons, providing phase-stable quantum circuits with core functions, including quantum state generation, manipulation, and single-photon detection. Compared with traditional desktop optics, optical quantum chips use a passive, low-loss, multi-dimensional and easy-to-control device library of silicon-based photonic platform, which is convenient for large-scale integration, and is expected to carry out large-scale quantum information processing with hundreds or thousands of photons, which can rapidly promote the practical application of optical quantum technology. In recent years, optical quantum chips have developed rapidly, which is expected to promote the development of quantum computing, quantum communication, quantum sensing, quantum simulation, and basic science. Quantum light sources have important applications in quantum communication, quantum computing and other fields, and are divided into single-photon sources, entangled state sources and continuous variable light sources. On-chip single-photon sources can be implemented using silicon waveguides or spontaneous four-wave mixing (SFWM) in MRR. SFWM is a third-order nonlinear effect that converts two pump photons into a pair of frequency-entangled signal photons and idle frequency photons, and the photon pairs can be used to predict a single-photon source after unentanglement.
Figure 7 The Working Principle and Specific Structure of the Xanadu Optical Quantum Computing Chip
The development of silicon-based optical quantum chips also faces many problems and challenges: (1) the need for fast and low-loss optical switching networks. Processing a large number of single-photons in the short term requires multiplexing and demultiplexing of single-photon sources, and recently LN, Si-LN and Si-barium titanate switches have shown promising applications in this regard. (2) An optical quantum chip that fully integrates a quantum light source, a loop, and a detector is yet to be realized, and the challenge lies in the removal of pump light and the low-temperature manipulation of photons. Cascaded MRRs and MZI with rejection ratios of up to 100 dB have been reported, which are expected to address the first challenge, which is expected to be solved by a low-temperature operating Si-barium titanate switch. (3) How to further improve the performance of MBQC, and how to overcome the error and variability during large-scale manufacturing. Combined with the high programmability of optical quantum chips and machine learning algorithms, it is expected to compensate for manufacturing imperfections. The development of optical quantum chips is closely related to silicon-based optoelectronic technology, and the key performance of optical quantum circuits needs to be driven by new materials, advanced integration and packaging processes. To meet the challenges of integrated optical quantum and market demand, there is a need for a coordinated approach, investment in the development of new photonic integration platform components and supply chains, and the establishment of hybrid and heterogeneous integration infrastructure.
Send Inquiry