One way is to move the grating in and out of the evanescent field. This is present at the surface of any conventional light-guide and if the grating is inside the field, a guided ray sees it, otherwise not. The field is only ~ 3 microns thick so movement is tiny and several groups have made panels like this. But it's like trying to lift a carpet up and down: air doesn't have much time to rush in and out so it's not easily done.

Another approach is to use liquid crystal gratings. Liquid crystal is almost unique for giving a huge electro-optic change while using very little power: that's why it makes such great displays. But in order to put voltage across the liquid crystal, you need a transparent conductor and these are never all that transparent. For the display, that doesn't matter, but light bounces many times through a light-guide and will gradually be absorbed by the transparent conductor.

Instead, we are experimenting with deep gratings, often called Bragg gratings. These switch on in the sense of diffracting light only if you illuminate them at the right angle. No transparent conductor is needed but we do have to change things around somewhat…

Replace the LEDs in a conventional backlight with lasers and replace its randomly rough surface by a grating and it will produce collimated illumination. The angle of collimated illumination depends on the angle of the lasers so we have a low power backlight which can target rays towards a user as they move around, provided they stay far away.

More likely is that a user will want to be close to the display and it might seem a good idea to add a Fresnel lens in front of the backlight so that we can concentrate rays into one eye of the viewer at a time. But the focal length of a lens is fixed, and viewers tend to move forward and back: how do we vary the angle at which rays converge?

Instead of embossing a single permanent grating on our backlight surface, suppose that we tile it with a line of gratings, each of which can be switched off. Now switch on our laser so that collimated light emerges from the backlight and there will be a ray from one grating which enters the viewer’s eye. That means we can switch off all the other gratings.

Alter the angle of collimation a bit, and a ray from a different grating will reach the eye, so we switch on this grating and keep the others off. Carry on like this and we can time-multiplex a fan of rays which concentrate into the eye, and we can simply alter the timing of things as the eye moves away or closer. That’s why we are interested in switchable gratings.

It would have been hard with old-fashioned TV’s, but liquid crystal displays have the great advantage that they work in two stages. First there is a backlight which is just a slim lamp, then there is the liquid crystal panel which is an electrically switchable stained glass window.

Until now, it hasn’t been possible to know where the viewer is so conventional backlights just glow – they are made deliberately diffuse. They comprise a transparent sheet surrounded by LEDs pumping light into its edges: the LEDs emit light at all possible directions into the sheet, then the sheet is made rough so as to make everything as random as possible.

If instead, we line one edge of the sheet with parallel lasers and the sheet is smooth except for a single grating on its surface, then light from the lasers all travels in the same direction within the sheet as if through an optical fiber, and the grating diffracts light out from all parts of the sheet in the same direction. Instead of emitting diffuse light, our sheet emits collimated light so that when we put a Fresnel lens in front, all the emitted rays concentrate into an eye at the focal point of the lens.

Lasers are expensive, perhaps 100 times more so than LEDs, but we don’t need many because, remember, we are no longer wasting any light, so we expect our components to cost no more than is normal.