Infinera Technology

Integration of active and passive optical components on a single substrate has been the "holy grail" of the optical R&D community for over 30 years. There are myriad reasons why it has taken so long for the practical realization of highly integrated optical devices. These include technological barriers associated with achieving the requisite process uniformity, reproducibility and manufacturing scale for InP semiconductor processes such as epitaxy, lithography, and etching. There were also design difficulties associated with the design of active/active and active/passive transitions and electrical/optical isolation.

In addition, the development of EDFA technology in the early 1990's enabled the elimination of many digital OEO-based repeaters, which for a while dampened the drive to reduce OEO cost. However the biggest issue limiting PIC development appears to have been the single-minded industry focus on maximizing optical component performance in response to the system-level drive to implement "all-optical" networking technologies. This drove extreme device optimization, which in turn required the development of discrete, single function devices built on different materials platforms so as to maximize performance for each function.

Until recently, the availability and use of photonic integrated circuits has been quite modest. One example would be single-wavelength DFB lasers integrated with a 10Gb/s electro-absorption modulator and PIN power monitor photodiode, also known as an Electro-absorption Modulator Laser, or EML. This device integrates three functions in InP along a single serial direction (ie: operation across a single wavelength). However, the modest degree of functional integration in an EML compared to discrete devices limits the extent of the system-level benefits that can be achieved.

Figure 6: Infinera has introduced the first "WDM system on a chip" by monolithically integrating all the optical components required for a 100Gb/s DWDM system onto an InP PIC.

Infinera introduced into live network operation the first large-scale PICs with a capacity of 100Gb/s, representing the industry's first "WDM system on a chip"(see Figure 6). Consolidating over sixty optical devices and six different functions into a one pair of PICs, these PICs offer the same capacity and performance at only a fraction of the size and cost that would otherwise be incurred using conventional discrete components.

Designed to work as a pair, the transmit and receive PICs developed by Infinera incorporate all of the active and passive optical components required to implement a 100Gb/s DWDM system operating with ten wavelengths at 10Gb/s per wavelength (see Figure 7). These PICs include both active components such as wavelength-specific lasers, modulators and detectors, but also passive components such as multiplexers and attenuators. All these components are monolithically integrated into an InP substrate, and interconnected by "on-chip" planar optical waveguides that eliminate the need for complex and unreliable fiber couplings between each device.

Figure 7: Infinera's InP photonic integrated circuits include a 100Gb/s transmit PIC and 100Gb/s receive PIC that together implement a 100Gb/s DWDM system.

Why has Infinera been able to implement such a significant step forward in an already well researched field?

First, because Infinera undertook a holistic approach to component and system design, it allowed component development to remain unencumbered by prior industry notions of what the "right" device specifications needed to be and enabled component and system specifications to be defined not in isolation, but rather in such a way as to ensure they were mutually optimized to maximize network benefit.

Second, by attacking the challenge of photonic integration from scratch, Infinera did not need to evolve existing InP manufacturing lines, but instead implemented best in class practices for both InP and Silicon processing. Combined with a focus on manufacturing repeatability and process control, this allowed Infinera to address and overcome many of the practical problems previously encountered in the volume manufacture of PICs.

Large-scale photonic integration has since shown promise for enabling even greater functional integration. For example, Infinera R&D efforts have demonstrated a 400G PIC comprised of hundreds of devices to implement complex modulation functionality, as well as PICs capable of operating outside the conventional band of DWDM. In parallel other R&D activities have shown the ability to increase PIC channel counts, such as integrating twenty or more DWDM channels into a single device, or increasing the degree of functional integration by integrating additional optical functions such as semiconductor optical amplifiers into a PIC.

Taken together, these advances will enable ever greater consolidation of system functionality and cost compared to discrete optical devices, and for the first time brings the learning curves of Moore's Law to optical networking. This unprecedented level of optical component integration and packaging consolidation enabled by monolithic integration will allow future optical systems to benefit from greater functional integration, increased density, and higher reliability.