Posted by: Qing Xu on March 09, 2017
Fairly new on the data-transmission scene, silicon photonics is an exciting technology that promises inexpensive, mass-produced optical components through photonics integration – it transfers data among computer chips by optical rays that carry more data in less time than electrical conductors. Since the first commercial product was introduced in 2005, the last decade has seen significant progress in technology and development.
A main application area for silicon photonics is cloud data centers, whose footprints continue to grow substantially to accommodate massive amounts of servers and switches. Starting in 2016, many hyperscale cloud content and service providers, such as Facebook and Microsoft, began deploying 100G Ethernet using singlemode optics-based infrastructure in their new data centers to support business growth.
100G Ethernet data center deployment has drawn much attention to silicon photonics; many of these singlemode transceiver modules are made with silicon photonics technology.
Let’s take a closer look at the applications, technology basics and market potential for silicon photonics.
The first proof-of-concept experiment was conducted by Professor Richard Soref in 1985, but it took close to 20 years for the technology to advance from a core idea to a product (Kotura’s variable optical attenuator). The last decade has seen significant progress in product development.
Silicon photonics is also referred to as CMOS (complementary metal-oxide-semiconductor) photonics; these new types of photonic chips are produced with the same well-developed production infrastructure used for CMOS electronic chips. This ensures low costs, high yield and extremely small sizes to unleash a range of new optical devices for different applications:
To better understand why silicon photonics is so appealing in certain industries, let’s take a look at the CMOS integrated circuit (IC) industry:
Silicon photonics is attractive to the semiconductor industry: The investment and the developed expertise (i.e. CMOS processing, wafer-level testing and chip packaging) can be revamped for data transmission.
Historically, the optical components industry is much less mature in terms of component integration and volume shipment; the majority of photonics devices are assembled with discrete components and very high costs. Silicon photonics chip manufacturing can reuse the established CMOS IC process (200 mm and 300 mm wafer) with modifications to add photonics building blocks, such as light generation, modulation, waveguide, detection and passive components like multiplexers and splitters.
Automated wafer-level testing on silicon photonics wafer
The IC chip achieves better performance by shrinking down the feature size. This is not equally true for the silicon photonics process; the minimum size of the waveguides is governed by the wavelength of the light source.
Typical light source wavelengths are within the range of 1310 nm, 1490 nm or 1550 nm; this is several orders of magnitude larger than the size of an electron. As a result, the larger-size features of the old CMOS processing nodes, such as 65 nm, 90 nm or 130 nm, are sufficient to make high-quality optical waveguides. These processes are currently well mastered and can operate at a much lower cost and higher yield as compared to the state-of-the-art 14 nm node CMOS IC process.
Mother Nature does not empower silicon with the best physical and chemical attributes for photonics integration. Historically, optical components are made using group III-V materials (on the periodic table), such as indium phosphide, gallium arsenide, lithium niobate and silica.
Silicon photonics is still in its infancy. As suggested by LightCounting Research, it won’t be a disruptive technology in the next five years; product sales may reach $1 billion by 2020, accounting for 10% of the market.
As a matter of fact, its most important target application and market is data centers; it still competes against indium phosphide-based technology (InP) and VCSEL-based technology.
CMOS silicon photonics chip
Moving forward, on-board optics will help silicon photonics become even more pervasive, especially for high-performance computing platforms, also referred to as supercomputers. It will become an ideal technology that enables ultra-high density and high-speed I/O through the integration of photonics and electronics on a single platform. It is also a suitable technology that can enable large-scale Systems on Chip, which will emerge over the next few years.
If your organization is planning a next-generation Ethernet deployment and has unanswered questions, we can help! Belden data center experts can explain how silicon photonics may impact your future planning and budgeting, and help you decide if it’s worth exploring.