The fabrication process, known as laser reactive deposition (LRD), exploits a high-power industrial laser to create nanoparticles of optical materials that are then deposited on a silicon substrate. LRD is both faster and cheaper than other common deposition processes, and also allows the composition of the layer to be varied across the substrate to create more advanced optical devices.
"Planar technology is a very good way to make devices smaller and more cost-effective, and to integrate different components on the same chip," says Jane Li, vice-president of sales and business development at NeoPhotonics. "But most conventional deposition processes have limited capability when it comes to complex compositions that enable new optical functionalities."
Most manufacturers of planar waveguides have borrowed production processes from other industries, notably chemical vapour deposition (CVD) from the semiconductor business and flame hydrolysis deposition from fibre manufacture. But Li argues that neither of these techniques has been optimized for making optical devices. "In CVD the deposition rate is very slow, which is fine for small chip sizes and thin layers. But photonic chips are much larger than semiconductor devices and the layers are so thick that it's not very cost-effective."
Depending on the application, Li estimates that CVD costs about $700 per wafer and can take several hours to deposit the optical layer. In contrast, LRD takes just a few minutes and costs less than $200 per wafer. It also produces more uniform layer properties, which leads to higher yields.
The LRD process works by ejecting vapours of precursor materials from a nozzle into a high-power laser beam. The laser drives chemical reactions between the precursors, creating solid nanoparticles that stick to a substrate moved above the reaction zone. The process exploits a continuous carbon-dioxide laser to initiate and sustain the reactions, and this also controls the temperature and chemistry within the reaction zone.
Li explains that it is crucial to control the distance between the reaction zone and the substrate precisely to ensure that the nanoparticles are deposited at the right time. The company's current process operates with 100 mm wafers, but Li claims that the design can be easily scaled to 150 mm wafers and then to 200 mm.
The process can deal with all sorts of precursors in vapour, gas or aerosol form, allowing new optical materials to be synthesized quickly and easily. And the ability to vary the composition of the layer across the silicon wafer - and with it critical material properties such as phase, surface morphology and particle size - makes it possible to optimize the optical properties of the device.
"Planar amplifiers is one functionality we are pursuing, since we can increase the concentration of erbium without much clustering," says Li. The company also plans to integrate a planar amplifier with an arrayed waveguide grating, which would deliver variable optical gain after demultiplexing signals in dense wavelength-division multiplexing (DWDM) systems.
Existing DWDM technology combines a demultiplexer with variable optical attenuators, but a planar amplifier would raise the power of the weaker signals to match the strongest rather than dropping the level of the strongest signal to match the weakest. Other devices in the pipeline include integrated optical add-drop modules that would also provide loss compensation.