With the recent acquisition of a Silicon Photonics (SiP) and Indium Phosphide (InP) optical company by Ciena, I have been asked whether or not this is a unique Ciena move or part of a larger industry trend towards vertical integration. Personally, I believe that this is part a larger trend of companies pulling critical technology back inside as a way to compete on an field being increasingly leveled by technologies like SDN. However, I end up spending most of my time in the interviews explaining what exactly SiP and InP are before I can get to the vertical integration discussion.

Silicon is an amazing material – so useful that it seems to have been purposefully designed for circuitry. It is abundant, easy to manipulate, simple to create transistors from, and has a native insulating oxide (silicon dioxide). Without going into the details of why each of those properties is important, the result is that silicon is ideal for creating the chips that run almost all electronics. The silicon manufacturing industry is enormous, mature, and flexible – meaning that the cost of going from concept to chip is relatively inexpensive and easily outsourced to one of several specialty companies.

The one useful property that silicon does not have is the ability to efficiently create light. While companies like INSiAVA (a startup from South Africa that I have consulted for) have done work on generating light from silicon, silicon light emission is inefficient and cannot be used to make lasers due to the inherent properties of the material.

Indium phosphide is the base semiconductor material used to create most lasers used in longer distance telecommunications due to its ability to emit light at the low-loss point of optical fiber (1550nm). Other materials such as gallium arsenide are used for shorter reach optics, but InP is ideal at the longer distance. The problem with InP is that it is much more expensive to buy and to work with, the fabrication facilities for InP are much fewer than those for silicon, and InP is less suited to creating transistors, primarily due to less experience creating InP circuitry.

As a comparison of relative costs, when I was in graduate school we regularly etched holes in silicon wafers to use them as holders for more expensive components. Silicon wafers were provided to each student in large boxes. On the other hand, I one broke a small piece of an InP wafer and was concerned that my entire budget for the project was lost. InP wafers were provided as one wafer per group of researchers.

Silicon photonics tries to take advantage of the significantly lower cost and huge background of experience with silicon to move as much functionality as possible out of the InP and into Si. Generally, these processes use waveguides of silicon or silicon dioxide that move light around on a chip while changes in voltages from silicon circuitry produce desired effects on the light. Many of the most advanced coherent devices are built using silicon photonics to modulate the light source. The problem with silicon photonics is that light has to be coupled onto the chip from a light source that is not silicon. That process has high tolerances, relatively low yields, and losses that drive up the cost of the solution.

Enter the alternative of an InP “photonic integrated circuit” or PIC, a term popularized by Infinera and now used industry-wide. PICs are generally agreed to be a complete circuit with all of the optical sources, waveguides, and manipulations performed natively in InP. By combining all of the devices on InP, the light source can be built at the same time as the circuit, on the same device, with direct coupling between the two parts and no need for expensive integration. The downside of a PIC is that the InP process is inherently more specialized and therefore more expensive.

Infinera and others (such as Ciena, via their acquisition) believe that there are places where the additional cost of InP PICs will pay off and give them a competitive advantage in functionality, bandwidth, and scalability. SiP adherents claim that eventually the low cost advantages of silicon will win the day. The good news for customers is that the technology fight between the two approaches should lead to a continuing series of innovative products and increasingly cost-effective solutions for optical networking.