Is it possible for you, with the HN forum limits, to describe some such interesting major issues with LED (for a physics illiterate)?
So, to start at the beginning: how do we grow these things?
At the heart of every LED is a crystal. For blue LEDs, this crystal is gallium nitride (GaN). How do we grow these crystals at scale? This achievement is behind this Nobel prize, but it's far from a solved problem even today. They found out you could grow GaN by flowing hot gasses containing Ga and N on top of an artificial sapphire film, which would act as a template for the crystal to grow. The problem is that GaN crystals and sapphire crystals are slightly different sizes (the gaps between their constituent atoms is different) so they don't match up exactly, and this resulted in a lot of strain and defects in the GaN crystals–but sapphire was the best we had, and it worked. The trouble is, sapphire isn't great: it's expensive and you can only make small crystals.
The big push now is to find a way of growing GaN on top of silicon. This would make it a lot easier to grow larger crystals and would also make it a lot easier to incorporate GaN into silicon-based devices. But growing on silicon comes with its own problems, so it's even harder to grow high quality films. Specifically, cracking as the crystal cools from its growth temperature is a big problem because the thermal conductivity of silicon and GaN is so different.
There's also a push to try and grow GaN directly from a liquid, which would be more like how we grow silicon 'from scratch' (rather than growing on top of a template). This shows a lot of promise but it's a long way from commercial viability yet.
I'll try and come back and comment on some other issues later on if there's interest :)
I work on biomolecular crystallography and nucleation is half the battle! The other half, used to be size, but thanks to microbeam beam lines, at synchrotrons like APS, we can get away with very tiny crystals for X-Ray diffraction data.
I am fascinated by the idea of using semi/synthetic materials as seeding agents. But then again, biomolecular crystals have huge unit cells compared to semiconductors. People have tried zeolites in the past, but surprisingly a random speck of dust sometimes works better than the best designed substrate.
This is exactly what's often done. It's all about scale fundamentally. It's slow to grow crystals, they have to be very high quality single crystals, and they can't have even the slightest trace of impurities, and ideally they're going to be large and easy to process too. If an alternative substrate is also hard to grow at scale, it's not going to work. But I don't want it make it sound like it's just a scaling issue, because to my knowledge better substrates haven't been found even as a proof of concept. It's not just lattice parameters, but a whole host of other things too. But people are still looking :)
> People have tried zeolites in the past, but surprisingly a random speck of dust sometimes works better than the best designed substrate.
Hah, that sounds both incredibly frustrating and good fun!
By the way, your series of posts on this topic have been superlative—many thanks.
> They found out you could grow GaN by flowing hot gasses containing Ga and N on top of an artificial sapphire film, which would act as a template for the crystal to grow.
This must be what [3] refers to. Mix molten gallium with nitrogen at 100 atm, 1000 ˚C. Alternatively, mix gallium with ammonia. Get a powder of GaN, then vapor deposit it into layers.
> The problem is that GaN crystals and sapphire crystals are slightly different sizes (the gaps between their constituent atoms is different) so they don't match up exactly
Right, several articles mention matching lattice constants. Seems to be a big problem. In fact, [2] mentions that the first substrates used for growing GaN were sapphire, zinc oxide, and silicon carbide. A chart [4] shows lattice constants, which I don't fully understand, but GaN's 3.186 Å is pretty close to SiC's 3.086 Å. So this seems to make sense.
How do you compare a single lattice constant like ZnO: 4.580 Å with a pair like GaN's 3.186 Å, 5.186 Å?
[1] https://en.wikipedia.org/wiki/Light-emitting_diode#Ultraviol...
[2] https://en.wikipedia.org/wiki/Gallium_nitride
[3] https://en.wikipedia.org/wiki/Gallium_nitride#Bulk_substrate...
[4] http://sector7.xray.aps.anl.gov/calculators/crystal_lattice_...
Not quite. If you want something to search for, search for "metalorganic vapour phase epitaxy" (MOVPE) or "metalorganic vapour deposition" (MOCVD).
> How do you compare a single lattice constant like ZnO: 4.580 Å with a pair like GaN's 3.186 Å, 5.186 Å?
This is a harder question than it might seem!
You can easily calculate a lattice misfit as a percentage if the crystals are the same shape: (a_substrate - a_film)/a_film. If it's low, the films will be strained, if it's higher then the films will have to relax through some deformation process resulting in disruption and defects at the interface. It's a complex process, and there's no easy rule for what will happen (keyword to search for is "Matthews Blakeslee" who came up with a model to predict how thick a film could be for a given lattice misfit before you get these defects, but in practice it's quite limited).
Care must be taken to directly compare lattice parameters though. To pick a simple example, imagine you have one crystal with a lattice parameter exactly twice that of another. On paper, that'd be a lot of misfit, but because they tile perfectly in practice it might work really well. Likewise, you can imagine lining up two square crystals, you could imagine being able to line up the diagonal of one crystal with the sides of the other crystal if one lattice parameter if the ratio of their lattice parameters is 1:sqrt(2). So it's not as simple as just looking to see how similar two numbers are, you have to consider the geometry of the crystals too.
This is where it gets a little complicated. For your specific example of ZnO and GaN, the ZnO value you have is for cubic ZnO so its three lattice parameters are the same (a=b=c like the sides of a cube) which is why only one is quoted (a = 4.580 Å) whereas GaN is hexagonal (a=b!=c) which is why two are quoted (a = 3.186 Å, c = 5.186 Å).
[Aside: GaN is often grown on its c-plane, in which case we can neglect the c parameter for working out the lattice misfit. This is something that's difficult for me to explain in words, but if you're interested in understanding it a bit better, search for "Bravais lattices" so you more easily visualise what these lattice parameters refer to. This means we only need to consider the a values when working out the misfit.]
So you'd want to compare the 4.580 Å value to the 3.186 Å value and ignore the 5.18 6Å value. But because the GaN crystal is not just a different size but also different shape to the ZnO crystal (hexagonal vs. cubic), it's actually more complicated. However, luckily for you, ZnO also exists in a hexagonal form just like GaN and in that case has lattice parameters a ~= 3.25 Å and c ~= 5.21 Å, so the misfit between ZnO and GaN in this case would be about 2%?
If you're curious, it seems like people do grow ZnO on GaN and vice versa, so you picked a good example to ask about :)
[Edited a few times for clarity.]
In the crystal, we have two types of particles flowing around. Electrons and holes. Holes aren't 'real' particles, but they still exist: much like bubbles in a bottle of water. It's an awkward analogy, but we can imagine two bottles of water: one that's completely full except for a few bubbles, and another that is higher in energy that's almost empty apart from a few drops of water. These drops of water sloshing around are the electrons moving through the crystal. What happens when an electron (water droplet) and hole (bubble) meet? They annihilate each other, and in the process give off a little spark of energy in the form of a piece of light. For gallium nitride, the light's blue because of the difference in energy between the electrons and the holes.
To get higher energy light (more blue->purple->ultra violet) we can replace some of the gallium with aluminium. To get lower energy light (green) we replace some of the gallium with indium. So far, so good.
One of many outstanding problems though is that LEDs have much poorer efficiencies when we want to emit green light. Lots of different combinations and permutations have been tried, but none are great, and we're still looking for a better solution–this might be in the form of finding a brand new material, or of growing nitride-based crystals in more unusual forms, such as nanowires (imagine a forest of crystals standing on a sapphire 'floor') or quantum dots (tiny little pyramids). This can help because electrons act in very different ways when they're confined in certain dimensions, e.g. if they can only move in along a straight line, and this can be exploited to make better devices. So lots of people are trying this, not just for green LEDs, but for all manner of different devices.
So this is what we do when we want to intentionally change the colour that's emitted. But there's lots that happens to unintentionally change the colour too, which is a problem if you're trying to make thousands of light bulbs that should all look the same! This is the result of much more subtle problems that exist on the very smallest atomic length scales...