Next generation RGB edge-emitting lasers for Augmented Reality

See-through AR smartglasses can be split into two main categories. One combines a waveguide structure, employing total internal reflection to couple light into a lens, with an extraction grating that delivers images to the user’s eye. The common alternative involves free-space projection and reflection off the lens. As we’re about to see, both have pros and cons (see Figure 1).

One of the merits of the waveguide approach is that it can be designed to deliver a good level of visual comfort, due to large eye boxes – thanks to them, when the pupil moves around, there is no difficulty in seeing the image. However, the use of waveguides creates a number of problems related to image quality and display efficiency. With this architecture, there are issues that result in a less than satisfactory AR experience, arising from a combination of chromatic dispersion and artefacts related to the waveguide structure and extraction gratings.

Another downside of waveguides is that they can produce extremely lossy coupling of light to the eye, with efficiencies of 1 percent or less not uncommon. Due to the large optical loss of the system, light sources have to compensate by providing an optical power per colour ranging from several tens of mW to 100 mW. Such a high output drives up power consumption and leads to a larger projector size, needed to handle heat dissipation. This state of affairs is hampering the fulfilment of the consumer’s wish for the development of slim glasses weighing no more than 70 grams.

The strengths of free-space reflective displays are a simpler optical design, the generation of higher quality images, and exceptional efficiency in relaying light from the projector to the user’s retina; thanks to this, they require a far lower power from the light sources. Despite a smaller eye box, the simplified design and the strongly reduced power consumption make this option an excellent candidate for consumer-type AR-glasses. In this case, the required optical power from each laser diode is only a fraction of a milliwatt, rising to a few milliwatts when accounting for the optical loss of the projector unit and the reflective element.

Is it possible to fulfil these requirements with red, green and blue laser diodes? Well, there’s hope, thanks to a major evolution over the last two decades in blue and, in particular, green laser diodes based on GaN and its related alloys. During that time there have been fundamental technological breakthroughs in efficient p-type doping and the realisation of high-quality active regions emitting in the green. In addition, much effort has been directed at developing reliable devices that deliver ever-higher output powers for industrial and large-area display applications. However, low-power regimes with reduced power consumption remain largely unexplored. This explains why commercially available laser diodes are yet to meet the demanding requirements for free-space AR smartglasses.

Over the last few years much work has been directed at closing this gap by developing visible VCSELs. Promising results have come from both red VCSELs, based on AlGaInP, and blue and green cousins formed from the AlInGaN material system. However, these devices are at the proof-of-concept stage, with some major challenges holding back commercialization. The manufacturing process is complex, and there are technical issues related to the production of good multi-layer mirrors and efficient active regions. So far, these challenges are imparting a heavy toll on production yield and performance homogeneity.

Tiny threshold currents

We are championing an alternative to the VCSEL through our development of a range of edge-emitting lasers with very low threshold currents. These devices, manufactured with conventional semiconductor wafer processing tools, address the performance gap by delivering a few milliwatts of optical power at a fraction of the drive current required for the typical laser diode of today. Our low-power blue and red laser diodes have threshold currents near 5 mA and operate below 10 mA; and our green variants start lasing near 15 mA, and run between 20 mA and 30 mA (see Figure 2). Operating voltages at threshold are 2.3 V, 3.7 V and 5.5 V for our red, blue and green laser diodes, respectively, and corresponding figures for electrical power consumption at an optical output power of 3 mW are roughly 25 mW, 35 mW and 160 mW.

It is clear from these figures that the performance of green-emitting lasers lags those in the red and blue. A great deal of optimization must still be directed at the crystal quality of the active layers. We expect to make progress on this front in the coming months, through improvements in epitaxial growth by MOCVD.

As well as offering excellent performance, our new laser diodes provide great stability and long-term reliability. Lifetime curves, acquired under constant-current and continuous-wave operation, reveal that for more than 1,000 hours of operation at an output power near 5 mW, there is little or no power drop, indicating that these devices have much promise for deployment in consumer display applications (see Figure 3).

For those that are not familiar with III-nitride compounds, this lifetime for blue and green laser diodes might appear a routine result – but it is not, it is actually a major achievement. Our success is hard won, building on the breakthroughs we have made with our 405 nm SLEDs, which have a projected lifetime of about 5,000 hours.

There are multiple mechanisms that plague the long- and short-term stability of GaN laser diodes. To obtain reliable devices, every threat must be addressed in a manner that ensures high performance.

A well-known issue for GaN laser diodes is a high level of defects, which degrade performance. Like many of our peers, we produce our devices on high-quality free-standing GaN substrates with a low dislocation density, a foundation that helps us to avoid crystal imperfections, particularly in InGaN-based quantum wells.

Another challenge associated with the production of III-nitride laser diodes is the realisation of p-doping. Our approach involves introducing magnesium atoms at carefully controlled dopant levels. We take care to avoid too high a magnesium level, as this could increase optical loss in the chip, and could eventually generate clusters and defects if crystal densities exceed a few 1019 cm-3. Note that it is crucial to not head too far in the other direction – if magnesium concentrations are too low, series resistance climbs alongside the device’s operation voltage, impairing device reliability.

In addition to considerations related to crystal growth, all makers of ridge-waveguide lasers need to ensure that the dielectric mirrors have very low absorption and scattering losses, and that the device has an appropriate package. It is imperative to have an inert package atmosphere that is free of contaminants, as this prevents degradation of the output mirror by photo-induced deposition processes that can take place during operation.

While the production of red laser diodes is more advanced than blue and green variants, success should not be taken for granted. Production draws on a more mature GaAs-based technology, with epitaxy, wafer processing and packaging having been mastered for more than 15 years. However, even though long-term reliability is not a big concern, it is still important to pay attention to the potential onset of sudden failures. While the long emission wavelength, corresponding to a lower photon energy, reduces the likelihood of photo-induced degradation processes, there is still the need to carefully optimise the output mirrors of the laser diodes. Ignore this and there is the threat of sudden degradation by catastrophic optical mirror damage, a common failure mechanism in AlGaInP lasers.

One promising feature of the low-power laser diodes developed by our team is that all three colours produce a similar single-lateral-mode output pattern. This lack of variation significantly simplifies the integration of red, green and blue lasers with common collimation optics. Emission from the cleaved facets of our edge-emitting lasers has the typical elliptic shape, with a divergence that depends on the optical confinement of light in the chip.

With edge-emitting lasers, an important role is played by the two-dimensional refractive index architecture, defined over a cross-section of the ridge-waveguide structure. Along the epitaxial growth direction, angular beam divergence is determined by the thickness and composition of the core and cladding layers that provide waveguiding. For the direction parallel to the growth surface, divergence depends on the ridge-waveguide width, the etch depth and the refractive index of the dielectric layer that provides electrical insulation on each side of the ridge.

Careful optimisation of these parameters enables a great deal of control over the shape of the light emission. With this approach, we have realised extremely similar light patterns for red, green and blue laser diodes, despite their different semiconductor technologies. For all three colours, the typical slow-axis divergence, defined in terms of the full-width at half-maximum, is of the order of 9-10 °; fast-axis angular divergence is 20 ° for the green, 22 ° for the blue, and 24 ° for the red (see Figure 4).

For the red edge-emitter, which has a slightly larger divergence along the growth direction, the conventional core/cladding geometry is derived from that used previously for SLED epitaxial structures. We expect that with further optimization in future designs, we can adjust the optical confinement so that the red fast-axis divergence is even closer to that of the green and blue equivalents.

Our laser diodes set new standards for providing a few milliwatts of optical power at a very low power consumption, and they pave the way for efficient, compact pico-projectors for AR technologies. In these systems, the red, green and blue sources have to be integrated with micro-optics to generate a collinear, collimated white-light beam for the MEMS scanner.

We have significant expertise in this area, as we can draw on experience in producing tuneable lasers and spectrally-combined SLED sources, which make use of highly automated robotic assembly processes to position optical filters and lenses with sub-micron accuracy. We are currently prototyping ultra-compact light engines based on red, green and blue lasers that have a volume of about 50 mm3 (see Figure 5), to enable the next generation of AR smartglasses.

At this year’s Digital Optical Technologies conference, held in June, we presented a module 4.4 mm in length, 4.15 mm wide, and 2.9 mm high. With a perfect blend of semiconductor and packaging technologies at hand, we have a bright future ahead of us.