So it's a real pleasure to welcome back to I don't know, prodigal son, I guess of Georgia Tech, sparrows politest. An extra I forgot to ask you. Where did you get your bachelor's degree? I was in England at Imperial College and parentage and then got his master's degree also at Imperial College London. Before coming to Georgia Tech where he got his PhD in electrical and computer engineering. And then stayed on for a postdoc and materials science and engineering before starting at NC State, where he is currently an assistant professor in the Department of Electrical and Computer Engineering. He's also associated with a couple of big centers, namely the power America, which is part of the manufacturing USE network, as well as the assist Center, which is an NSF ERC. 2022 has been a good year so far yet, a number of awards including Bennett Faculty Fellow as well as an NSF Career Award. So I think only get better from here. So with that, I'll turn it over to Spirit. Wonderful. Thank you very much for the intro. Yes, I'm very excited to be here. A lot of familiar faces and feels sort of strange for me to be on this side of the podium. I'm used to it was used for a long time to be sitting in the audience. They're eating a slice of pizza or something while watching the stock. So it's nice to be on the side of the podium as well. So yeah, I'm gonna be talking today about the work that I've been doing at NC State for the last few years. One of the main thrust of my group has been to focus on the development of gallium nitride devices. And a big part of this has been enabling next-generation vertical can power devices using selective area doping techniques. So a couple of words, I guess, about my group before I jump into the technical detail. When I moved to NC State, I started a group or a lab and we call ourselves and CSU leads, which stands for laboratory for electronics and advanced devices and systems. There's a lot of words in there, a lot of things that you could, I guess, interpret from that. But this diagram names that kinda gives you a sense of what we're doing. The core of it is development of devices, electronic devices, high performance devices. And you can see that that really intersects with a lot of other areas as well. So as was mentioned, I got my degree in ECE, but I also spent some time working with Eric's group as well in material science. And I think one thing I've come to appreciate very much, very much is the importance of good material to enable a very good device. And so we work closely with material scientists to understand how their synthesis approaches are there. New technologies can enable new devices and better performance in these devices. And of course, all these devices ultimately have to triple the way up into circuits and systems. And so we look to collaborate with other folks on that and also do some basic demonstrators, sometimes with the circuits as well. Big picture, all of this has to lead to some sort of progress, some sort of societal change or improvement. And the context of what I'll be talking about today. It'll be for power devices. And I'm sure many of you are well aware of the importance of power devices and power systems. They really form the backbone of our modern society, ranging from the integration of renewable energy sources to novel transport, to resiliency. And you name it really, it really touches all the different areas of our lives. So what we're really thinking about here is opportunities to improve efficiency, to be proven, performance and reliability as well, and unlock even new applications for solid-state devices to define their way into. So this is an outline of what we'll be talking about today is sort of a high level outline. I'll start off with giving a brief introduction to power devices. I understand that there's a broad community here in the audience. I want to make sure that we're all on the same page about what I'm thinking about when I talk about power devices. And as the title already hinted, that there's a really strong need for selective area doping. And for Gan, gan has a lot of attractive properties that make it a great fit for power devices. But a big technological challenge for a long time has been the ability to dope areas of the device laterally with different selective very doping techniques. And so that's one of the things that I'll be touching on today. The two devices at all focus on to demonstrate that are a gallium nitride super junction device using a structure known as a lateral polar junction. I'll give a bit more information about that later. And then a JBS diode with implantation and ultra-high pressure and kneeling. And so these are different doping techniques, different ways of doping gan devices in different regions and different devices that go along with that. And I also want to highlight some very, very important to people that have been involved in this work. Dollar and Shane are two of my first students. Dollar actually graduated recently. And so a lot of the work that you'll see is his work that he was driving, spending a lot of time in the clean room. Testing devices as well, and Shane has been following up a lot of that work as well. And then I mentioned it earlier, the importance of good material and collaborations with material scientists, slot go sitar and Ramon Cajal AZO in the Material Science Department at NC State have been really key collaborators and the work that you'll, you'll see here and we are groups really overlap quite a bit and work closely together. So I want to make sure that they're acknowledged. So again, if you're in the world of wide band gap semi-conductors and power devices. You've probably seen a chart or a table like this before. But again, if you're not, I want to make sure that you're aware of why we're talking about GAN in this context? There is a few different properties that we can kind of stack side-by-side when we compare gallium nitride with silicon, conventional silicon devices. The first is looking at the band gap. Clearly, gallium nitride has a larger band gap that makes it a little why we call it a wide band gap semiconductor effectively, if you skip to the bottom of this table here, the second-to-last row, you can see a clear relationship between the band gap and the critical electric field. And that is the main property that we are really trying to leverage here as we scale these devices and try to make them more efficient. Now, one of the figures of merit, and this is the ball league of figure of merit. That's a lot of people cite when discussing wide band gap semiconductors. You can see here as a very important and cubic dependence on critical electric field. And so clearly, once you have a couple factors of improvement there in critical electric field, that very quickly translates to a blown-up reason to go after gallium nitride. Now, I would never make the argument that you'll get a 4 thousand x improvement in a gallium nitride device versus silicon. But I think the leap here is large enough that you can motivate the idea of going after this and see something like a Tenex improvement, for example, in performance. More qualitatively, what are we looking for or what do we gain? Well, if you look at a unipolar device or a Schottky diode as an example, what you're doing is you go from a conventional silicon device where this drift region here, this is a low dose and thick region. And the device that has to sustain the electric field. That region can be redesigned in gallium nitride or wide band gap semiconductors more generally, to be thinner and higher doped. And so that ultimately reduces your fifth region resistance. And if that plays a key part in your devices overall loss and we'll really reduce the losses in your device as well. So this has an important outcome. Not only are we improving conduction losses and lowering our on, but it now means that we can think about using unipolar devices like a Schottky diode or a mosfet in place of other bipolar devices. So IGBTs, for example, are devices that are typically used silicon IGBTs. And now we can think about replacing IGBTs with mosfets are other unipolar devices. And that has system-level outcomes as well. So unipolar devices can switch faster and by switching faster, we can miniaturize the size of the system. That then unlocks the opportunity to go after. Very mobile applications are applications where size and weight are big constraints. So this is another way that we visualize this and you'll see curves like this show up in my presentation and a lot of other presentations as well. Basically, we're able to draw a relationship between our on and breakdown voltage for different technologies. When you go through and work through different materials, you'll find different coefficients here. And then this breakdown voltage relationship with our on, with the power of five over two. And that kind of gives you this slope that you see constant. But overall, what you're seeing is that you are shifting down to the bottom right-hand corner of this curve. And clearly for a same a constant breakdown voltage for a wider band gap device, you're able to reduce her conduction losses. So this is really why, again, is exciting. But one of the things that I want to highlight here, and this will be especially important when I started talking about the super junction devices that as it is noted here, the relationship between breakdown voltage and are on is pretty fundamentally you scale it by material parameters like the critical electric field. But this slope is basically the same. And so the super junction device that I'll talk about later, one of the things that we're looking to do there is actually bend this curve downwards and get even more efficiency than lower loss for the same breakdown voltage. So we're trying to go beyond those conventional relationships and rules. So I promised you that I would talk a little bit about the need for selective area doping. And so these are some devices that are very relevant to power. What you have is the JBS diode or the junction barrier shocky diode that I already talked a little bit about my outline. You can see here that this is a, basically a Schottky diode. There's a Schottky contact are here at the top, but underneath that there's a grid of a P type doped regions, as well as these regions off this side. These are floating field rings that are often used as well to improve the. Determination efficiency or basic electric field distribution of the device. And so you can see here that yes, we have n-type doping and that needs to be very well controlled, but we also have a need to make lateral p-n junctions for these devices to work. And so that's an important question about how we do this. Same thing here. You need junction terminations, you need P gate as well. And then another one over here, which is one of the, another device that I'll focus on today, is a super junction diode where you have these lateral p-n junction regions to alter the distribution of electric field and improve the performance of the device. So clearly throughout all of these and even other devices as well, you need to be able to introduce regions of p-type doping laterally within the device structure. And this has been traditionally a very big challenge for gallium nitride. So one of the reasons is that with silicon, one of the things we typically learn when we learn about semiconductor fabrication is we can leverage something like diffusion doping, right? And so that's a very convenient way to get started, but with gallium nitride, that's not an option. These diffusion, diffusion processes are very, very slow, so it's not practical to use them. You can also deteriorate the surface with those high temperatures that are needed. And then another very common practice that's used, especially in cases like this, is etch and regrowth. And that edge and regrowth process can lead to damaged interfaces. And then that can also reduce the performance of your device. So we want to think about other ways of doing that. And I'm going to talk about two of those ways today. So let's start off with the gallium nitride super junction device. And the structure that we're going to use there is the lateral polar junction. I already showed you this chart and I just wanted to remind you, one of the main outcomes here is that there's this fixed trade-off or relationship between Oran and breakdown voltage for unipolar devices. And what I wanna do now is think about how we can break that rule. Build up to that by looking at where this comes from. So in a conventional unipolar device, like a Schottky diode here you have this low doped region underneath your Schottky contact or underneath your anode. And you are distributing the electric field across that. And so what you're trying to do is basically make sure that you do not hit your critical electric field of your material too early. Otherwise you'd have premature breakdown of your device. And the slope of this triangular distribution of or linear distribution of electric field is proportional to your doping. So generally speaking, you the doping and then lower that slope as well. And the design rules that you'll see, and this is pretty generic. The numbers are tuned to Gan here, but the relationships are fundamentally the same across all semiconductors is that as you go to a larger breakdown voltage, you've gotta do a couple of things. The first is that you've got to reduce your doping concentration. I already alluded to that over here. And the other is that you also have to scale up your drift region thickness. So you can see as an example that if you want to make a device that is five KV, for example, you have to be into the ten to the 15 range, which is very, very difficult. You have to have very well controlled growth processes, control over impurity incorporation. And if you're gonna do this, for high voltages, you need to be able to grow layers that are tens of microns thick. And so ideally you don't want to do that with a very, very slow deposition process either, that'll take forever. So we want to be able to achieve low doping. We want to also have low doping in combination with high mobility. If you have a lot of impurities that you're either purposely or unintentionally using as compensators. Those will ultimately, ultimately also lead to a hit and your mobility and therefore your losses. And if you can, you need to think about ways to do this with high growth rates. But this is sort of the challenge that everyone is dealing with when you try to make a unipolar device. So is there another way for that, that alters your design rules, that gives you another way to scale up and make high-voltage devices. And that's where this super junction structure comes from. And this is a little bit of discussion about reverse bias design and behavior. So the super junction, unlike the Schottky diode that I showed you earlier now has these lateral pillars are stripes of p and n regions side-by-side. So when you look at the electric field distribution, it goes from being this linear or triangular shape that you had before. Then I'll having this rectangular shapes. So vertically There's basically a doping independence and it's laterally now that you start to see that triangular shaped come out, we're now doping width of those stripes. These WP and wn layer parameters become very, very important. So we're flipping the structure over a little bit in terms of what we have to think about with doping requirements. And we did a lot of analysis and others have done this as well to understand how these relationships work for gallium nitride. And these are some examples kinda give you a very direct feeling of how the benefits are. So let's take the five KV case. If you're doing this with a unipolar device, again, you'd have to be in the mid ten to the 15 doping range. So very low doping with thick 35 micro meter thick drift regions. Now if you're a super junction, got a super junction device and you can scale this to have. Very small pillar widths. Now you're doping relaxes and you can do this at 317 and a slight improvement or relaxation of your thickness requirements as well. But really the main benefit is, you can see here is almost a two orders of magnitude increase in the doping that you need. So that the relaxation in doping requirements means that we can do this with technologies that are available today and devices can be made from Adobe perspective alone. But the big problem that you'll see here is again, how do I do that with selective area doping, right? So I can make an n-type region that has tended to sit 17 doping, Yes. But how do I make these stripes side-by-side so I can make the super junction work. So overall, those larger doping levels lead to smaller are on and better efficiency. Then the other really critical thing that you have to think about when you're designing and making a super junction is charged balance or the other way around, be worried about not having charge imbalance. Because the math on how these devices work assumes that your charge is balanced between your lateral p-n junction regions. And you can see the consequences of not having that in this chart here. If you start to deviate from a charged balance scenario, you see a roll-off in your breakdown voltage. And the other really important thing to keep in mind here is that that slope or that roll-off is very dependent on the doping level that you're using. I mentioned earlier I want to have a high doping level to reduce my losses. But you can see here that that puts me in a very risky position. The higher doping I'm using, the more aggressive that roll-off becomes as well. So the need for charge balance becomes even more important. So this really places a critical need on doping and then being able to control doping in both of those p and n regions side-by-side. And you can see another consequence of that happening here. If I start to see charge imbalance happen, I start to move more to a 1D case and I can see a very significant reduction in my breakdown voltage. Basically a complete collapse of the super junction feature that you get with these two faces. Now, the other side of the coin is also thinking about how these devices behave in Ford. So with a Schottky diode, one of the great things about it is that you're turn-on voltage will generally be low. It will be dependent on the properties of this metal semiconductor contact and current will flow through that drift region if you apply a large enough bias and you'll be majority carrier flow or unipolar current through there. In this super junction case, you now have two regions and basically two important barriers to think about. One is going to be that top metal semiconductor barrier, which is that same Schottky barrier over here. And then you have another P-N junction down here. So the idea is to be biasing, able to turn on your device or engineer that schottky barrier to be somewhere between that p-n junction below that pn junction and then still get high current flow with low turn-on voltage or low forward voltage drop to keep your efficiency is low as well. We want to avoid getting into the p-n junction regime where we have bipolar action happening. The challenges well though, is thinking about how to engineer this so that your turn-on voltage is not too low because if that barrier becomes too low, then you reverse leakage currents also becomes significant and you have issues with the efficiency of the device and blocking. So also thinking about schottky barrier height engineering is important here and I'll address that later too. So we all show a couple of these charts here just to drive this message home. When we went through and analyze these devices for different geometries and so on, we can come up with these analytical equations to very quickly understand what the benefits of the super junction device are. And you can now see what I was talking about earlier in terms of breaking that slope, that conventional 1D limit is here with that conventional relationship between breakdown voltage and are on. And when we get to the super junction and we see a new relationship, and that relationship has an extra tuning parameter which is the pillar with clearly if I can scale these devices down to smaller pillar widths, I can see even more benefits in my performance. So we have to think about ways that we can control doping, control the incorporation of doping in different regions. But also, can we do this in a way that can be scaled down to very small pillar widths. If we do that, then we can improve the performance of our devices. And so we're looking at trying to make these one micron pillar widths or potentially even smaller in the future. And like I mentioned earlier, engineering the various parts of the device here. So to put this into context, why I'm gonna, I'm gonna talk about the lateral polar junction approach. But to understand why this is so different, you have to understand how people make these things today and with conventional technologies, namely in silicon, but also some of these things are being translated over to silicon carbide, which is probably gans biggest competitor at the moment and the wide band gap market. So one of the approaches that is used is to do something like a multiple epitaxy approach where you sequentially grow with film implant into it, grow another film implant into it. And if you're in silicon, you can also leverage diffusion of the species to merge all these things and create a continuous PN junction as well. Silicon carbide Afghan don't really have that. It's, you basically don't have significant diffusion there. Well, silicon carbide, actually, I'll show why GAN may actually have a pathway for that later. But all of this requires very well-controlled lithium because you have to do this multiple times, efficient activation of implanted species. And in order not to do this too many times, hopefully the ability to implant in a very deep level as well. You can also do trench etching and re growth at shadowed region and put them in a layer within that. Not only requires good litho, But also requires high-quality interfaces. And we can't do wet etching and Gan and we have to use dry etching. Those dredging processes usually leave behind some type of damage. Those interfaces are going to have defects and then those can be leaky interfaces and then also lead to problems with the device performance. With GAN. It's really hard to look at what people have conventionally done and just do a one for one swap for processes, we have to think about a new way to handle this. So when we were all talking and it's Lotka and Ramon had been thinking about this for a long time as well. There was this idea of leveraging a unique property of gallium nitride, which is the fact that GAN is a polar material and you can basically grow it either gallium polar or nitrogen polar, which basically is, you can see comes about as a result of the differences in the direction of the polarization fields there. And there's a few important consequences of thinking about gallium polar and nitrogen polar. They're essentially the same material, but they actually have some different properties, which almost you can treat them as different materials in a way, from an Electoral Engineering perspective. In some ways they almost behave like very different materials. Because of the difference in the direction of the polarization fields. Actually when you make a contact and nitrogen polar again, you have a different barrier height, so different Schottky barrier height and different turn-on voltage for your diode. So that brings in already hints that we have to think about how to engineer that interface. Chemical sensitivity is very, very different for nitrogen polar can. So conventionally for again, you have to dry etch it and you can expose it to a lot of different acids and it'll survive quite nicely. Nitrogen polar again, is very, very sensitive to acids and bases. I'll show a little bit of that later as well. But to make this clear, is that early way of understanding whether you grew and nitrogen polar or gallium polar GAN, was just to dip it in KOH. And if you started seeing etching, it was nitrogen polar Ganz, so it's that easy to see it. Another very important feature is asymmetric defect incorporation. Gallium polar Gan when you're doing growth. And most CVD, for example, if you can tune it so that the growth conditions incorporated low amount of oxygen, a low amount of background oxygen with my chin polar again, you'll generally see more oxygen being incorporated. So now you can think a little bit about how doping might be different in these two types of orientations. And so that's where the lateral polar junction comes in. We can take a template and aluminum nitride template, pattern it. And then now go directly to a growth step. We can grow gallium polar and non polar layers or regions side-by-side in a single growth step. And in doing that, while doing that, I can, for example, introduce magnesium as a acceptor impurity. And for the gallium polar regions, there'll be low background oxygen. Museum will be incorporated. Lab material or region will become p-type, where the nitrogen polar regions, I'll still have oxygen incorporated in there which will be net and tight. And so I can basically create NNP region side-by-side with each other. And this is really a very different approach compared to what you see in conventional fabrication techniques with implantation and so on. We can do this in a single growth step. And as a result, that also leads to the ability to scale this down even further. We can do a single lithium. You can use something like EBL, for example, in the future and get submicron features and not have to worry about layer two, layer alignment and multiple lithography steps along the way. So we have this structure, we can grow this to some degree. But what is missing here. So we need to be able to demonstrate 317 doping. I mentioned that number earlier to be able to be relevant for kilovolt range types of devices, we need to be able to do that for one micron features or smaller. We need charge balance, which is really important. And we want to think about Schottky contact engineering. So we have low reverse leakage currents as well. So for nitrogen polar shocky began shocky diodes is one of the first papers that we published in my group. We went through and made these end polar. We worked with Ramon as lacO who grew these N polar regions. And then we, I went ahead and made Schottky diodes on these devices. And the goal was to demonstrate mid ten to the 17 doping. We went through when we made these devices, we characterize these as a control on sort of unprocessed, pristine surfaces, if you will. And we set a baseline for very high-quality shocky diodes with low ideology factors. And now you can see turn on or threshold voltage of 0.4 volts. Gallium polar GAN typically has a larger turn-on voltage when you make a Schottky diodes on in the range of 0.6 to 0.7 volts or so. So this is already a little lower and you can see that that will trickle down later when we make these devices for high voltage applications. We also in the same paper did CV. And from the CV we can confirm that with the growth conditions that were used, we now had n-type doping effectively a 317 range. So this was in that range that we needed to be able to make these super junction devices down the line with high voltage behavior. Another important study that we did as part of this paper was looking at the impact of our processing. So conventionally, when you make a high-quality Schottky contact, you want to prepare the surface before metal deposition. And that often involves treating the surface with some sort of acid to clean up oxides or residual carbon, for example. A very common way of doing that is hot HCl path. And you can see here these are SCM side-by-side or different surfaces after different chemical treatments that will immediately lead to very rough surfaces at damaging that interface. Before you put the metal down. What that does is increase your barrier height. And you can even see this coming about if you use something like Tim age, which of course is a common base in developers for litho, right? So we even have to think about how we do our lithography and these kinds of processes. So this is very, very sensitive, right? As far as the material is concerned and the way that the processing is done. So that has impacted a lot of the ways that we've done our processing down the line as well. So we've shown now that you can make a high-quality Schottky contact if you control the surface and keep it clean, a demonstrated relevant doping levels, but also show that there's sort of a low barrier there that could lead to leakage currents. And so we want to increase that barrier height and think about ways to improve chemical and thermal stability. So one of the things that we leveraged for this is making a thin silicon nitride inner layer. And we specifically have been doing this with low pressure chemical vapor deposition. And so this was another study that we published where we deposited these LP CVD layer is relatively thin at seven nanometers and then put down our Schottky contacts on top. And in this case, we also looked at different surface treatments to that silicon nitride prior to shocky metal deposition, one case where we dip the samples in HF to remove oxides are forming there at the surface. And another one where we let the oxide grow after letting it sit at the sample set out for a few days and then put the Schottky contact down. And you can see that this has a very important implication on the device performance. This curve here, the red one is the sample with the oxygen terminated surface with a large turn-on voltage. But if you use HF to treat the silicon nitride before putting down your Schottky contact, your turn on of that HF clean surface is slightly higher than your bear or control device, but you still get high ideality factors. And the silicon nitride is just a thin, thin inner layer there as well. So what did we conclude from this? I'll just briefly mention this. This was an interesting sort of interpretation where we sat down other results in this paper here as well, we did XPS and other techniques, but basically we were able to understand that depending on what the surface termination is and how you treat that surface, the dominant barrier will either be at the surface, which is this oxygen terminated case where you have a large barrier here at the surface, or that surface barrier will be quite small and the dominant barrier will be at that interface below there with a silicon nitride, again isn't. So you have to be very careful about how you treat these samples and how you fabricate them. And then ultimately that will have an impact on their performance as well. Now, not only did we tune the barrier to be larger through the inclusion of the inner layer. But we also demonstrate that we can now go to higher temperatures. And so I'll just briefly show this here. This was a control device that already at 250 degrees was basically resistive. I mean, there's no rectification effectively in this device. But with a five nanometer silicon nitride layer, we can still see a rectifying diode at 400 degrees Celsius. That takes us, that takes advantage of the larger barrier height that we're engineering through the inclusion of silicon nitride layer. And we even showed that if you go to 500 degrees Celsius, your device still survives. It does change a little bit in its behavior. So you probably wouldn't want to do this on an everyday basis, but you can deploy these devices at hundreds of degrees and still see diode like behavior. When we looked at trying to understand what was going on here, we can see that there's a little bit of a change in surface morphology there. And at the temperatures that we're working with here, There's well-documented alloying between nickel and silicon. So we think that we've created some sort of silicide at the interface there. So the reaction is probably more to do with the metal to silicon nitride interface changing rather than the underlying silicon nitrite again interface. So that is all well and good because we can show low ideology factors and engineered interfaces and so on. Because the inner layer is so thin, we have negligible change in the on-resistance as well. But the biggest problem in terms of high voltage is that if you use a silicon nitride layer, ultimately you might. Breakdown happening are bottlenecked by your dielectric. So we were thinking about a different way that we can do this that would really leveraged the wide band gap properties of gallium nitride, but still allow us to tune that barrier height and get better performance. So the idea here is to use what's known as a camel diode. And that's what I'll talk about here. So this is again your conventional Schottky contact or Schottky diode. And the camel diode is sort of a hybrid or in-between state almost of a shocky and pn diode. Effectively, what you're doing is you're introducing a highly doped p, Again layer beneath your end layer and your Schottky contact. But you control the thickness of that so that you are able to introduce this hump in the conduction band and therefore a larger barrier compared to your conventional or reference shocky diode. But you don't go all the way to the point of turning this into a p.sit. And so this becomes a depleted layer effectively. And we did various t cat simulations to look at how this relationship exists between thickness and as well as legal discharge between thickness and doping. So there's two sort of design knobs here that you can use. But effectively this gives you a way to kind of add to your reference Schottky barrier diode. Schottky barrier for your shock that your barrier height for your Schottky diode and add to it via engineering this, this p layer. So that's what we aim to do here. And we design this structure where we introduced the thin peak and layer with very high doping on top of now the lateral polar junction with gallium polar and the nitrogen polar regions side-by-side. And this was meant now to start to prove out the different concepts that we have here in terms of making our way towards the, the super junction structure. So you can see here a cross-sectional SCM and you can actually make out the various regions of the device. The gallium polar region is here and the nitrogen polar region is here. So we then went ahead and analyze this in a few different ways. The first is we did what people normally do when they wanna make sure they have nitrogen polar gamma, which is we did this in hot KOH. And indeed we can see selective etching. So we can see that we are removing these pillars where the nitrogen polar again was grown. So you can see a periodic arrangement of these nitrogen polar and gallium polar domains side-by-side. If you zoom in and do a TM on this cross-section. At this scale, you can see that there's a very abrupt interface between nitrogen polar and gallium polar as well. So these are atomically abrupt inversion domain boundaries as well. I mean, full disclosure, I would say we still have some work to do at a macro scale. You can see there's a bit of a roughness there along that edge. But at this scale here, there's good control of these inversion domain boundaries. Then we wanted to think about charge balance or impurity incorporation. And so we did a scan here to look at depth analysis. And we tracked magnesium through this structure here. And you can see that there's a 118 level. Keep in mind that when we're sampling the Sims here, we're doing it across a large area. So it's capturing impurities that would be both in the nitrogen polar regions and the gallium polar regions. So we're calculating, we're seeing about 118 and these regions. And then we're also seeing about to 18 from an oxygen. But we know that there is very low oxygen coming from the gallium polar regions. So we can attribute the magnesium from both regions and the oxygen from only the nitrogen polar region where it is predominantly being incorporated. When you go through and do the math on that, you basically now see that you have this charge or impurity balance of 2018 when 18 n-type and p-type between the two regions here. So this was very, very happy result for us to see that the doping is being controlled well at these dimensions. And then we're able to move forward here with characterizing this device. Now you can also see that for this structure, we did not hit the low one is 17 range that we want it. So there's clearly work to do to not just control the charge or the impurity incorporation in both regions, which is what we're achieving here. But to do that while also reducing the doping, so we can hit low doping when we're only growing niche and polar again. But the moment that we tried to put these two together now it becomes a more challenging proposition. So that's one of the big areas that we're all working on right now. As we've worked with the material science team. Of course, I'm an electron near, so I wanted to take this device and make something with it are these layers and make a device on it. So he made Schottky diodes on these. These are the I-V curves that you see here. You can see that we have a larger turn-on voltage greater than a volt, which we tribute to this camel doubt structure that we have here. We also did some simulations to try to understand what combination of thickness and doping best matches that. And you can see here that it's not. What we found is the best fit as many 518 an a and 15 nanometers thickness, which is different than what we intended up here. So that also leads us to now think a little bit more carefully about how we control the growth to get abrupt interfaces for those of you that are familiar with gallium nitride epi, especially in MO CVD. You can have magnesium carrying forward to these layers. So it can sometimes be hard to create an abrupt PN junction. And we think that some of that is probably feeding into why we don't exactly get the 20 nanometers thickness or that week we anticipated. So this was one important result for us to understand what was going on in these devices. And I will briefly show this so I make sure I have enough time for the next section. We also did CV on these devices. And from the slope of the CV, we're able to go and back out what the doping concentration is inside here. And now electrically, we can also see that we've got 118 doping in these regions, which is an agreement with the sims results that we have as well. So we have a couple of ways of showing that we have consistency of doping incorporation within these structures. So this stove cv analysis matches with Sims as well. So this is a very unique approach to making these lateral, to making these super junction structures with these lateral polo junctions. Up until now, we have not been able to demonstrate charged balance between these. And so that's a major, major development in this area. As I mentioned earlier, when we're just growing nitrogen polar again, we can see the Mideast are low 17's doping. We're working to try to get that alongside charge balancing in the lateral polar junctions. And I've also shown you a few different pathways to be able to engineer these Schottky contacts to tune the barrier height and reduce leakage currents and even pathways to making high temperature devices as well. These are some of the areas that we're currently pursuing to take this and make it a more practical implementation of these devices from increasing drift region, decreasing doping, getting better control these interfaces for the camel diode design and then scaling these things down to sub-micron pillars. And of course, you may have noticed there's a lot of what I showed them. This particular case is on sapphire substrates. These are all quasi vertical devices. We'd love to be able to move this over to gallium nitride substrates for truly vertical devices as well. So this, this is a very active area for us. But I think a lot of room to grow as well. I'm going to switch gears to a different approach to selective area doping and a different device that we're thinking about as well. So this is work that we've been doing to leverage magnesium implantation for p-type doping and GAN, the formation of JBS diodes. So if you're, if you look through the literature, what you'll find is information like this about the challenge of doing selective area doping with ion implantation and can effectively, the problem that you have is that you will implant and then after you implant, you will have generated a whole bunch of damage in your material and you need to come back and anneal that material at high enough temperature to repair the crystal and activate the implanted impurities. Now for something like gallium nitride, the temperatures needed to repair the crystal and activate those impurities are very, very high there above 1000 degrees Celsius. And at those temperatures, the material actually also starts to decompose. So now you have this problem that you can't just throw it into a conventional RTA, for example, and just activate everything and start moving forward. So one of the things that people have done is to look at the use of different capping layers, effectively, materials that you can deposit on top of the semiconductor before you do the anneal to hold the material together. But the main challenge that has come as a result of that is that at these high temperatures, the capping layer and began actually start to interact and it becomes very difficult to remove that capping layer and leave behind a clean surface to which you can make high-quality contexts. So we've been looking at a approach to doing this with that does not involve a capping layer. And the idea behind this is to use ultra-high pressure annealing. So instead of relying on a capping layer to hold the material together, let's heat this material up in the presence of a very, very large nitrogen over pressure. And we started off by doing this one gigapascal, so very high pressures, but more recently, scale this down to a few 100 megapascal. Now, if you're involved in this area within, you've built tools before, you probably realize this is a gigantic number. There are very few places in the world that will do this for you and we're lucky enough to have a collaboration with a team at the high pressure Institute in Poland where we send our samples and they have a system set up for being able to do this effectively at these high temperatures, we can go to up to 1500 degrees at high pressure and prevent decomposition of the can. And then look at recovery of the material and activation of the implant and magnesium. So for this first experiment, what we did is we implanted the magnesium. And after we implanted the magnesium, we went and Neil did one giga Pascal and nitrogen over pressure at 30 degrees and relatively long time as well, about 100 minutes here as well. And this is some materials analysis on these structures that we did along with flood going Ramon. So importantly, I mentioned I don't want the surface to decompose. You can see here AFM morphology and roughness is similar before and after that anneal. So we're holding the material together. And XR D, If you look at this as well, you can see that when you do the implantation and do the XR Tea directly afterwards, you had these peaks that are indicative of damage in the crystal, but after the high pressure, Neil there, you're able to remove those peaks and return back to the high-quality again as well. Now, Sims here is also something that's very interesting, especially from an electrical design perspective. You have this very sharp magnesium implantation distribution here, which is this is Sims capture on the sample right after implantation. And you can see that we design this for a shallow implantation depth. We basically we're trying to use this for contact layers. But when you do this high pressure annealing for the temperatures and the time they were doing it, the magnesium actually redistributes, it diffuses through the material. This is actually quite interesting because I mentioned earlier that wide band gap semiconductors don't typically have. We don't typically think about diffusion of these impurities. But for GAN, and especially in the case of magnesium that's very different from silicon. Magnesium will move on you quite a bit. So you have to think a lot about that when you design your devices. And I would also argue that there may be opportunities here to leverage this for device fabrication and the future of doing similar things where people did with silicon that we maybe weren't thinking about when they move to something like silicon carbide. A little bit of Hall here, I'll just point out that once we do this activation step, we're able to see high hole concentrations as well as high mobilities. And if you work through the literature and these numbers in this chart here, you'll see that the implanted began layers have mobilities on par with what you get with epitaxy, so we're able to achieve high-quality material there as well. So again, I want to see a device, what I showed you previously as a lot of the material science behind it, but I wanna make a device. And one of the first devices we made was actually just trying to make a pn diode. So we did this now in a conductive gans substrate, implanted magnesium, annealed it and made it a truly vertical. Now can diode where we're making it backside contact. Another feature of this design here is that we created beveled edges. These are low sloped edges of the device. And if you go through the design and engineering of these types of devices that helps distribute your electric field and bring it in from the surface so that you can really leverage the bulk critical actual field of Gamma. And so this was reversed bike bias characteristic that we got for this device. And you can see that it's breaking down at about one KV. So this was one of the first devices that we were able to make very, very happy with this result originally, but we knew that there were a lot of problems. And what I'm not showing you here, for example, is what happened when we make Schottky contacts on these samples that went out for annealing. So we had good breakdown voltage, but it's the surfaces for the formation of contacts were clearly not ideal for us. And so one of the things that we've been doing over the last couple of years. It's thinking about different surface treatments that we can use to improve the performance of these devices. And so before I get to that, I just want to show that my student did some modeling here to try to understand what the distribution of the electric field is. He also tried to match the simulated distribution of magnesium to what we had through sim. So it was kind of a better representation. And you're able to see that we're getting critical electric fields that are near the ideal in that three to 3.3 range for, for again, so we're leveraging the Gann at its maximum here. So this is very recent work. It's actually just been published. And over COVID, we were really big heads up against the wall trying to figure out how to make good contacts to this material, right? So not just showing that we can deal with high electric fields and large reverse bias, but also forward bias characteristics being a very good as well. And so what we did is we went ahead and wanted to demonstrate this through our target device, which is that JBS diode, where now you are creating a grid of implanted p regions. We mask the sample off with resist. Basically send it off for implantation, strip the resist and then send it off for ultra-high pressure and kneeling in Poland, you can see here that we've been tweaking our conditions a little bit. So with a lower pressure here, slightly lower at times as well for anneal. And then we create these devices that have this kind of structure with these P stripes as well. So the best way that I can really convey to you how good the surfaces are now with the improved processes that we have is through the Schottky contact. Because as we know, Schottky contacts are very, very surface sensitive. And so this is a control shocky diode that we make on that same wafer that's gone through all these processes. In this case, it's shocking contexts. There's no implanted region here, but it has gone through the anneal. We can see near ideal Schottky barrier contact Schottky behavior here as well. And actually, one of the things that's particularly fun for us as we worked through and did some calculations for what we could estimate the mobility in these regions, in this and then drift region to be. And we're seeing what we think are very, very high. Vertical mobility is of near a thousand centimeters squared per volt second. So this shows that we're getting low doping 1.3 times ten to the 16. I have separate CV for that. Demonstrate that as well, but I'm not showing here, but we're also seeing low losses which are coming from the high mobility in these films as well. So low carbon incorporation basically as well. This is the CV that we have here to demonstrate the 1.3 times ten to the 16 doping inside here with a low built-in voltage as well. So now you may be intrusted. Okay, well, what's the JBS diode doing? So now we have the JBS style that I'm putting side-by-side with this. And now you can see through this chart here that we have the JBS diode with slightly higher turn-on voltage. And what I didn't really tell you early about why I'm doing this JBS diode is that I'm willing to pay this price of a larger resistance, I should say with button, because what I'm getting in return is a lower leakage current and reverse bias. These are all measured results over here have a schottky barrier diodes are the control one. I have one with an edge termination. I have my JBS out here, and I have another pn data that I created on the same sample. And what you can see here is that schottky barrier diodes, because they're very surface sensitive, have more leakage. Current barrier lowering effects are happening there. But with the JBS, I am in reverse bias merging these P and regions, basically driving the junction below the surface and reducing my leakage current here. And I'm able to show that I can get breakdown voltages on par with a PN diode on the same wafer. So where do all these things stack up? If you look at what's been published in the past, we have now demonstrated for gan devices, records are on back to 900 volts here. This little unfilled arrow is sort of what we think is also maybe a path forward here. If you go ahead and view thin the substrate as well, you should be able to reduce those losses even further. But even with a thicker substrate that we have here, you put the whole thing together. We still have records are on it and performance here for these devices. So I'll wrap up here to make sure that there's a few minutes for questions. But basically what I'm demonstrating here is our ability to use magnesium implantation in conjunction with ultra-high pressure annealing to achieve p-type doping and Gan, we use that UHV approach to recover damaged and activate the magnesium. And through hall measurements we can show that we have hi p type conductivity on par with at the material. And these low ideology factors that we're showing now are very, very promising. We have really tuned in this process a lot and we're able to get good contacts and good behavior from these devices with, with low resistances that come from that as well, which are really record-breaking compared to what was published up until now. So overall, between the LPGA and the magnesium implantation approaches, we're closer than ever to select a very doping and gallium nitride. If we make these super junction devices scaled up to higher voltages, I think that could be a real game changer in the power field, especially because we can do this at scale and approaches that could really go beyond what silicon carbide is able to do with more conventional cases. And now we're even unlocking JBS styles at high voltages, which I think will be very, very promising for next-generation electronics. And the last teaser that I have here, which is something I skipped over is this question of Kennan magnesium implantation be used for super junction devices? In a sense, can I make these pillars side-by-side and then drive the magnesium down via implant, via, via the high pressure annealing. And now look at an alternative way of making super junctions as well. In a way that would not be something that you can do in silicon carbide. So this is kind of an interesting question that we're thinking a little bit about as we move forward. And these are some of the folks that have funded this research. And if you have any questions, I'd be more than happy to discuss them offline as well. Thank you very much for your attention. I'd be happy to answer questions. Thanks. Any questions from people who are electrical engineers? Not me. Very nice and clear talk. I'm curious about the high pressure or kneeling. Yes. So what is the mechanism that I mean, what changes the activation energy for both dopant incorporation as well as the activation energy for diffusion or both of those changing when you go to higher pressures. So that's an interesting question. I don't think we are changing those parameters as much as we are allowing the GAN to be contained or not decomposed at those temperatures that are needed for the activation to occur. So in other words, if you could, for example, use something like a capping layer to be able to sufficiently control or contained again, at a high enough temperature, I think you might see similar activation energies in something like the high pressure. The only technique that we use or in a captain approach. The problem is that other techniques that people have looked at using for annealing these things, the GAN typically degrades at those temperatures that are still annealing at the same temperature as a reward, he did not lower, but that's, yeah, that's correct. We've we've lowered it a tiny bit from what we did early on that like 1400 degrees. Now we're still at 1300 degrees or something like that. So it's a small reduction, but yeah, we're not using the pressure to come down in temperature. In fact, it's the opposite. We're using the pressure to be able to stay at the high temperatures that we need to activate. Hey spirits. Thank you again for the talk. On the maybe the next to last slide that JBS diode where you have the selectively patterned p-doped here. So you showed the magnesium drive in. Did you design those to account for the spread, the lateral spread? Yeah, that's a good question. We have. We still need better control over that as well, but definitely weave through a lot of experimentation. We have an idea now for a given doping profile, initial implanted doping profile, and then a set of conditions that we use for annealing where that magnesium should finally end up. Then as a result, we can design our device to be able to handle that. So what we've seen in groups out of Japan that had been working very actively in this area as well as that we mainly see diffusion vertically. And then the lateral areas outside of where we did the implantation actually do not allow the magnesium to penetrate into those that efficiently. So it seems to be driven by a damaged initially and that allows quick movement to the magnesium. And then once you fix that or repair that crystal, that magnesium diffusion slows down as well. So it's mainly accounting for vertical movement or diffusion of the magnesium there. But yeah, it's definitely been something we've been thinking about. Any other questions? Alright, if not, okay, got it under the wire. Really nice talk. I also curious. So for the vertical is pia and especially dealing. Templates are nearly what happened between the interface. There is no diffusion. I still really not that clear. You're asking about like this lateral interface here, which all around that device a picture so you have really big. Which picture? Can make the pack more? Yeah, Even more. So you souls the interface are you talking about? Like these pictures? Let's select vertical. One is the PI. The lateral polar junctions earlier part of the talk like this, this stuff here, right here. Yeah, yeah, yeah. So you can see one big as halfway is p or not a half is what happened between the pia and so he's very long interface. Yeah, so let me go back a couple of more slides. Here. Oops, Here you go. So we are doing two things, right? So we are through the initial patterning of that aluminum nitride templates, we are controlling how the growth will initiate and what polarity the GAN will have as it grows vertically from there so we can control it so that it's gallium polar or nitrogen polar based on whether or not that aluminum nitride template is there. And then those gross side-by-side. And you could then after that, you can control based on what you're injecting into the reactor. For example, whether or not you're putting magnesium, whether both will be n type or one or the other will be p or n. Or actually if you put enough magnesium in there, they could even both be p-type. And so we are this paper actually, if you look at the title here, this was a paper that was led by going around on, you can think of it almost like three-dimensional doping control. You can, you can control the doping laterally between those nitrogen polar and gallium polar regions based on what is being incorporated in those films. But by tuning what you're putting into the reactor as you go vertically, you can also control the vertical arrangement of those layers that doping in each of those layers vertically. So we are, as this cartoon shows, here, we are growing these gallium polar and non-polar region side-by-side. And then we're able to control whether each one of them is p or n. I'm not sure if I'm answering the question that I want to make sure that I'm addressing. Yeah. Yeah. So the leaner hard templates are nearly what happened between the pia. Let them know diffusion. Okay, sorry, yeah, yeah. So this sample, we do not do this ultra-high pressure annealing. So these are two completely separate approaches. The ultra-high pressure annealing that I showed at the end of the presentation is something that we're leveraging in combination with magnesium implantation. So that is to repair the crystal and activate the implanted magnesium. In this case, the magnesium is incorporated during the epitaxy itself. And so we follow up with a lower temperature and kneel to drive out hydrogen, which is very typical for MO CVD growth when you're trying to do p-type GAN. But we don't do an ultra-high pressure and Neil, and so we don't expect there to be lateral diffusion or even for that matter, significant diffusion as a result of annealing step. So you may have a sort of smeared profile of magnesium because of other things related to the growth step itself and magnesium being in the chamber and so on. But there's no anneal that we're doing to move the magnesium around after the structure is grown. Does that answer your question? Yes. This is the radical conditioned I see now. So there should be something between the pia and the interface in reality. So what do you mean by something in between, I guess? Yeah, you'll go back to last draft. So there should be some disease, the p-n junction letter law, right? Yes. Yes. So it does affect your device? Yeah. Yeah. Yeah. So okay. So keep in mind that current will flow kinda falling the lowest barrier, right? So these p-n junctions have large barriers between them, right? With, with Gan, your ideal have a three-point for AV barrier there that you have to surmount. These barriers here lets us design, this is the camel dies. So this is a little bit of an exception because it's thin and we're depleting it. But after it climbs through that, it's going to go through the lower barrier pathway, which is down here. So we don't use the gallium polar or this p-type region for conduction. We use this to balance charge, which gives you that difference in E field distribution. Which then leads you to have the larger, larger, lower conduction losses basically in the super junction structure. So if i, in other words, biases device at, let's say four volts or five volts, then I would turn on this pathway as well. And I could inject current all the way through this as well. And then I'd have a p-n junction or bipolar conduction with those. Well, but in this case, I would restrict myself to two volts, let's say, in which case I'm only providing enough bias there to go through this side. And this is just there to passively balance out the charge when I'm in reversed bias. Okay, Great. Thank you. I'm also happy to talk afterwards if I'm not addressing the question outer percent, no good discussion, but I think in lieu of the time we will finish up. And so let's think spirits one more time. Thank you very much.