III-V Multijunctions Text Version

This is a text version of the video III-V Multijunctions, a lecture given as part of the Hands-On Photovoltaic Experience Workshop.

>>Myles: Okay. My name's Myles. I work in the III-V group. I've been here a little less than 12 years. I came to post-doc and then, I got a real job here and I've been here ever since. So, I work on III-V's and multijunction solar cells.

I guess Paul has told me that I'm gonna say a lot of things but I'm not sure I made slides on all the things. Paul said [Inaudible].

>>Audience: How much more real is your current job than your post-doc job?

>>Myles: Well...

[Laughter]

It might actually be less real than the post-doc. [Inaudible] Okay. So, Paul talked a lot about silicon PV. This was what was off what was called the first generation of photovoltaics. The second generation was toward lower efficiency, but cheap and large areas – silicon films, things like [Inaudible].

And now, we move to third generation, which was multijunctions, concentrative cells – very high efficiency; not necessarily the cheapest. But idea we produced in form pack was it could be cheap. So, that's what I'm gonna talk about. This is sort of a condensed version of a four-hour tutorial, so, I'll try to get through all of this. We're gonna talk about the materials and growth techniques, the III-Vs, and basics about multijunctions – why would you want a multijunction, how does it work?

This is a form factor of the multijunction that we had developed at NREL. There are some other ones that – we'll talk about this one and then we'll talk about concentration and, if we still have time, I'll talk about [Inaudible]. Please, ask me questions as we go along. So, this is a – this is sort of a zoomed in version of [Inaudible] table that shows group three, group fives, and sort of the main elements. You have the group three elements – aluminum, gallium, and indium were the most important ones.

I've yet to see a set like [Inaudible], but this could be a research not reporting. Under group five side – phosphorous and arsenic, nitrogen, antimony – there's some work on bismuth. So, these are the major elements of the III-V system, and then, you see dopants in the adjacent columns – [Inaudible] carbon are P-type dopants; cell [Inaudible] and [Inaudible] are end-type dopants. [Inaudible due to door creaking loudly open] carbon are considered [Inaudible]. This is a band gap, which is [Inaudible] of the III-V system, and I think this slide kind of helped illustrate why these are such a versatile material system.

You can make very, very highly crystalline material – which I'll show in a minute. They are almost entirely direct band gap. You get some indirect if you get some of the higher band gap concentrations, but throughout most of this system, you get direct band gaps, but you don't get a lot of material. The wide range of band gaps expand all the way from 2 to 21 [Inaudible] all the way down to below half a [Inaudible]. Certainly small.

We can dope these n-type. We can dope these p-type. And I think what's really important is that you can grow a lot of the materials [Inaudible] to each other. So, here's germanium. Here's gallium arsenide.

This time line here between gallium phosphide and indium phosphide crosses over the last constant of gallium arsenide. So, you can grow a triple junction on germanium that has one band gap, two band gap, three band gap, and they all have the same last constant. And to back up, that's just one example of gallium arsenide, but if you think about indium phosphide, you could grow indium phosphide on top of a lower band gap material at the same last constant, and so on for anything – of any substrate that you pick without having substrates available. But, if you manufactured one, then [Inaudible] to grow various materials [Inaudible] to each other. Okay.

This is what I mean by my last [Inaudible]. I assume most of you understand this concept. This is the [Inaudible] structure, which is the crystal material – almost all of III-Vs. So, the unit cell here contains four groups threes, four group [Inaudible], and four groups fives. So, most of these are grown by MOVEE – sometimes called MOCVD. These are chemical paper deposition reactions.

So, they're high temperature. You have some sort of source gas material in the [Inaudible] phase, paralyze those molecules at high temperature because of the heating from the substrate or from whatever the substrate is sitting on. Everything lands on the substrate and it's very strong, [Inaudible], and a driving force for the arrangement in crystal form. That's a snapshot and there are – that is obviously over-simplifying it. There's face separation possibilities that can come up, but this is an illustration.

So, you have a chamber. The gas is coming in the top here. You have this gallium – tri-methyl gallium molecules – just an organic metallic, which is kind of expensive – and then, here you have arsine, which is a hydride, and the heat – the substrate is sitting over here. It's sitting on top of this graphite susceptor we heat up to 700 or 800 degrees, and then, this heat paralyzes these molecules [Inaudible] on the surface. And whatever doesn't go – whatever doesn't land on the surface, goes through an exhaust and we try to clean the exhaust.

So, those of you that haven't watched growth [Inaudible]. We get growth rates in the 6 to 10 microns per hour. There's research in industry to try to push this higher towards 60 microns an hour, but the source materials here are expensive. This is a large-scale production for anything that's not PV – transistors and all kinds of sensors.

The space industry uses this. This is what we use at NREL, for the most part, for our high efficiency cells. So, that's an MOVEE. An area of research that's kind of captivated a lot of people's attention in the last couple of years is called [Inaudible]. It's an old technique that has only recently been applied to PV.

And here, what you do is you start with solid sources. So, gallium, in this reaction. Solid level gallium – you flow a chlorine gas – hydrogen chloride [Inaudible], and it forms gallium chloride, and then, that reacts with whatever happened on the group five to precipitate out material onto the surface. And again, you have a strong, [Inaudible], driving force for crystallinity here. So, what's attractive about this is that these metallic sources are very cheap.

Much, much, much, much, much, much, much cheaper than metal [Inaudible]. So, that's one of the main reasons why this can be less expensive, and also, they can get growth rates in the 200 to 300 microns per hour – so, much faster than the growth rates for MOVEE. A lot of research is being done at NREL on trying to grow all of the materials, all the alloys – the ones we need – for solar cell. So, I wouldn't say it's in its infancy, but it's certainly considerably less mature as a PV growth technique than most [Inaudible]. There's a lot of research opportunities.

Here, you see a schematic of what's going on. There's one chamber over here for arsenide; there's a second chamber over here for phosphides. So, we prepare the liquid gallium or the solid gallium and the tigrides over here and the [Inaudible] or whatever, and they all flow down onto this as a substrate. And then, you shuttle the substrate into the adjacent chamber and you grow the phosphide – shuttle back and forth. The reactor here has two chambers.

You can go back and forth. You can imagine scaling this thing up to more of an inline process where you had one layer that puts down n-type gallium arsenide and you shuttle your substrate over and you put down n-type gadolinium phosphide; you shuttle over to the next one and so on and so forth. This is drawn as a circle, but it can actually be one long line. So, you have substrates coming in this side going out the other side and it's a continuous reactor. Yes?

>>Audience: How do you control the views in between the layers? Say, between like – as you go through, do they start to mix in a certain way? Does the – like, the [Inaudible] – will matching – just [Inaudible] kind of frozen in the layers that they want to be in?

>>Myles: You could get something to mix in. Most of the driving force for crystallinity happens at the growth surface. Once you have buried that layer, it's not completely stable, but mostly stable. There is dopant diffusion since they move a lot. So, they can move a little bit.

Carbon stays put. And then, there's other reactor effects. So, MOCVD for [Inaudible] is what's called memory effect. You can't just turn off selenium. You can turn off the selenium, but it doesn't go away [Inaudible].

Continue to dope the next [Inaudible]. So, things are fairly stable. So, I think this is the last slide I have on growth, but this is work by [Inaudible]. You see the VOC of gallium arsenide cells. It's a fairly basic gallium arsenide cell, and the VOC's a function of a growth rate.

So, these are all kind of stable 1.02 – not the best gallium arsenide, but certainly, not the worst. And then, oh, we have here – 180 microns per hour. That's three microns a minute. If you run your gallium arsenide cell for one minute, you're getting [Inaudible]. So, that sort of gives you a little bit of the glimpse of the possibilities of this technique.

And so, that's the gallium arsenide cells we've made here by MOVCD or MOVPE as it's sometimes called. We get VOCs about 1.1 volts and by [Inaudible]. So, [Inaudible], basically the same VOC. This is kind of representative of the state of the art of HVP. This is actually almost a full percentage point lower than the best gallium arsenide cell.

This is a self-marker here – the best one's [Inaudible] device to cell. That's about 28. Even higher. All right. So, III-Vs enable a lot of PV technologies.

This is a graph that kind of schematically plots temperature and intensity, and we've kind of come to love this point in the middle we call One Sun. It's about 1,000 watts per square meter solar insolence and 25 degree Celsius. But, if you – you can plot a lot of different technologies on this map. So, if you raise the concentration, you get concentrative cells. If you raise the concentration, lower the temperature, you could have an incredibly efficient cell appear.

This space can even be – kind of probably inhabits this space over here where you're basically not changing the intensity, but you might be changing the temperature from very cold to very hot. Of course, [Inaudible].

>>Joe: So, Myles, my take is that the color scale here is the efficiency?

>>Myles: No. Oh, this? Yeah. This color scale's the efficiency. It goes from low efficiency to high efficiency.

I used these colors to represent something that I [Inaudible]. So, you have crazy high efficiency over here and you have very bad efficiency over here. This would be like [Inaudible]. Hydrogen production I put over here is more or less room temperature but may be somewhere between one sun or higher. So, there's a lot of different technologies that are useful here that can be enabled by III-Vs.

These are probably the three most common – space, terrestrial, and [Inaudible]. So, this is – well, let's start at this phase. These – I've taken the III-Vs – I got them for the biggest market for III-V EB and here, you're looking for very high efficiency and very high radiation resistance. So, they fly up into space – you get all these iodine's particles that can really kill off your diffusion length and lower your efficiency. So, you want to design shelves to be tolerant and everything else is – cost design is important.

Weight is important. So, high efficiency, lightweight, and high [Inaudible] are really the most important things for a space cell. On Earth, you're competing against natural gas. You're competing against silicon. So, high efficiency's important, but cost is very important.

Being able to manufacturer at high volume is very important. So, we'll talk about concentration at the end. This is kind of new for III-Vs. It might be enabled by a high throughput MCV or [Inaudible], but you want really high efficiency at one sun; you need very, very low cell cost at [Inaudible]. Okay.

So, I want to talk to the next bunch of slides on why a multijunction and how it works, all right? So, Paul kind of, I think, outlined fundamental losses in solar cell, and, to my mind, there are two. One is that photons have a lower band gap, don't get absorbed; and the second one is that photons with energy in excess of the band gap thermalize two band gaps. So, it's thermalized advantage. You lose all that excess, okay?

There's other losses, but those are the main two. All right? Yes?

>>Joe: If you're a [Inaudible] PV devices –

[Crosstalk and laughter]

>>Myles: It [Inaudible] your material than the other losses, but let's start with these two. So, broadly speaking, the ways to mitigate these losses are to extend the range of absorption and to divide that absorption into smaller bands to reduce the thermalization loss. Okay? So, you want the highest energy photos to be captured – rather than all the way to gallium arsenide, you'd rather divide that into two, have it captured closer to their original energy. So, you'd get a high voltage, all right?

So, imagine it was this nice [Inaudible] Let's suppose it's 1.4 UV. This is the solar spectrum, and you have some optimal band gap, single junction cell. Okay? So, now, you divide this [Inaudible] into two.

Let's suppose you use the 1.8 gallium indium phosphide, and at 1.4, you produce those thermalization losses. Okay? Everything that is absorbed up here only thermalizes this [Inaudible] rather than all the way down to here. So, you increase the efficiency and now, you add another junction. Let's say it's .67 [Inaudible] germanium – forget about the fact that it's indirect, but now, you've captured all of this light and all these photons thermalize way down to this band edge here.

So, you're gonna increase the efficiency a little bit. Not a lot, but a little bit. Okay. And now, if we divide this bottom purple range into two – so, now, all the photons here are thermalized all the way down to here. So, now, you have a four-junction cell, okay?

So, compare this cell to this cell. The absorption range has gone from this small – well, not small, but this chunk of the spectrum to all the way up to 1,800 and you reduce the thermalization losses here one, two, three, four [Inaudible], okay? So, that's the power of multijunction solar cells. This is another way of looking at it. Now, this is in terms of energy.

So, we put our highest band gap on top of that second, third, fourth junction, and there's many, many ways of – well, there's many calculations out there of the efficiency. Depends what assumptions you make, the [Inaudible] the concentration, whatever. The main point here is that you increase the efficiency as you add the number of junctions, but it rolls off. It doesn't go up linearly. It doesn't continue forever.

It's a dimension [Inaudible] here, okay? Question? So, there's a lot of different ways of implementing this. Here's three. One is to split the spectrum with optics.

This is a project I worked on when I came here as post-doc. This is another [Inaudible]. You can't see the range so well. The idea is that you have light coming in and then, an optic would separate the photons into different bands and send these photons to this solar cell and the second set to the next solar cell, the third set to the third solar cell and so on. So, if you have to – you can make really nice solar cells and then, rearrange them in space, and then, we do all of the respective splitting with the optics.

Another way is to make it a kind of a stack. So, here, you have a gallium arsenide cell on a substrate. Let's suppose it's a transparent – it's a double-side [Inaudible]. Then, you have a second junction on the second substrate, and you make an interconnection that is transparent. This could be conductive, or it could be non-conductive.

So, you could make this into a two-terminal or a four-terminal device, but now, you're splitting it sort of in the more conventional way where all the light comes through the top and the highest energy light you capture here, and the lower energy light is captured here. But ultimately, you're connecting this together and this is what we've done a lot of work on in terms of making pilots between III-Vs and silicon. And then, the last one over here is this monolithic growth, where you start with one substrate and you grow everything together. This is a Venn diagram of a multijunction. You may have seen this.

I'll show this, but everything on the left of this yellow line is just a single junction. So, you have an emitter, you have a base, you have confinement on the front and back. Same thing on the second junction. But, we put a tunnel junction between them. So, this is drawn as an N on P solar cell – N-type emitter; P-type base.

So, you have to be able to translate from p-type majority carriers to n-type majority carriers, what you use to tunnel junction. So, it's an oppositely polarized very highly doped junction. It's [Inaudible] doped [Inaudible]. What you see is that the valance band here, the conduction band here, the [Inaudible] the conduction band red are the same, okay? So, there are available states on either sides for electrons and holes to fall through.

So, one way of thinking about a tunnel junction is it simply swaps majority carrier. It takes you from a majority carrier to a majority carrier electron. So, this is a two-junction, and you can imagine just stacking this all the way to make a six-junction cell if you want. Okay? So, if you have a single junction cell, you've got [Inaudible] that's there [Inaudible]. How do you combine the [Inaudible] for a multijunction? And you're not just gonna focus on the serious connected multijunction, right? That [Inaudible].

>>Audience: [Inaudible] slide, sir.

>>Myles: This one over here, we have a terminal on the body and a terminal on top and everything else is continuous. All right? So, the major point here to note is that every current, the voltages have to act. So, you have one current going through the entire stack. You just added voltages at each current – at those currents, right?

So, schematically, this is a tandem where you have a gallium arsenide cell in green and a gallium indium phosphide cell. The gallium indium phosphide cell has a high band gap, so, it has a high voltage. At every current, you add new voltages. So, you can see this light, for example. The red is the VOC of the tandem.

It's just something the VOC's been – the two underlying cells. This is a maxed power point. This is [Inaudible] – getting max power point is supposed to be some of two voltages. Notice that because these two are not current matched, the gallium arsenide cell here is at it's maxed out point. But the gallium indium phosphide's actually a higher voltage, right, and it's already on this nearly vertical portion of the [Inaudible].

And here, I [Inaudible]. Okay. So, I'm gonna pause for a second and show you this plot of the spectra. This is the AM0 – the space spectra – the global and the direct. And the main point that I want to point out here is that you have these absorption bands – this is water absorption bands here.

Here is water absorption bands, [Inaudible] absorption band. If you just consider the space spectra, which kind of outlines this as envelope with a nearly black body – there's no bands at all, right? So, you have the freedom to choose the band – well, you always have freedom to choose the band gap, but you can pick the band gaps wherever you want. But, in these terrestrial spectra, you got these big dips. So, you want to be careful that you don't put a band gap smack in the middle of the dip, right?

Because you're gonna be losing voltage without gaining current. So, if you think, just for a second, imagine you have a band gap that's right here and then, its sister cell had a band gap right there. They would have the same current, but they would have very different voltages. So, one would just be a much worse version of the other one. So, when you design multijunction, one of the things you're doing is you are optimizing for the spectra, and one of the things that you're doing is you're situating the band gaps around those absorption bands, all right?

So, this is – these are contours of efficiency for just the two-junction, which is the simplest to understand. So, this is a top-cell band gap, the bottom cell band gap, and we've done some thinning of the band gap at the top cell here to [Inaudible] match a little bit. But you see these contours and there's a dotted line here and kind of through the middle. So, everything on the top side is top limited – meaning the currents in the top cell are always gonna be lower than the bottom cell, and everything lower here, there's gonna be bottom. Right along this dotted line [Inaudible] current match, right?

See? There's an [Inaudible]. And you can draw 3D contours, 4D, 5, et cetera – however it is you could visualize these things for how many junctions you want, all right? So, here, I've [Inaudible] two different tandems. So, in black, it's kind of the conventional gallium indium phosphide, gallium arsenide – 1.4 gallium arsenide [Inaudible] gallium phosphide.

And note that it doesn't actually coincide neither with the bull's eye nor, even with this dotted line. Now, gallium indium phosphide can actually expand a few band gaps, so, you [Inaudible] 1.9 would actually come pretty close to the [Inaudible]. But this is kind of the conventional cell, because its lattice matches. You get really good material quality, right?

Whereas this bulls eye – 1.1 EV on the bottom, 1.7 EV over here – actually, the bulls eyes a little bit off that. The only band gap that is lattice matched here is silicate, but it's been 30 years of work trying to grow on silicate – with progress, but never ultimately what I would call success. So, we have managed to grow something that's close to there using lattice mismatched technique and I'll get to that in the next slide. But the main point here is that you get different voltages and currents between these two cells because of the different band gaps. So, this black one is a lower current/higher voltage cell and the red on is a higher current/lower voltage cell.

This is actually an old slide, so, the LG – which is a company in Korea – has actually recently made a [Inaudible] gallium arsenide cell that was 32.8 [Inaudible] pretty close [Inaudible]. Okay. So, you can only go so far being lattice matched, and the reason is – there's only so many band gaps that are available – or lattice matches. So, one strategy for accessing more band gaps is to change the lattice constant. Does everybody know what I mean by "lattice constant"?

I showed a graphic for it. All right. So, the problem with changing the lattice constants is that you have to engineer around these dislocations. But, if you can do it properly or if you can do it well, then there's a lot of energy space to be harnessed. So, here's gallium arsenide in green.

I'm gonna skip over three, four, and five junction to just jump to our sixth junction work here 'cause it's pretty up to date and new, but there's been many years of work on stuff between there and two – [Inaudible]. So, these are the band gaps that we've been working on here. So, this is gallium arsenide, gallium indium phosphide looped in over here. Instead, we're dividing this top region into three, so, we're using aluminum gallium arsenide – a cell at 1.7 – and aluminum gallium indium phosphide cell of 2.1, but, you know, these are all lattice matches. So, for this guy, there's not a lot to it.

So, if you lattice mismatch, then, here's the 1.2 EV cell. Here's the .9 EV cell. There's a 7 EV cell. There's a lot of alloys on the gallium indium arsenide line you can access, and you can activate [Inaudible] phosphorus – you get gallium indium arsenide phosphorus. So, there's a lot of band gaps to be met [Inaudible] change the last constant.

But, changing the last constant is not [Inaudible]. It's somewhat akin to taking a small cardboard box and trying to cram a large cardboard box into it. So, you have to kind of fold and bend around that perimeter. If you look far enough away, the top cardboard box is gonna look nice and the bottom one's gonna look nice. But, right around that place where you jammed one into the other, there's a lot of folds and bends and crappy material, right?

So, that's sort of the engineering part of the tier to make this work. So, just briefly, this is how we grow – and you can read this well. This is how it's grown. It's grown – the inverted metamorphic multijunction cell is grown [Inaudible]. So, we start with a substrate, we grow the top cell and the second cell, and we change the last constant and we grow the bottom cell.

And then, we flip this thing over and you bond this to a handle and you edge off the substrate and you shine [Inaudible]. So, in the final orientation of this triple junction cell, highest band gap is on top of the middle band gap, followed by the bottom band gap. Now, let me back up a second to tell you why you'd want to do this, why you'd want to grow this inverted. The main reason you may want to grow this inverted is that if you start from the top of what you'd finally want – the highest band gap – this is gonna produce the highest voltage. This is gonna produce the second highest voltage and so on.

So, the largest contributor to your power comes through this top cell. The lowest contributor to your power comes from this bottom cell all the way down here. The same current through all of these, but the voltage goes with the band gap, okay? So, it makes sense to make these the highest material quality, and if we're gonna have to sacrifice some material quality, sacrifice it in the lowest power producing sub-cells. Thus, the inverted.

You could have grown this upright. You could have started here and grow this if you had – [Inaudible] grown this lattice matching indium phosphide and then change the lattice constant downward, okay? But then, all your cells [Inaudible] mismatched, including the top power resource. Okay? So, I showed you this.

So, this is sort of a schematic of this lattice mismatch. So, here's your substrates. It's got a narrow band. And [Inaudible] blasts constant, then, you slowly increase the lattice constant. When you do – just like I said with this cardboard box analogy – you're getting all this – the folds along the edge.

Here, you're getting misfit dislocations, all right, 'cause you're [Inaudible] to the interfaces between the various size steps. So, [Inaudible] you're separating the lattice constant down, okay? And every time you change the lattice constant, you have to mediate that with that discontinuity by introducing dislocations. All right?

If you look at it [Inaudible] point of view, here you have – between this vertical line and this vertical line here – 10 atoms, 11 atoms. So, just a dislocation. Okay? So, in terms of engineering – here, you have these dislocations gliding very nicely through beautiful material. But the material – you have to admit – it kind of looks like this.

It's not smooth. So, dislocations that you're gliding can get trapped, all right? If the dislocation stays in the interface, then it's fairly innocuous. But, if it churns up, then, it runs into your active material and it's a source of – it's a location [Inaudible]. All right.

This is a TM that shows when we separate it upwards we're changing lattice constant. You can see that at the bottom and at the top, you have really nice material quality, but all the dislocations are confined to these layers in between. Okay? And this is some thesis work from Ohio State a number of years ago that showed that voltage is a function of the threat of dislocation. That's key.

What you see is that you have nice material quality. You have low [Inaudible] dislocation. [Inaudible] As you add more and more of these structural dislocations, discrepant dislocations, you get [Inaudible] drop [Inaudible]. All right. Let me see [Inaudible].

So, this is a light V metamorphic-ing gas with two different dislocation entities and there's a measurable difference in the voltage here that's on supporting [Inaudible] between these two when we change the dislocation [Inaudible]. All right. So, we've been working, for the last few years, here on a six-j cell – six-junction cell. Our charge is to hit 50 percent efficiency at concentration. Here, you can see the [Inaudible] quantum efficiency for all six junctions.

This is grown monolithically. So, we start here and grow in this direction and then, we re-orient this whole thing during processing. This takes about seven hours to grow, plus a little bit of preparatory work, and internally, we're getting over 90 percent ITV across the whole wavelength range. I'd like that to be a little bit higher, but we were pretty excited about that. And then, when you add an anti-reflection coating to the front, we re-measured – you get [Inaudible] – see some additional losses. So, here, you're losing some [Inaudible] because of absorption loss in [Inaudible]. But, at one sun – this is the 35 percent cell. So, that's kind of where we're at here at NREL, in terms of our III-V program.

>>Audience: Myles?

>>Myles: Yeah?

>>Audience: Quick question. So, can you, I guess, confirm my suspicion of the primary driver for the multijunction – the model type one versus the stacked one – is more like a cost standpoint as to why you would pursue one [Inaudible] another? Is it that the benefit you gain – cost of the model – is trying to – like, from all the dislocations – that that loss you gain in the reduced cost relative to the [Inaudible]?

>>Myles: So, I don't think there's a clear winner. I think that you have simple structure in terms of growth. Well, you have simple structure in terms of processing. You don't have to worry about that initial error. [Inaudible] You don't have to use two substrates.

Depending on what the application was, those substrates can be expensive. I'm sure the substrates are expensive. And added to the top of that substrate [Inaudible], but there's a lot of work out there trying to remove the material from the substrate and reuse it. So, if you only have to remove one substrate, that's gonna be, approximately, half as expensive than if you have to remove two substrates. It's probably less expensive, because that other substrate's probably [Inaudible] phosphide.

That said, if you make a mechanical stack, you can grow all the material lattice matched for the substrate. That will certainly have better material quality. So, it depends on the application.

>>Audience: But that is – in simplified terms, I guess, is the general tradeoff that in what you gain with material quality with the stack, you lose in cost relative –

>>Myles: You're losing complexity. Adele's gonna [Inaudible].

>>Adele: Yeah. So, I think it's more an issue of whether you're operating at one sun versus concentration, because if you have a stacked device, then you have layers in between where you have to worry about getting [Inaudible]. Either it's a two-terminal device or you need a really low-resistance contact layer in between or you're [Inaudible] versus growing directly [Inaudible] one thing [Inaudible] resistance – which, if you're looking at concentrator applications, that's the dominant loss mechanism area. This is the losses?

>>Myles: Yeah. And concentration.

>>Adele: Yeah. Versus at one sun, it's not a big deal. So, you know, I think that's one of the main design criterias. Like, what's your operation? What [Inaudible] need to be done there?

>>Audience: Okay.

>>Myles: Okay. So, this shouldn't shock you that solar cell's not [Inaudible] electronic [Inaudible]. It takes in light and it puts out electricity. But, when you think about that in the context of a multijunction, you have a very, very powerful way of looking at the eternal function of the device. So, this is kind of meant to show all the different junctions, since now we've made three and four and five and six junctions.

You have light coming into each of the junctions. You have light being emitted from each of the junctions, and some of that light gets emitted – comes out to the front – and some of that light then goes to the next junction. There's a lot of optical interactions. The reason I say light is emitted here is that there's a very – there's a reciprocity, really, between – or there's a – what's the word I'm looking for? An LED and a solar cell are more or less complimentary devices, right?

One of them, you absorb light and put out electricity; the other one, you put in electricity and emit light, okay? So, they're designed differently. You would optimize them differently for the two different functions. But, at the core, the physics of them are [Inaudible]. So, these solar cells are not the best LEDs ever made, but they still have that LED light property – that they emit light.

We can use that emission to our advantage. So, this is called electroluminescence. You operate like an LED, you put [Inaudible] electricity – you put current into the device and you get an emission peak at each of the junctions. So, this is a six-j – top junction, second, third, fourth, fifth, sixth. And the intensity of that emission is exponentially related to the voltage across that [Inaudible], right?

So, if you're calibrated your electroluminescence measurement carefully and you have good measurements of the intensity as a function of the ejection current, you can essentially measure the voltage of each sub-cell. You can't do that by just measuring the dark IV. If you just measure the dark IV, normal [Inaudible] voltage, you get the sum – you get the total dark IV. But, if you use the light that's being emitted, you can actually resolve into various sub-cells. All right?

So, these are two ways of looking at those data. On the left, you have the external ray efficiency – so, that's just the current coming out in terms of photons divided by current going in in terms of [Inaudible]. And here, we used the mathematics called the reciprocity theory to translate these data into essentially the dark IV – something they call the optimal dark IV. So, just current versus voltage. And then, the total dark IV would be off scale.

If something's on, you get the total measured dark. Okay. So, if you look over here, what you see is that at high currencies have all gone towards [Inaudible]. I think Paul talked about that. You have – have lower currents, you have [Inaudible] to behavior, where you're dominated by recombination and you get the space chart [Inaudible], whereas at high currents, you have a [Inaudible] dominated by a combination of [Inaudible].

So, all six of these junctions have evolved to work [Inaudible]. And here, you can see it as external rating efficiency of the function [Inaudible]. We can get to flavor, if you want, but it's essentially above a 10 percent grade material quality down to about .1 percent. Okay.

Well, you can see the material quality of each of the junctions a little bit more. It's a very powerful technique. And then, we can also look at the voltages. So, for example, you see the green is the gallium arsenide cell in this multijunction, and it's about .98 voltage. Pretty crappy gallium arsenide cell.

A good gallium arsenide cell I showed you is 1.1. [Inaudible] maybe 1.05. So, we're losing 60-70 millivolts. I'm not sure why. But only [Inaudible].

All right. So, before I run out of time here, I want to talk about concentration, all right? So, the power density of sunlight is 1,000 watts of a square meter. We convert a fraction of this and we want a lot of electricity. So, how do you do this cost effectively?

Well, one is to minimize the cost of the cells and the other is to minimize the area of the cell makeup – that intensity with concentration. So, instead of making a cell that's this big, make it 1,000 times smaller and concentrate the light 1,000 times. Okay? So, maybe you do this as a kid. You could put a solar cell at the focus of this magnifying glass and a smaller cell and more intense light – instead of burning an ant or a leaf.

All right. And it's not that hard to make small solar cells. These are some we made. We typically make cells kind of in this 5 by 5 or 3 by 3, but we made some that are 250 by 250 [Inaudible]. So, very, very small solar cells.

Here's some photos from industry of concentrated systems. This is an industry that has fallen on some hard times. Silicon, making up 95 percent of the market share, has kind of squeezed out a lot of things. I still maintain – a lot of people here still maintain – that manufactured at scale, this could be a more cost-effective way of [Inaudible] electricity. It uses a lot less area and you have area between all these panels to do stuff – farming, whatever it is.

Manufacturing to scale is a challenge. So, what does concentration do for efficiency? So, imagine you have a one sun IV curve or an IV curve at one sun. Here it is. It's got a JFC of 10.

So, your efficiency – that's just what Paul said it was – JFC times VOC. You now put on 10 times the intensity of light. You increase your JFC by a factor of 10. That's called the [Inaudible]. So, your JFC goes up by a factor of 10.

Your input power also went up by a factor of 10. These 10s cancel out. But, notice that your voltage increased. Voltage increases longer with intensity. And so, the efficiency here is more than the efficiency here, even after you've normalized for the [Inaudible].

So, that's – in terms of the performance of the solar cell, that's what's driving the higher – the better behavior of the concentrative cells – that you're increasing your voltage. And physically, what's going on is that as you increase the intensity, you are increasing the density of minority carriers. That pushes apart your quasi [Inaudible] levels, which ultimately increases your voltage. And you can see this in data. So, this is the gallium phosphide cell.

Here's your voltage going up longer [Inaudible] with intensity. Your [Inaudible] factor goes up a little bit until it rolls over – which is what Adele was getting at – in series resistance. So, your efficiency will go through some peak and then rollover. So, here, we've gone through the 12 and a half percent at 1 sun, so you look at 14 and a half percent and 200 suns. And I wanted to show you this status of our six-j work here.

So, these are concentrator IV curves that we've taken here. There's some weirdness going on in JFC, but we see that the voltage is increasing. There's a little [Inaudible] heating. Fill factor goes up. The efficiency kind of goes up.

[Inaudible due to door creaking loudly] which has to do with how well you conducted the spectrums of these points here of the data. And we did some mathematical parameters, and this is the curve that IV used. But, you see the efficiency here is about 42-43 percent. If you try and pass that with the [Inaudible] – earlier, which is about 36 or so [Inaudible]. Okay.

So, I'm just gonna leave you with this key challenges slide. There's lots of applications. We didn't really talk about hydrogen, but the main ones that we've been focused on here are one sun, concentrator, and space. And in terms of one sun, lower epitaxial growth costs and cheaper substrates. The substrates that can be used are probably two of the three key challenges.

The third isn't so much a key challenge, but hydro-efficiencies are always beneficial. So, going forward to the next 3 years in our group – and hopefully, for the next 10 years – we are really focusing on once an application's trying to make a III-V cell that is cost competitive for one set of applications – so, that entire efficiencies, lower epitaxial growth cost, and less expensive substrates – we have worked on concentrators for a long time here. There are places in the industry that worked on concentrators. Continued increases in efficiency will drive down costs, but I think it's probably not correct to say that the major problems with concentrator lie at system level, at the cell level. So, we can make a 50 percent cell, which is better than a 46 percent cell.

I'm not sure that's gonna suddenly make that industry viable [Inaudible]. And then, the space industry continues to roll out III-Vs. They're – these [Inaudible], increased it to 32 or 33.3 and [Inaudible]. So, here's some photos. Questions? Yes?

>>Audience: [Inaudible due to being too far from microphone]

>>Myles: So, you have current going through the whole thing. It's the same [Inaudible], right? So, ideally, you would design this with each cell as the same – generating the same current. I wouldn't say each layer. I would say each junction, or each solar cell is gonna be composed of probably four layers – maybe a few extra ones – and you're getting full scale of that stack.

You're getting [Inaudible] the next one. So, yes, you get voltage out of each of the junctions. Does that make sense?

>>Audience: Yes.

>>Audience: As much as I have a question about the [Inaudible] products, is it something you actually can control? Like, I understand III-Vs have some intrinsic – that intrinsic radiation may be hard, but is there something you can actually make in your design that makes it actually better for [Inaudible]?

>>Myles: Yeah. That's a new question. The main thing that radiation does is it makes – it lowers your diffusion length. So, if you can design a cell so that you are less sensitive to the diffusion length, then you have basic radiation arc. So, if your diffusion length is a micron and you made your cell a micron at the beginning of its life, you are not getting a radiation arc, because as that diffusion length drops below a micron, now, you have a cell that's too thin.

So, if you can thin the cell to a 10th of a micron, then, you might have some initial loss of efficiency, but your [Inaudible] radiation arc's gonna have the same efficiency throughout its life. And that's [Inaudible] So, what satellite manufacturers do is they figure out how long of a mission do I need? I need 10 years and I need a minimum of 31 percent efficient [Inaudible]. So, they design it so that after 10 years and all the radiation damage that you've accumulated in 10 years, you still have 31 percent.

So, they don't care what the beginning of life efficiency is. It's probably gonna be higher at the beginning of life. So, that's plenty of room to do [Inaudible], so that the end of the mission, it still produces enough. So, the main way to do it is to make things – one of the main ways is to make the cells or the culture radiation most susceptible to making it thinner. Let's see if I get this right.

The phosphides tend to be more radiation hard than the arsenides, and the thinking there is that the phosphorous atom is a lot smaller than the arsenic atom. And so, the whole thing itself can be [Inaudible]. But that seems very handy. I don't know if there's a huge number of fundamental [Inaudible] radiation arcs. There's a lot of phenomenal [Inaudible] studies and we would love to have funding to do some of those fun things. But –

>>Audience: But if you just [Inaudible] them, the more [Inaudible] product than the multijunction is more radiation than a single junc.

>>Myles: Well, I don't think that – they're more efficient at the beginning of life, the more efficient at end of life. No. They're probably not as radiation [Inaudible]. The reason is that unless you – is that as you change the thicknesses, you're going to change the photocurrents. You can change the current [Inaudible].

So, a tandem is going to be much more susceptible to your spectra variation. So, you have to make sure you design it for end of life. One more question? No?

>>Audience: Are there any scale on the [Inaudible] plant [Inaudible]?

>>Myles: Yeah. There's some plants. There's one in Alamosa, which is four hours away from here. It's some number of megawatts. It's not a big plant.

There's one in South Africa. There's one in Spain. There's one in China. There's a dozen around the world. This is sort of a technology that kind of had its sweet spot maybe 10 years ago.

Maybe it'll come back again. [Inaudible] There's nothing like gigawatts and these are all sort of low megawatt scale.

>>Joe: You are getting some things away, right? You don't have to worry about the stability, say, of your device in the same way, but you need it to be specular and you need [Inaudible]. It doesn't do well in diffuse light and weathering of the reflector is such a huge issue.

>>Myles: We have moving parts, so you –

[Crosstalk]

>>Joe: Yeah.

>>Audience: I agree. Maintenance.

>>Myles: All right. Well, I'll be here all week. Thank you.

[Applause]

[End of Audio]