Scanning an ultraviolet diode laser over a cheap phosphorescent screen should give you a very inexpensive, high-resolution, very-low-refresh-rate display screen.
Projecting an image with a laser suffers from a few problems. One is that, if the part of the laser beam that would fit through someone’s pupil is bright enough, it can be an eye safety hazard. Another is that the beam must be scanned very rapidly over the screen in order to provide the illusion of a stable image, and equipment for scanning the beam very rapidly is energy-hungry and expensive.
We can solve both of these problems with multiple ultraviolet lasers.
Ultraviolet light below about 400nm is blocked by the lens of the eye, so it will not be focused onto the retina and cause the instant retinal burns that are the principal danger of visible and especially infrared lasers. However, it can cause lens damage if continued over a period of time, including cataracts. Below 315nm, it cannot even penetrate the cornea, so it instead will cause only acute photokeratitis, which will heal unless it is very severe indeed.
However, in discussing the kinds of damage that can be caused, we risk losing sight of the quantitative safety factor. The IEC 60825 maximum permissible exposure for a 355nm ultraviolet laser over the course of a millisecond is 100 watts/cm², falling to under 1 watt/cm² if the exposure extends to an entire second, and down to 1 mW/cm² if the exposure continues for a kilosecond. By contrast, the MPE for a visible-light laser is 10 milliwatts if for a millisecond, or 3 milliwatts if for an entire second, converging with the ultraviolet MPE levels at 1000 seconds.
That means that the safe power levels for brief exposure to ultraviolet lasers are around 100 times higher than for visible-light lasers. This means you can send a great deal more energy to your screen at a safe power level. It might be a good idea to wear UV-blocking goggles and to couple your UV laser with a lower power, but painfully bright, but not dangerous, red laser, in order to trigger people’s blink reflex. 1mW should be plenty.
An alternative way to provide eye safety under normal circumstances is to project the laser onto one side of a screen while you look at the other side, enclosing the laser, scanning apparatus, and back of the screen inside a sealed, opaque box.
One unfortunate aspect of this approach, though, is that there aren’t any laser diodes of shorter wavelengths than 370nm commercially available yet. https://www.thorlabs.com/NewGroupPage9.cfm?ObjectGroup_ID=5400 offers Thor Labs’s new 375nm 70mW ultraviolet laser diode L375P70MLD for US$4300.
A phosphorescent screen will not only convert a certain fraction of the laser illumination into light, but also continue to glow over a long period of time, exponentially decaying. This acts as a single-pole low-pass filter on the image signal, attenuating frequencies faster than the time constant of the phosphor by 3dB per octave.
Zinc sulfide’s phosphorescence decay time constant is a few seconds to a few minutes. (I found some paper claiming 9', but that seems implausibly long to me from experience with glow-in-the-dark toys). This means that once a glowing image has been drawn on the zinc sulfide with the laser beam, it will stay there, gradually fading, for a few minutes.
This means that you can draw an image on the screen with a laser over a period of seconds or minutes, and it will continue to be visible. This means that you can draw a fairly complex image even with a fairly slow apparatus for scanning the laser beam across the screen. It also means that you can’t erase anything: you have to wait for it to fade.
It also means that it takes seconds to minutes for the image to reach full brightness, but because of the logarithmic brightness perception of human vision, this is not as much of a problem as you would expect. (I’m guessing this from my experience with analogue oscilloscopes with zinc sulfide screens.)
Copper-doped zinc sulfide is by far the most common glow-in-the-dark material.
It might be worthwhile using a secondary, say, red laser to draw a smaller amount of graphical information that can be instantly erased. This will work better if the screen is not sensitive to the wavelength of the secondary laser.
https://physics.stackexchange.com/questions/79860/why-is-a-laserpointer-able-to-erase-a-glow-in-the-dark-sticker reports that a red laser pointer was able to erase the glow from a glow-in-the-dark sticker (presumably ZnS:Cu). There is a video of this phenomenon at https://www.youtube.com/watch?v=kUteUH7mz0A, but the erasure seems temporary.
Laser diodes themselves are relatively inexpensive; Digi-Key has 1.5mW infrared lasers at US$5.76 and red 5mW lasers at US$12.52. But as the power goes up, cost increases sharply. Their cheapest 20mW laser is US$46.07 (green), their cheapest laser over 40mW is a 120mW 405nm near-ultraviolet unit for US$78.44 (this is the wavelength used by Blu-Ray players), and the only more powerful laser diode for which they list a price is an 175mW near-ultraviolet unit for US$452.
Given this price curve, you can probably get not only more visible light output but also more information on the screen by using several different laser diodes, each pulsed rather than CW, so that sequential points on the screen are often drawn by different lasers. Using two to six separate lasers will increase the energy throughput of the laser bottleneck significantly, without affecting the rest of the system.
Also, laser diodes can be controllably pulsed much more rapidly than mirrors can scan the beam — MHz in the common case or GHz in exceptional cases — and you can draw minimally readable letterforms by interrupting three or four vertical lines:
# ## # # ## ### # # # #
# # # # ### # # # # #### ## #
### ## ### #### ### #### # ##
# # # # # # # # # # # # # #
# # ## # # # # ### # # # #
Given three or four laser beams with a slight angle offset between them horizontally, you could sweep them vertically with a single movement of the mirror while pulsing different dash patterns on them to draw the letters.
You should be able to do a few megapixels this way, but probably not much more.
Suppose you’re using a 2m² screen, with the laser 1.5m away, and you have a 1mRad-divergence beam, which is a pretty normal divergence for a laser pointer. Then your spot will be 1.5mm across, so your screen is only about 1300×1300 “pixels”, for a total of 1.8 megapixels. If you can get a better-quality laser spot of 0.5mRad, you can get four times that, or 7 megapixels.
(The diffraction-limited divergence angle is 2λ/(πw), where λ is the wavelength and w is the beam-waist radius. So for a 650nm red laser to have 1mRad divergence, you need w > 2 · 650nm/(π · .001) ≈ 0.4 mm. Larger collimating optics can produce smaller divergence, but you aren’t going to get that beam waist below 0.4mm even if the beam waist is on the screen.)
However, you may only be able to illuminate a small fraction of these pixels at a time; even expensive laser-show galvos are rated at under 50kpps, and even in two minutes, 50kpps is only 6 million points. Simpler scanning apparatus, perhaps driven by scavenged hard disc voice coils or by paper cone speakers, might only hit 1kpps, and thus only 120k pixels illuminated per laser. With multiple lasers and dash patterns, you could actually paint all of those pixels, a few thousand per second.