>> Large-scale integration of this platform paves the way for fully reconfigurable chip-scale three-dimensional volumetric light projection across the entire visible range.
That's what I thought. Do this a wafer-scale and you have a fully holographic display. Curious how they phase synchronize so many light emitters but the full paper looks like $$
There's only a single light emitter, a fiber laser that feeds the device. That's split in treelike fashion. At each branching part, there's an optical resonator that ensures light is fed evenly to each branch. Each branch has its own phase delay, which requires a bit of software tuning to account for variations.
This is quite a cool paper. I haven't read the full thing yet, but any talk of displays is EXTREMELY premature. This isn't even one pixel- it's steerable in only one direction. In order to steer in two directions, you'd need to stack up hundreds of these. That would give you a pixel... one that's at least a millimeter wide, or 10x wider than a color block on the screen I'm typing this into. Certainly it would be incredibly expensive to make an entire display out of this.
It's even pretty premature for lidar- for safety reasons lidar is in the infrared, usually around 1 micron wavelength, which makes a big difference in terms of manufacturability and price. One of the biggest problems for solid state lidar is the beamwidth, which can be a couple degrees or even >10 degrees in some cases. This is .17 degrees, which is a fantastic result, but still quite large compared to conventional lidar. Top end lidar has angular resolution below .1 degrees and you can integrate data over time to find edges as sharp as the beamspot, which can give you some really insane precision. If we used this system, the intense blur over the spot would require deconvolution and strongly limit precision.
Back of the envelope math: at .75 m (2.5'), .17 degrees between pixels would give you ~2.25 mm pixel pitch. Even at that distance, the blurring will be intense- the middle picture here[1] is semi-accurate. Obviously fidelity wasn't the goal and this doesn't represent the limits of the technology! Don't give up hope. Just expect to see someone demonstrate something that exceeds the requirement for fidelity before it actually gets made into a consumer product.
This paper has similar results to an MIT paper[2], but with 3x higher wavelength. Naturally, 3x higher wavelength way more than 3x harder. This device is also smaller- the MIT device is ~1 mm by .1 mm vs. ~1 mm by ~10 nm. That's mostly due to the fact that the MIT device can steer in two directions- the first direction works with phase shifters just like this device, but the second direction works by heating up the emitters so that they change size (genius).
The device here basically just pipes laser light out through the side of the wafer, which syncs up excellently with the MIT approach. MIT used very long (500 micron) antennas that have a thin-thick-thin wavy pattern[3]; at each wide part (IIRC) refraction causes light to escape at a particular angle. Changing the length of the wide parts changes the resonance of the light inside and therefor the angle light escapes at, and that's how they did beamforming. Normally this would require atomically-perfect manufacturing, but the long antenna averages errors out (although you still end up with just under 1 degree of beamwidth). With a suitable intermediate layer you might be able to plug this work into the MIT antennas and get full steering without needing to stack arrays.
This is all way better than current commercial solid state lidars, which mostly use rectangular patch antennas to project lasers. The optical properties of those systems are really bad. It's a shame there isn't more money in research like this, because it's necessary to make that final jump.
Wouldn't the display application here be something like a scanning AR/VR headset which constructs an image directly on your retina with a small number of beams moving very quickly (kinda like an old CRT display)? In which case you really just need three "pixels" of different colors (assuming it's sufficiently responsive to high frequency control).
1/0.17 = 5.9 ppd is about half the angular resolution of the commercial Index VR headset, so it does still need a bit more progress before it would be competitive. Or a lot more, giving it a few additional factors of 2 for the fact that it's not emitting from directly inside your pupil, and you probably don't want adjacent pixels bleeding into each other (fwhm is still... half maximum which is quite a lot).
> Wouldn't the display application here be something like a scanning AR/VR headset which constructs an image directly on your retina with a small number of beams moving very quickly (kinda like an old CRT display)?
That would only work if the display stays exactly (.01 degrees) in the same spot relative to your eyes. What's it going to do, project a million images, each the width of your pupil? You'd need extremely good eye tracking. You can't put this on a contact lens because the substrate is too thick, in addition to all the manufacturing problems.
> 1/0.17 = 5.9 ppd is about half the angular resolution of the commercial Index VR headset, so it does still need a bit more progress before it would be competitive.
Factors of two is a massive underestimate. The screen in a VR headset is waaaaaay bigger than your retina, which you're projecting onto. Not only that, but you're really concerned about projecting onto the fovea, which has about 1 "pixel" per 20 microns. At 10 cm (4"), 20 microns width would be .011 degrees per pixel. The beam is actually pretty sharp, so halving the beamwidth would lower the bleed to <<10%, which is fine. All together that's ~30x improvement in beamwidth.
That would be what is required for an ideal display. Current VR displays get closer to 0.1 degrees per pixel so it would be a ~3x improvement to be competitive with existing tech, as I said.
No, I'm talking about what is required to get to the same as current VR. Beamwidth and pixels per degree are NOT equivalent. Beamwidth causes adjacent pixels to blur into each other, which is a much bigger issue when projecting onto the eye because it's such a smaller surface. The equivalent factor for a display would be the diffraction blur introduced by a pixel.
Back of the envelope math: at 10 pixels per mm, 500 nm light through a circular aperture[1] has a half-power beamwidth of .0013 degrees, so two orders of magnitude better than this technology. That is what you'd need for an ideal display. The bare minimum is just that the blurring at the retina does not cause pixels three rows over to bleed into each other.
Being aperiodic should mean there's no well defined grating lobes. Basically each emitter pair has a different spacing and thus different grating lobe ___location, so across the whole array they get smeared out to an even sidelobe level. But yes, wondering what the mainlobe contrast is.. it's also >lambda spacing so the efficiency is probably poor.
That's what I thought. Do this a wafer-scale and you have a fully holographic display. Curious how they phase synchronize so many light emitters but the full paper looks like $$