Hybrid Perovskites Text Version
This is a text version of the video Hybrid Perovskites, a lecture given as part of the Hands-On Photovoltaic Experience Workshop.
>>Joe: Okay. So, I'm gonna talk about perovskite solar cells, and first thing I want to do – these are some of our recent people who've just matriculated. So, Jeff's last week was beginning of this month. He's now gonna be at Hill College. Same name as the program.
Philip Schultz, who was a post doc with me for a number of years and is now at DVF outside of Paris in French and this is done with Macron – making our planet great again – and Jean Lee, who's recently left – he left – it was Wednesday, May – back home to China. And oh, I forgot to do this. [Inaudible] And then, these are kind of the current set of people who are engaged in activities here at NREL. So, it's a large collection of folks, and you'll see a lot of them around the hallways. Jeff – he just left, so, he still makes the slide with people who are still working here.
All right. So, you've already heard about this, right? We talked about this [Inaudible] told you everything you need to know. It's everywhere. Everybody likes it, invested lots of money on it in the form of sea moss.
We probably know more about [Inaudible] than we know about any other material. That's not to say that we know it all, but we know a lot. Then, there are EASLIES. That's what Myles talked about. These are the – leaving materials as God intended, and they have the ability to go to extremes in efficiency by doing the strategy of essentially making multijunctions as the building inspector required. [Inaudible] the loss mechanisms.
And this is where we find them, in addition to the [Inaudible], which we can do in concentration. And there are efforts, as Myles was saying, that'll lower the cost. Then, we have what we call "emerging" or "newfangled" fuel technologies. So, that is done by these things down here. I've worked on a number of these over the years.
There are things like quantum dot solar cells, there are things like organic solar cells, and then, there are the perovskites. And the perovskites are what we're working on now, and so, I'm happy to answer questions about some of these other things, which I've spent time working on, but nobody really cares. We want you to care about the perovskites anymore. And the reason why we care about the perovskites is obviously, why?
>>Joe: Cheap? Well, [Inaudible], 'cause there are lots of things that are cheap. I heard somebody else say it.
>>Audience: Solution process.
>>Joe: Solution process? That's part cheap. Yeah. That one's cheap. But why do we really care?
>>Audience: Potential high efficiency.
>>Audience: Potentially, really high efficiency.
>>Joe: High efficiency. Not potentially, right? This is as good as any of the other single junctions. So, this will [Inaudible]. Pretty close, anyway.
Yeah. It's pretty steep – that shows good things. Partly, it's as hopefully that we know how to make photovoltaics after all this time of doing it. But – and, I must say, I was super excited to see that both my colleagues showed Venn diagrams. How many people know who Herb Kromer is?
All right. Herb Kromer – 2001 Nobel Laureate for the semi-conductor heterostructure. Right? I didn't include my slides on this, but there's the Kromer [Inaudible] which is "If, in discussing a semiconductor problem, you fail to basically put the Venn diagram up, that shows that you don't know what you're talking about." And I think you've seen this for all the technologies, and this is really, really important.
There is an underlying thing that has been discussed that I want to highlight in this particular talk, and that is the question of what defines the properties of a particular semiconductor. So, we've talked about loss mechanisms in the context of [Inaudible]. We talked about – okay, what is a semiconductor good for? Why is a semiconductor special?
>>Audience: You can tune the conductivity.
>>Joe: You can tune the conductivity. How do you tune the conductivity?
>>Joe: By doping. And what is a doping?
>>Audience: Oh. It's an impurity.
>>Joe: It's an impurity. What is an impurity?
>>Audience: It's a defect.
>>Joe: It's a defect. It is a defect. What other kinds of defects are there?
>>Audience: The bad [Inaudible].
>>Joe: There are the bad kinds. [Laughter] [Inaudible] Well, no. This is important because you have to understand that a lot of the properties we hold dear – they have both sides to them, right? A defect can be a very bad thing in the context of a device. Or, it can be the thing that actually enables the functionality of [Inaudible].
So, that's something to keep in mind. But yeah, basically, what we want to do [Inaudible], electrons, holes. You can do multijunctions to avoid the [Inaudible] loss, but then, we want to extract the barriers out. And, we might do that selectively, so, we might want to tune the details of the interfaces to hopefully make that as efficient as possible. But, at the end of the day, we need efficiency and stability and we need the scalability kind of aspect, too, if we're gonna do photovoltaics.
This is a typical kind of stack. That basically is an environment of this. So, here, I have, basically, the holes collected here, the electrons collected here, absorbent here. Here, we have a very similar thing where I've got my absorber, which we're calling that APBX3 [Inaudible] possible refer to this as the [Inaudible] prospect active layer. We have a hole transport or an electron transport lining the side – the top and bottom of this, and, of course, we can flip the order to this [Inaudible] our needs.
We have a metal electrode, and then, we need some kind of transparent conductive contact on the bottom side. All right. So, perovskites. Perovskites – so, I think the states are something like, 70 to 80 percent of the Earth's crust are just perovskite minerals. Obviously, those are not silicon materials.
We call perovskites "perovskites" and we seem to ignore all the perovskites that came before. We really are talking about the halide [Inaudible] perovskite systems. And more to the point, for the most part – though they aren't always – they are hybrid systems. That is to say, they contain organic paths. All the ones that we've been looking at here [Inaudible] for solar, can be grouped by some combination of these.
The methylammonium is the [Inaudible] side – I've just drawn it as a module, but it isn't – pardon me, as a round sphere. It is a module. It's got hydrogen [Inaudible] off the side. For [Inaudible] and you'll see what this looks like later, but [Inaudible] is the one case of an inorganic halide [Inaudible] for this same kind of application. Lead and tin are the things.
Lead is really the thing. Bromine and iodide are also the primary halides that we're talking about. These are not things that basically appear in nature. They are synthetic. It is noteworthy that Mitzi – Mitzi of CC Guest fame – actually did these materials way back like, in the '80s.
I say that, and I can remember the '80s so, [Inaudible] about me. But his work was for transistors, so, this is stuff that IBM basically invest in. And I'm happy to talk about – there was this – actually, I had this conversation at one point where somebody said, "Well, yeah, David Mitzi did the stuff for transistors and it wasn't sustainable, so, they gave up." Well, that's not the story that David Mitzi tells. But, all right.
Getting back to what is it we care about in terms of [Inaudible] properties, this is a plot of the fraction of VOCsq. So, that is the theoretical limitation for the voltage based upon the bandgap. And you can see here – so, the closer we are to one, the better we are – or the closer we are to theoretic limit. And we've already talked – Myles did a really nice job of telling you about how we arrive at VOC, right? This is a question with the splitting.
Our ability to split the [Inaudible] levels and basically, push them right up to the band difference. So, we have gallium arsenide – again, the material that all other materials wish they were. Then, we have the [Inaudible] phosphides – again, [Inaudible] – and then, we have perovskites. Then we have [Inaudible]. Then we have silicon and then we have [Inaudible].
So, perovskites are very, very close to this, and we often refer to them as solution process gallium arsenide, with a couple of important caveats. In terms of the absorption area, it's also a really strong absorbent material, so we can really make it in a thin film kind of format. So, that's great, too. So, this kind of explains why efficiency is as good as it is. Yes? No?
>>Audience: Not entirely.
>>Joe: Not entirely. What? Okay. So, what have I left out?
>>Audience: Defect [Inaudible].
>>Joe: Defect [Inaudible]. See? [Inaudible] know this. All right. So, I kind of gave you the answer since Myles had already told you what limits [Inaudible] see, all right?
So, what else limits the performance? We've already talked about the defects limiting performance. How do they limit performance? So, Myles started off – and I'm, again, still gonna [Inaudible] with him from the back, because then he can [Inaudible] me later and we'll feel good about it. He said, "Well, yes, there's only, essentially, these radiative losses, right?"
But they're also non-radiative losses in materials in materials that are not good. And –
>>Audience: I'd say fundamental losses.
>>Joe: Fundamental losses. There you go. But, most of the materials have defects and they result in, essentially, non-radiative recombination. And if they don't, then we should be able to push higher and higher and closer and closer to that VOCsq limit. So, what determines whether recombination is radiative or non-radiative?
>>Audience: If there's no [Inaudible].
>>Joe: That's pretty reasonable. So, expand on that. Is that clear to everybody why that is the case? So, if I can do a direct transition, then, I'm free to emit, essentially, a photon – in which case, it's not loss of the same type. Right?
Of course, then again, if I have a trap, then I basically lose two, and then [Inaudible] it'll work out. That is a loss, I guess. But, if there's no photons involved, then, yeah. Right? And photons represent thermal energy in the context of this, which is a loss mechanism that Myles talked about pretty extensively [Inaudible].
All right. So, what types of defects? Surfaces are defects. Grain boundaries are defects. Impurity – that was the big defect.
That was also the key working – what makes [Inaudible] perspective, so to speak. Also, dislocations, vacancies, interstitial – this whole zoo of these types of things. So, I showed you the plot of the VOCSQ versus gap. And somebody's already mentioned this concept of defect tolerance. So, okay.
So, I'm gonna move that again. So, what is defect tolerance? Putting you on the spot. What – so, if I said something is defect tolerant, what does that mean?
>>Audience: The performance is not affected.
>>Joe: The performance is not affected by the defect to a point. You guys are sophisticated. I like it. Right? So, defect tolerance really is a lot about what the functional property of material is and how the introduction of one of these zoo of things modifies that functional property.
Outside the context and the functional property, the notion of defect tolerance becomes a pretty mushy type of thing, right? What does that really mean? So, it turns out that – and that's basically what that test is. I'm not gonna do this, 'cause we kind of covered this. In the context of solar cells, what makes perovskites special is the fact that they're like gallium arsenide but they're solution processable.
So, that means at the end of the day, I'm left with a thin film that is, from a crystallographic perspective, a hot mess. Right? It's got all the whole zoo of defects and maybe some that aren't even up there. It's got interstitials, it's got vacancies, it's got grain boundaries – it's got everything all over the place. But yet, it makes a really, really good solar cell.
And, at least for the case of the prototypical – okay – how many people here work on perovskites? I should have started that at the beginning. Okay. We've got lots of people. Okay. How many people have worked with perovskites that are non [Inaudible]? All right. What?
>>Audience: The [Inaudible] Quazi 23.
>>Joe: Oh. Quazi 23. All right.
>>Joe: Okay. That's a good line to this. Who else? Yeah.
>>Joe: How's that working out? How's the germanium doing?
>>Audience: All right.
>>Joe: Really? How efficient?
>>Audience: Well, right now, it's really brand new.
>>Joe: Okay. You say that like that's some kind of like, free pass [Inaudible]. "It's new, so, therefore, it's okay if it's a crap PV." I'm giving you a hard time. Right?
There are no free passes here, right? 'Cause there's a legitimate question – especially in the context of the things my colleagues just talked about, right? Anybody who thinks that there aren't good PV materials out there that you are competing with in order to justify why it is we work on this stuff is kidding themselves. Silicon is awesome, because we know so much about it and it makes really good PVs even though it has no business doing so. It's an [Inaudible] material.
Come on. Nobody makes optoelectronics out of [Inaudible], right? They [Inaudible] guys that – and gals, for that matter – who, if they have their way, are gonna make it cheaper than silicon. In which case, why would we need anything else? I'm sorry. I don't mean to pick on you.
But this is the thing that keeps me up at night. We are trying to do something different with this technology, but we shouldn't kid ourselves that this is gonna be easy. But this is basically the explanation for the defect tolerance and using that as kind of a prototypical material. Now, it is not clear – so, the general statements that I'll make about this Venn diagram – and this is the data from this particular paper that was done by a colleague of mine, Vladimir Stojanovic, along with some folks at MIT. And it explains kind of why it is that lead is magic in the context [Inaudible] defect tolerance.
And if you are wanting to go away from lead, you should think very hard and very long about whether that makes sense and whether or not you're just making another CVTS. But, at the end of the day, the things that makes MAP-E special is that essentially, the conduction and valance bands – and normally, in most of the compounds and conductors that we talked about, the conduction band is made out of s-like orbitals, right? PV is basically what we find down here. That is not the case here. What we see here is that the lead s-states comprise the top of the valance band and the iodide p-states – or no, pardon me, the lead p-states – are up here.
So, that means that when we basically introduce a vacancy or a defect, the energetics of them are such that it is possible to form states that are in a gap that would be kind of defects that would be viewed as catastrophic in the context of our functional property – i.e. the absorption of light and the long-live carriers [Inaudible] extract in the form of energy. But, most of these are not energetically favorable to form. That is to the say, these are not the things that happen. What happens are these things. And those things are not in the gap, so, they don't basically impact our – the critical thinking we care about for moving forward.
And that's the magic of the halide perovskites. Now, the mechanism by which – okay, but somebody said – who was the person that said, "There are limits to defect tolerance"? Who was that? I know it was somebody. So, there are limits to defect tolerance. Please, expand on that.
>>Audience: Well, I mean – and I'm not as clear, but –
>>Joe: That's all right. I say things that are wrong all the time. Don't sweat it.
>>Audience: It's kind of the – it's at the inverse of like, the dimensional curve where you can reach a point where you basically know like, [Inaudible] defects that you filled all those states – the [Inaudible] ones or –
>>Joe: Yeah. No. I think that's a pretty good starting point, right? The other way to think about it Myles talked about essentially what happens in terms of engineering for [Inaudible] right? So, you make the cell thin so that as you get radiation damage and you get defects associated with that, you haven't killed off your drift so much that you modify the device. So, the same thing happens in the context of perovskites.
If we have these states, they will scatter carriers, right? They won't create recombination and catastrophic behavior in the same way, but they will have consequences to the material properties that we will likely be able to measure. And those can, and likely do, show up in the performance of the solar cells. So, eliminating defects is still a thing, even in defect tolerant [Inaudible], which, you know, kind of begs this question about – nah, I'm not gonna go there.
I'll be around all week, so we'll have this conversation as we talk through some of the stuff, but the concept of doping a defect tolerant material is kind of a – right? It's kind of weird, right? There is this whole question about – okay, giving that dopants are defects, do we want them in a solar cell? I would argue pretty strongly that no, we don't. Doping is what we do to make up for something else that's a problem – some other defect that we don't want to see in the same kind of way.
So, all of this stuff is theoretical. This is some work that we did a while ago where we basically subjected an active layer to prolonged exposure in a photoemission pool. Now, how many people are familiar with photoemission. Ooh. I did not prepare any slides. Okay. So, photoemission's [Inaudible].
>>Joe: Oh, perfect. Great. All right. Point being – it gives me information about the local chemical environment of atoms. So, I can basically look at what the local chemical environment of all my substituent atoms look like.
And I can essentially look at it and know what the composition of my material is. So, we essentially expose our active layer material for a prolonged period of time to extract radiation, and what we found is that over time, the composition changed. The lead to iodide ratio changed. And it turns out – [Inaudible] lab scale – it's a little bit difficult to look at the organic patterns, but we can do that. And so, what we have is the material's [Inaudible] from methylammonium lead iodide into lead iodide.
And we can knock that in real time. Now, the thing that's staggering is that this is basically the – so, it's MAPVI3. So, that would be the stoichiometric compound. We started out with a material that was sub-stoichiometric, then we took it even farther sub-stoichiometric, and this is basically a plot of the valance band of this material. And what's striking is that the valance band is not really evolving until we hit this 2.5 kind of value.
So, that means that on every kind of unit cell, we've essentially created a single defect. And yet, we haven't basically disturbed the valance band. Now, that's not to say we've done anything we haven't in the [Inaudible] band. Probe what's happening in the [Inaudible] band. But, this is a direct experimental verification that this mapping system is defect tolerant.
Now, the other thing you can do – and we heard about some of these compounds from some of you guys – is we can alloy these. And alloying is special because now we begin to become even more III-V like. We can tune our platonic gap – and here, you can actually see what these – the CSM is just an atom. This is methylammonium; this is forming [Inaudible]. This was reported by Jyles Eckeron.
So, some of you will be working with him on making some devices later this week. But, you can see that we can essentially change the wavelength and change the absorption offset, which means that we can make dams.
Not only that, Jyles reported – I didn't cite it here, but reported in the – I can't remember; was it Science or Nature? Some place. Some place that basically, you could come up with the ideal gaps – at least two-junction tandem – from just this [Inaudible] you can do that. Now, granted, both of those cells are worse than – yeah. Whatever.
Just 'cause you got the gaps doesn't mean you got the device. But, having said that, people are working on that part, too. So, that means that not only do we have here – and we can compete with other thin films – we have access to this part of the chart, and so, we're coming after Myles, but at hopefully, much lower cost. And this is the statement – "Why does the world need another PV material?" So, why does the world need another PV material?
>>Audience: The main thing that I see is that most of the III-Vs and the silicon are all plain [Inaudible] materials. They're all restricted to that kind of form factor, and the perovskites – you have opportunities to change that.
>>Joe: I like that. I like that answer. Anybody else got any other answers?
>>Audience: It allows every country to make their own PV.
>>Joe: Yeah. It could allow every country to make their own PV. And the kind of underlying thing with that type of notion – I mean, these two ideas are very connected, right – this essentially freedom from existing kind of form factors and what it might mean in terms of adding manufacturing and changing, at a fundamental level, the way we think about PV at scale. At least, those are the kind of things that I get excited about. So –
>>Audience: You know, I think the other thing is that simply – I mean, if you look at a little bit wider view, a new material that has different tolerance – I mean, this is maybe bigger than PV in some sense. It's LEDs. It's many different things that they [Inaudible] manipulate on such a scale. I mean, that's a – it might be not an answer to that question directly, but it is, I think, quite informative.
>>Joe: So, I do – I have a slide that captures – well, I guess it does a little bit. I say this a little bit. LEDs, displays, QC – do people know what QC is – Quantum Computation? That kind of stuff. Right?
So, my observation is the following. Any material that's worth actually spending a lot of our R&D dollars on should be good for more than one thing. That is true of silicon; it is true of [Inaudible]. I would argue that it has been true, to a lesser extent, of even CATO, right? CATO was in a material that was interesting in the context of T6 materials for blue lasers before gallium [Inaudible] came in.
I think that the challenges for some of the other thin films are how do you basically get to the place where you can impact more than just PV? Because, at the end of the day, nobody makes money in PV. Right? We save the world with PV. We can add value with PV to other products.
We can change the world with PV. But, I have yet to see a really, really strong business case for making lots and lots of money with PV. I could be mistaken. That's why I'm a scientist and not an economist.
>>Audience: It's the [Inaudible] of things.
>>Audience: Like, Internet of Things, several things –
>>Joe: Yeah. So, that's this very different business model. And whether or not that's in – the economics of that really kind of work, I think, is a – right? But then, you're selling your widget. You're selling whatever it is and you're basically putting a solar cell on it because the solar cell is so dirt cheap, it didn't cost you anything and it saved you some energy.
Right? But yeah. So, this is why we think – and then, this kind of discussion is probably more interesting than any of the text I got on the slide, but it's generally this idea that there's this opportunity to basically change the world. Not only that, because of the aspect – and Paul mentioned, I believe, in his first talk – that we do have the opportunity to make existing PV technologies better. Oxford PV just announced a 27 percent silicon-perovskite tandem.
That is better than what you can do with silicon by itself. Now, on the one hand, there's a very different kind of – I mean, there's still some challenges sort of. So, I made the case for why perovskites are awesome. I hope. And I'm preaching largely to the choir, so, that's always nice.
So, what's the problem? Myles?
>>Myles: They don't last that long.
>>Joe: Yeah. Okay. Ooh, sorry. I forgot. This is Dack, who's just graduating.
This is like actually doing role to role. So, this is a 40-meter substrate across flexible glass substrate. Yeah. We talked about this, right? Gonna integrate it and tense, enclose, just where – yeah?
>>Audience: One slide back. It looks like you need to print like, 3,000 days to get to [Inaudible].
>>Audience: So, we'll be printing all the time.
>>Joe: Yeah. And so, it turns out, if you actually do the calculations – depending upon the details – this is one calculation which was done by a colleague of mine. I did another one. So, if you look at the total volume of the top like, 10 worldwide publications, yeah, you can – newsprint is one thing to be produced at the kind of volume that would be comparable terawatts and it seems like we should be able to do it. There are things that are manufactured more graphically, but they tend to be, I believe, limited not by supply, but by demand.
Like, saran wrap. You only need so much, I guess. I could be mistaken, but that's my understanding. But the thing Myles' mentioned – they don't last very long. How am I doing on time? Am I in decent shape?
>>Audience: You got a few minutes.
>>Joe: All right.
>>Myles: That's longer than some of the perovskites last.
>>Joe: Maybe your perovskites, but not mine.
Yeah. It does no good. So, fundamentally, you have to have these three things – and I said this on one of the previous slides – but, if you don't, you don't have really a PV cell technology. You just don't. You've gotta have all three.
Given where these are, we definitely got this one kind of covered – at least at the level that, okay, we're in the game. This was the problem for organic photovoltaics. To a large extent, it still is, to a certain extent. Making progress there, but not nearly as rapidly. And it's not obvious that a single junction kind of OPV will get you to where you need to be.
The scaling thing – this has be kind of the – people worry about this in the context of things like thin film – mainly CATO. Now, on one hand, you can't argue the first solar's doing good. They're clearly making a lot. What else can do we use _____ for? That's for some, again, economists to basically figure out.
But, at a basic level, this is also one of the challenges for – but the III-Vs can – for silicon – is this Cap-X kind of argument, right? A silicon plant costs a lot of money to stand up from nothing. And that's a lot of investment. And again, depending on how want – one has to always be super dodgy about – I can tell you some of the calculations we've done; I can also tell you that on one hand, I'll tell you them like I believe in them, and on the other hand, I'll tell you that I know that the assumptions of those calculations are probably sketchy. So, one has to be super careful with a lot of the technical economic analysis.
This is also, though, challenge for III-Vs. How do you become kind of cost competitive? Nobody wants to spend more on their solar cells than they're spending on in their house. That's tough. But both of those technologies benefit from the fact that they have been deployed and they're out in the field, and you can't make an argument that they can't last though.
Having said that, one should not presume that all of these technologies are, in fact, stable. Because to a certain level, they aren't. They simply aren't. Or, they're stable because of [Inaudible] unit put in. And, as we push to this and reduce the cost of packaging in the module, we also open up ourselves to questions about how stable those systems then are, in ways that may not kind of be appreciated. Right?
So, on the one hand, we have these cost drivers – gotta lower cost, lower cost, lower cost – but, we may be introducing instabilities in that, given that a lot of what we've done has been engineered away from kind of environmental [Inaudible]. All right. So, in terms of what we do here at NREL, we are focused on the stability and the scalability with the notion that we have to be at some level of efficiency in order to do it. We can make pretty good devices here, but we are not actively kind of chasing the record efficiencies. And to the paper – we've done a lot of papers.
This is one I mentioned before. This was one that was really kind of instrumental in identifying schemes to basically move away from methylammonium lead iodide and into other systems. And this is the first kind of stability assessment that was published kind of early on and showed that the interfaces are actually a primary source of some instabilities in devices. This is some more recent work that we've done in terms of looking at processing to scale, and this is some work – so, I don't have 1,000 hours, right? I only have 10 minutes now.
So, I won't take the full length of my perovskite solar cell stability for the rest of this talk – although, I'd love to do that. This is work that Jeff really did wrapping up, and so, I think it's –
>>Myles: Is that 1,000 hours measured on sun? Like, how is that measured?
>>Joe: Yeah. So, that's basically – for technical reasons, we're not quite at a full sun. We're at .77 suns. It's essentially continuously power producing or actively being swapped. And it is with no encapsulation. So, open air.
>>Myles: That's not an accelerated test. That's an actual –
>>Joe: No, it's not an accelerated test. Although, we would view it as such in the context of oxygen and water are both known degradation mechanisms. When we – I think I have some data, but I don't have the full set of data. We have just about 5,000 hours on under light, under power-producing conditions for devices that were kept in nitrogen. Right?
So, we would view this to be kind of like a 5X acceleration factor being in there. Having said that, I would argue that – you could argue it's 5X – I probably wouldn't argue it's a 5X is probably what I should say. So, there's certainly some acceleration there, but it's not – we don't have mechanistic insight into it and so, exactly how that translates across architectures is something that I think one we'll need to worry about a little bit.
>>Myles: No. You didn't measure it for 200 and then project it to 1,000.
>>Joe: No, no, no.
>>Myles: That's okay.
>>Joe: Yeah. So, I'll show that data, but keep me honest. It's a full-time job. So, this was when people first measured the silicon solar cell, [Inaudible] my colleague, Joey Luther – limitless energy from the sun. So, is it obvious to everybody how one should measure the efficiency of a solar cell?
I mean – so, the idea of a solar cell efficiency test is to get an assessment of when we put the thing outside and put it on sun, how much power are we getting? But everybody knows, who's tested solar cell, that we don't do it at kind of ambient temperature or pressure array. I mean, we do for modules – we put them on their outdoor test facility. But the outdoor test facility here in Colorado is different than an outdoor test facility will be in Tampa, Florida or London, England. So, it's worth thinking that it wasn't obvious, when we started on solar cells, how to test for efficiency and that we had to basically develop standard test conditions.
And so, the same thing, I would argue, is true for stability. Now, perovskites, given that they were catastrophic, made us – or placed demands, rightfully so, that we begin to get some way to assess stability. But it's worth asking the question – so, again, what is our goal with stability [Inaudible]? The goal with the stability study is to determine how long this thing can produce power. And then, we can ask subsequent questions about, "Okay. Well, am I interested in doing that in kind of allowed environment to understand whether or not just simply illuminating these materials is gonna be catastrophic?"
The [Inaudible] of that is – you're not gonna have a greater PV. Or, we'll put it another way. The other way people kind of talk about it is that you can encapsulate away a lot of things. You can't encapsulate away light and you can't encapsulate away heat. Right? So, those are things that we have to maybe consider.
But the point is, we really want this [Inaudible] over time. Yeah. Exactly how you do this is not straightforward. We just submitted a – and I was just talking to Henrietta Grevens, who does a lot of our module stability kind of testing, about kind of our perspective on what it is you need to do in terms of stability testing. And – I don't do that.
I won't do that. Where's – did I skip this? Yeah. I'm gonna do this one. We'll go back to some of those.
But really, you have to do these two things, and if you don't do these two things, you don't really ever know what the requirements are for this one. Right? If we don't have a material that's fundamentally stable under illumination, again, that's gonna be a problem. If it's fundamentally [Inaudible] unstable under kind of operational temperatures, that's gonna be a problem. Now, granted, operational temperatures on Earth are very different than they would be in space.
But just because you have a stable material does not mean that your device will be stable. Because when you've made a device [Inaudible] introduced a lot of other interfaces within the system – a lot of other potential defects. And those potential defects are another route, kind of, for things to go wrong. So, we do think that you have to do this and then, sooner or later, yeah, you will have to do this. Fortunately, this is what you've gotta do if you want to make money in PV and I've already let my feelings about that be known.
And so, we really kind of focus on these two things. So, all right. Now, I'm gonna zoom back down here. All right. So, these are our typical device stacks and typically, we are in what we call an NIE structure.
So, we have the electrons extracted through the transparent [Inaudible] and we have a hole transport material, which is often times, for these materials – or for these [Inaudible] – and we'll show you why, explicitly, that we don't like that. And then, gold or silver is often used. Those are bad. We like molyoxide aluminum. And again – I'll be here all week, so, we can talk about why that is given how much time I got left.
But yeah, these interfaces are a big concern for us. So, this is some – these are some diagrams of some stuff you'll see around the lab. Basically, this is a common [Inaudible] degradation setup. So, it allows us to test a bunch of substrates, each with a few cells on each substrate, and we could essentially run them under load, expose some light producing power over long periods of time. And in this case, there's just enough in the ambient lab.
They aren't quite at – actually, no. These are considered [Inaudible] for the experiments I'll keep talking about. The cells are all kind of held at 25C and that is a pretty important thing, right? 'Cause at elevated temperatures, other things go bad. We also have this flow cell geometry.
So, I mentioned – and I'll show you a little bit of the data – where we basically did – we eliminated oxygen and water and what that did to the stability, but, of course, we can also use these to essentially add oxygen or add water and hopefully separate it out – the impacts of those two things – differently. So, this is this early report in '16. So, this is an eon ago in perovskite years. But here, you kind of see the fact that when we had the [Inaudible] a lot higher, the devices would [Inaudible], and when the relative humidity was lower, it was basically more stable. You'd see the [Inaudible] of the JVs, but the key thing that we had done – and this is some EKV that basically really shows that we had a little change in the active layer happening because we basically were left with just a [Inaudible], which is the in [Inaudible] that I had in end state.
But yeah, the small asset aluminum contact actually is significantly more robust than gold or silver. Again, in place of silver, there are lots of reasons not to like it from a cost perspective and the interactions with halides are not good. Gold is equally problematic from a cost perspective, but this is often viewed as inert. It is not. There's been some nice work out of EPFL that showed that [Inaudible] is a huge problem into the halide active layers.
So, we like our molyoxide aluminum, not least of which is [Inaudible] cheap. Here's some photos so you can see kind of how things bleached. Yellow is obviously bad. Dark is better. So, again, the other thing that's kind of neat is that if we use this molyoxide, we separate out the role of the electrode – yeah, the metal's work function and how the device operates in some nice ways.
So, in this case, we still had essentially sprio-omitab in here. And spiro-omitab is this organic molecule that then you dope with lithium. How many people – everybody knows about lithium ion batteries. How many people have worked with lithium? Goes everywhere.
It's a mess. So, not only that, it's been clearly shown that basically, if you put lithium into spiro, it goes all the way down through your device stack. It will modify your charge collection and your lower contact as well as modifying the top contact. We looked at questions about lithium migration as well as other kind of extrinsic ions that we might choose as dopants for some of these different transport layer materials, and it's a huge problem, especially if you aren't cognizant of it, because it's gonna fundamentally change what the Venn diagram for your device looks like under operational conditions. And, if you know about it, then, you can kind of engineer around it.
If you don't know about it, well, you're kind of hosed. So, we tend to think that lithium being around is just not a good thing. It's also hydrophilic, right? So, it tends to grab up water, which we know is bad. And so, we basically had access to this module and these guys here had their devices that were like, eight percent efficient.
So, the delta here is not too good. It's not too large, so, that's nice, but these are not the kind of efficiencies that you can get things published in the high impact journals with. But these are pretty noteworthy kind of people in the field of perovskites. And so, we thought to ourselves, "Well, okay. Not everybody's on a good device case, so, we'll see how we can do."
These are our devices. And so, here you can see that when we optimized the [Inaudible] 44, which is this module by spiro, it doesn't require lithium for a dopant. We're actually able to get a device that's a slightly higher performance level than the spiro and [Inaudible] control. So, that's nice. So, we've eliminated this thing that gathers water, and we've replaced it with something that does better is hydrophobic – so, it keeps water out.
So, we're like, "Okay. We've solved the stability. We're gonna put it in a device." So, okay. I don't know what's going on here.
That's not good. And then, "Okay. So, this is better than that. But only just." So, what about the titania interface?
Titania is used as [Inaudible] to great whites. What else is titania used for?
>>Joe: Sunscreens. Yeah. So, what does that mean about titania? [Audience makes inaudible comments] It absorbs UV. Yes.
I heard UV. Okay. And not only that, it's self-cleaning. It's photocatalytic. So, it absorbs UV and it will do [Inaudible] on its surface.
So, a colleague of mine, when he told me he was working on plastic solar cells – this is before I came to NREL and was appropriately educated – I was like, "Why on Earth would anybody make a plastic solar cell?" You put a plastic toy up in the sun, it bleaches. So, why would anybody do that? But now, you're gonna put it in contact with a photocatalytic material. You're gonna put organic stuff – i.e. the a-site – in contact with photocatalytic material.
Nothing good can possibly come of this. Right? Stands to reason. So, it does stand to reason, but you've gotta have proof. So, we used a technique called Topsense.
How many are familiar with Topsense? How many people are familiar with SIMS? So, it's Secondary Ion Mass Spectrometry. So, basically, what we do is we bring in bismuth, we knock stuff out, and we measure what the masses are. And the Topsense is especially nice, because we can basically see the whole [Inaudible] without fracturing it.
So, we can actually see it come out. 'Cause often times, if the energy's too high, basically, it will blatant the material apart – which is another thing that I'll maybe mention right at the end. So, there's a lot of stuff going on here, but point being – we can basically look at all the substituents in our film as a function of depth and where it is. And here, I've isolated the a-sites and the iodide. I don't know why we have the iodide.
But, the point is, the – I believe it's the yellow curve – I should have kept that on there. Yeah. Okay. The black curve is basically when we first make the device, and the yellow curve is after we've run it for 20 hours. And clearly, on the one hand here, we see that these curves have changed.
Some of them more than others, but yeah, that's a pretty significant change. In contrast, over here, these [Inaudible] are basically lying on top of each other. And here, we've changed from having titania here to having tin oxide here. So, we've essentially removed this photocatalytic thing. We've replaced it with another oxide that doesn't absorb quite as strongly, doesn't have quite the same photocatalytic behaviors, and we're able to improve the overall – well, I've assumed that we've improved overall stability.
But, we've shown that there's some kind of photocatalytic effect. But then, the hypotheses would be that given that, this is the thing that causes the material to be non-stable. I will also point out, though, that we have seen very strong dependencies on the substrate on the active layer, generally. So, what we didn't point out but is reasonable to point out is that we may not have the same active layers to begin with. Certainly, these tracers won't be the same.
So, it could be that we didn't set up the material the same to begin with and so, therefore, it's not the same. Having said that, I have also seen data where they basically really hit the sample with UV and you see this kind of behavior in the active layer and you see it at the [Inaudible]. So, the photocatalytic effect is important. Is it the only thing? I don't know that for sure.
But if it's the primary thing, if I remove it, the device should be more stable. Yes? Oh, you're timing. Sorry, MJ.
>>Audience: That's all up to you.
>>Joe: Yeah. Okay. I'm –
>>Audience: Can I ask a question? So, the perovskites that you're working with are applied before maybe methylammonium [Inaudible] and what not. Can you confirm the stoichiometry with [Inaudible] while you're doing this or would that just account for the changes? Like, if you're talking about really small changes in stoichiometry, does that result in the change in these graphs? Or how do you confirm that you're actually working with the same [Inaudible] as [Inaudible]?
>>Joe: All right. So, we aren't from [Inaudible]. But all of these that we're doing are basically from the same batch of salts. So, there's a whole nother paper on how it is that you need to basically handle the salts in order to be sure that the active layer you're putting down is the composition [Inaudible]. But, that's a whole nother talk and I don't have time.
But, you can see here that we have, in fact, stabilized it. This is 1,000 hours with the data. This is in the Nature [Inaudible] paper. We talked about that. So, Myles, this is for you. This is basically 2014. This is where we're at for stability. This normalized efficiency, but the efficiency wasn't that good either.
>>Audience: The error bars are pretty big, though.
That's why the efficiency wasn't very good. But then, we basically moved away from the spiro hole to the molyoxide spiro. Got better. We went from just pure MAP-E to the FA – what we call the kitchen sink around here. That made it a little bit better, but this was getting the spiro gold.
Replaced the spiro gold with ES44 gold – again, that made it better. Went ahead [Inaudible] for molyoxide aluminum – it's even better yet, but again, have this initial burn in kind of thing. And then, this is basically when we replaced the titania with the [Inaudible]. So, this is where we are now. And we have – and this basically the breakdown.
And, if you look at this in detail, what this tells you is that most of our loss is of EOC, which is indicative, although not definitive, that the thing that we're doing or the way we're losing efficiency is in the active layer. And again, we know that both oxygen and water are bad actors for the active layer. So, we would contend we've engineered the interfaces to be stable and so, any improvements that we make are in the active layer proper, should therefore improve things. And then, this is data – so, this is a simple poll – we have actually better statistics than this – but for data and nitrogen, and this would again, confirm this thing that I just asserted, which is that we know that water and oxygen are bad actors, because when we remove them, we basically flatten this out and we see very little change in the VOC and very little change of efficiency overall. We can do the same thing for other active layer formulations.
Some of them have some other weaknesses, but, for the most part, handling the interfaces is the primary thing we have to do. And then, this is [Inaudible] scaling [Inaudible]
[End of Audio]