Fabrication and assembly of an AR/VR Light-Engine prototype in project MatEl

High-power RGB diodes were edge-coupled to our TriPleX® silicon nitride PIC. Each of these laser diodes can output at least 120 mW. As such, they can also get pretty hot. So, we placed the diodes on aluminum nitride (AlN) thermal sink submounts. The coupling loss across the interface was around 2 dB.  Fans of our hybrid integration will know that this is a disappointing result for us (under 1 dB/facet is what we are used to). It highlights the importance of close collaboration for us—due to missing info about the diodes, we were unable to optimally integrate them to our PIC. Quality control across the supply chain is crucial for scalability. It is acceptable in early prototypes to have some leeway as things are getting tested out, but without reliability, there is no scalability.

The TriPleX® chip had a circuit of combiners to mix the RGB colors, splitters that branched out to 16 channels, and directional couplers to adjust the color in each channel. Next, a variable optic attenuator array (made using Mach-Zehnder Interferometers), allowing us to adjust the output light power per channel. Finally, to outcouple each channel out-of-plane, 16 waveguide-level mirrors were fabricated with focused ion beam (FIB).

As usual, all the processing and assembly were done in-house at LioniX.

I should highlight here what really contributes to production scale-up. As with any early prototype, you never want to spend a lot of time trying to make many untested technologies demonstrate a new device. The first iteration should mostly use known quantities: reliable processes, well-tested designs, and familiar components. Edge coupling is relatively slow for hybrid integration, but it is a high-performance, well-known technique for us. Especially when using high-power diodes, it is better to rely on a low-loss integration technique so we can test the diode and circuit performance in the final demonstrator itself. The two “dependent variables”, so to speak, are the and the FIB mirrors.

The AlN submounts were a real relief. Because visible laser diodes (the bigger the better) generate so much heat, we had to keep the high-power diodes cool to keep them going at full strength. AlN is a great heat conductor. It is also nice to work with and super easy to adapt to any shape we want, which made it a perfect interposer between our standard copper submounts and the new diodes. Those will definitely speed up module production when volume is needed. The same goes for the FIB mirrors. After some tinkering with the process design, we are now able to easily add these to our TriPleX® silicon nitride waveguides. Having a scalable out-of-plane coupling option is very useful for a lot of applications that need a projection solution. An array of 16 mirrors was chosen for this design, but in fact, we could put hundreds of these in the PIC. I cannot imagine what you would make with that ability, but we are always surprised by the creativity of our clients!

So, the RGB diodes push light into the silicon nitride, the light is combined and split up, each split channel is independently tuned for wavelength and intensity, then the light is pushed out broadside. How could this be done better?

Well, for one, all tuning on this module is done with heaters. Thermo-optic actuation is fine, but stress-optic actuation is finer. For a power-sensitive application like AR/VR, piezoelectric actuators are a much better fit, since they consume negligible power in steady-state operation. Another easy upgrade is to flip-chip the diodes on the TriPleX®, instead of edge-coupling them. That speeds up the assembly considerably and enables parallelization, great for volume production. There are a couple more tricks we are working on that we cannot talk about just yet, but there is an obvious problem here that PIC designers will enjoy. How do you split light in a way that reuses power in unused channels?

We have already been working on the next demonstrator module, and we have some pretty good answers and some pretty interesting questions. We can already say that Mark II will be smaller, produce less heat, be more power efficient, and get considerably lighter. All great things for the stars, whether they are satellites or the ones who will wear cool AR glasses all day.