Comment by HCIdivision17 | Hacker News Reader

Comment by HCIdivision17 | Hacker News Reader

Parent Full thread

The manufacturing costs won't be too bad, I'd imagine. At least, in terms of the normal shocking costs of manufacturing. If the design is simple enough, that 300 layers may be just 75 repititions of four layers (no doubt not actually the case, but the machine spraying material onto the plates can adapt).

I'm a bit disappointed that the recipe wasn't included; I would have liked to see what the design entailed. All the layers are on dialectrics, and they're all oxides, so the process chambers won't be too hard to control. And I say that knowing full well some poor coater engineer is actually just going to have a devil of a time with it (or not! quarter wave filters are fairly standard stuff).

Most of all, the thing I'd be intersted in is how resistant to error is the design? It sure looks like it doesn't need to be too perfect per layer. For a pair of microscope lenses, precision is everything. In a lightbulb you can probably get away with a lot of error (per layer: as a whole you could easily make the color something goofy). Of course, if it's color swings wildly with minor errors in certain layers, well, that could hole the product up in R&D hell.

> Most of all, the thing I'd be intersted in is how resistant to error is the design?

I feel like this is what keeps the most impressive research out of real world application. Yeah, manufacturing is expensive, but tolerances are just so much more controllable in a lab.

Manufacturing also produces revenue, whereas funding the years of research necessary just to figure out a new nanometer-precise process is a hard investment for most companies to swallow, especially if the end product isn't guaranteed to produce a profit itself.

The good news is that the processes I described above are already well established. The trick is always in the dirty, nasty details, but in this case I'd be bullishly optimistic. I haven't personally worked on a large batch coater, so I can't attest to how hard it is to tune the things in, but your typical coater engineer is a bit eccentric anyhow and will no doubt make it work.

See, the weird bit of how these coatings work is that it's fairly easy to empirically tune the damn things to be angstrom accurate (-ish), but the geometry of how the material is sprayed messes up the uniformity, which in turn ruins your yields (one square centimeter in the center is good, chuck the rest!). But if the coating can withstand some variation between layers, tuning it such that the layers sorta cancel out in the bulk, you'll lose the efficiency, but your yield can absolutely skyrocket. There will likely be key layers at the optical half, quarter, eighth, etc. thicknesses, but if you get enough layers you can probably get away with a lot. You can also just build really neat source geometries to sorta flatten out the spray over a larger area, but that's something I never worked on.

My ignorance of this sort of coating has me worry about it immediately destroying itself from the insane material stresses... Though no idea how well it'll anneal after 1000 hours of constant heat (or worse: on-off heat-cooling cycles!) But since they're all oxides and likely put down together while the substrate stays hot, I bet it works out ok.

> Yeah, manufacturing is expensive, but tolerances are just so much more controllable in a lab.

Not least because, in a lab you are happy if one out of ten prototypes work for your experiment. You don't want 90% spoil in manufacturing.

Each of those layers of oxides are "quarter wave stacks" (from the paper) which means they're somewhere between 200 to 1000 nanometers (depending on the wavelength of light they want to transmit or reflect).

So one of those layers is say, 200 nanometers, you could end up reflecting the visible and transmitting the infrared, which is the opposite of what you want to do.

True, but when you have an awful lot of layers, you can play some silly buggers with it. Even if you botch a layer, it won't neccesarily ruin the whole thing, since the interference betwen the sets of layers also act as filters. I haven't worked on these sorts of stacks, personally, but I'd be surprised if the whole stack dies because one layer was a bit screwed up (I'd expect a degredation in the efficiency, but not necessarily a total reversal - now ask me what happens if your goal is to tune to neutral color and you make a slight error and then I'll flip sides and say "rainbow").

What I'd really expect is that if you make one layer slightly larger, your process almost certainly making all the repititions of that layer larger too, and now yeah, your bulk properties are just hosed.

Reply on news.ycombinator.com