Introduction
Scalable integration of photonic devices has long been a revolution on the horizon. The ability to mass-produce arrays of optical components integrated onto a single wafer promises dramatic improvements in the cost, performance and complexity of light-based technology. In contrast to the current labour-intensive hand assembly of optical devices, integration could put photonics on the kind of Moore's law trajectory that has made microelectronics pervasive.
Researchers have just overcome one of the remaining obstacles to this enticing goal on page 57 of this issue1. Barwicz and colleagues have demonstrated the ability to make devices that confine light on a microscopic scale and, in addition, are 'polarization-transparent': that is, an incoming light signal is processed correctly even if it has a randomly oriented electric field. Specifically, Barwicz and co-workers present an add–drop filter, a device that can pull out a signal at one wavelength from a communication line densely packed with many different wavelengths. The device can separate closely spaced (130 GHz) channels very cleanly and maintains polarization-transparency over a broad range (60 nm) of wavelengths — the kind of device you would expect to find on a next-generation telecommunications chip.
This breakthrough has come at an exciting time in the field of microphotonics. In just a few years, several major hurdles have been overcome. One barrier was the profound incompatibility between materials that are good at generating or modulating light and those that make good electrical devices. Scientists have now coaxed different materials to work together, for example by bonding optical-gain media onto silicon wafers used in standard electronics2. In other work, attempts to endow silicon — the staple of the microelectronic world — with the ability to control light are paying off, resulting in compact silicon modulators3, 4.
Even simpler structures that merely move light around a microphotonic chip can unfortunately be difficult to implement, because of their sensitivity to manufacturing errors. For dense integration into optical devices, waveguides must confine light within submicrometre dimensions using materials with a high refractive index. But even slight imperfections in the dimensions or surface roughness of waveguide structures can lead to crippling light losses.
If the light entering such a structure is randomly polarized, this inherent sensitivity becomes an even bigger problem. Arbitrary input polarization is often unavoidable, especially if light couples to the device through an optical fibre as is the case in many applications. A recent theoretical analysis5 showed that even if components could be fabricated with atomic-level precision, this would still not be enough to achieve polarization transparency for some devices. Subnanometre imperfections in modules such as add–drop filters can lead to different polarization contributions being filtered in different ways. The need for better than atomic precision puts severe limitations on the structures that can be made.
This apparent impasse has not stopped Barwicz and colleagues1. They use a technique called polarization diversity, in which light (of a random polarization) is split into its orthogonally polarized components travelling in separate arms of a photonic circuit (see Fig. 1). By rotating the polarization state in one of these arms, a single polarization is achieved on the chip as a whole. The two beams then pass through identical sets of polarization-sensitive structures and are recombined at the output. In this way, the researchers force polarization-sensitive structures (in this case distinct add–drop filters) to behave in a polarization-transparent way, offering an almost identical response to the different input polarizations.
Figure 1: Polarization-transparent devices.
Using the polarization-diversity approach shown, the two polarization components of the incoming light wave are separated, sent through two identical photonic structures (add–drop filters in this case), and combined at the output. The overall device has an almost identical filter response to orthogonal input polarizations, even though individual photonic elements are extremely polarization sensitive.
Full size image (33 KB) (33 KB)Such an approach is conceptually rather simple: as each filter receives only one polarization, its mismatched response to the other polarization does not matter. However, to implement this known approach, the team had to overcome some technological challenges. Not least, polarization diversity calls for extra complexity and the design of new elements in addition to the microrings that do the filtering. One such component is an elegant polarization splitter and rotator, which manipulates light in an intricate three-dimensional structure that can be constructed from just two flat silicon layers. Moreover, the filters in the device must be very nearly identical so that they offer the same response, otherwise the overall device will process the two incoming polarizations differently.
This advance is a culmination of several refinements that have been made in recent years, pulling together clever designs and fabrication techniques. By integrating these new building blocks, Barwicz and co-workers not only address the specific needs of polarization diversity, but they also move a step closer to producing sophisticated optical devices with improved functionality.
The emerging silicon photonics toolbox is impressive, and its ability to put optics on a chip with standard silicon electronics already has enormous potential. Nevertheless, tantalizing questions remain about the development of light sources that are compatible with silicon. In the meantime, non-silicon platforms are showing promise for optical integration6. Regardless of the eventual winner, in order to substantially change the way optical networks are made — and get other applications such as light-based computing and sensors moving — we will need to bring together the capabilities demonstrated at MIT and other research labs. With exciting improvements in the pipeline, the long-awaited impact of integrated photonics should be arriving soon.