 Good morning from Stanford University. It's great to welcome you back to the storage and symposium and for those joining us new welcome. My name is will chew on the faculty co-director of the initiative here. Today is a great day because it is the two year anniversary of our seminar series. So as most of you know, we started the seminar series two years ago at the beginning of cove it. And we launched with a really awesome talk by our Nobel laureates and winning him. And it has been a tremendous honor to host the seminar for the past two years. And what a great way to celebrate this occasion by inviting two of my great colleagues from slack national accelerator laboratory to talk about advanced characterization of materials for energy storage. So let's go ahead and get started. Let me first introduce a little bit about slack. Slack is Department of Energy National Laboratory right next to Stanford University that is operated by Stanford. And it is also the home to several unique scientific facilities, such as a x-ray synchotron and a free electron laser. And for the past decade, two of our scientists, Johanna Wecker and Eugene Liu have been developing advanced methodologies for characterizing materials for batteries and other systems, while they operate while they age. And today we're really thrilled to have both of them talk to us about the techniques they have developed, as well as the insights they have derived from these very advanced measurements. This is something that's very near and dear to my heart, the ability to see what is happening has really transformed many fields, for example, medicine, and it is now will on its way to also transform energy storage as well. Most specifically batteries but also extending to other forms of energy storage. So it is really my great pleasure to introduce Johanna first. So Johanna is currently the lead scientist at Slack National Accelerator Laboratory and she's responsible for one of the several x-ray beam lines at Stanford synchotron. And she has been doing pioneering work on a wide range of materials for energy storage from lithium sulfur to graphite to lithium metal, and really combining a wide range of techniques to understand what is happening as the materials operate. And she is one of the pioneering folks in this area, really developing the methodologies and then also applying them to real battery systems. I still remember one of the first images of lithium sulfur batteries undergoing cycling about 10 years ago, which is really blown away on what you can see and that there's been tremendous progress since then. So let me invite Johanna to come to the stage and then she will be sharing all the latest and greatest from her work here at Slack. Johanna. Yes, thank you for the invitation to speak at StorageX and thank you for that wonderful introduction. So today this morning I'm going to talk about the x-ray characterization that we can do to observe battery degradation. As Will said, my name is Johanna Nelson-Wacker, and if you want to learn more about the research that my group is doing, this is my website. You can feel free to reach out to me by email with any questions you have afterwards. So one of the great things about x-ray characterization, which we've been utilizing over the last few years, is the fact that we can characterize across the relevant battery length scales using x-rays. We've done a lot of research spanning full cells, going to half cells, looking at the particle level at individual electrodes, and also looking at the atomic level. And we also, simultaneous to spanning all of these length scales that are relevant, we also want to observe the batteries in as realistic conditions as possible, depending on the technique that we're using. One of the additional benefits of using x-rays to characterize batteries is that we can also characterize different chemistries by tuning our x-ray energy to different wavelengths. So if we want to look at the low Z materials like carbon and oxygen and fluorine, which are really important, for example, in the batteries in the SEI, in the solid electrolyte interface, we can go to soft x-rays, which are low energy, long wavelengths. If we want to look, for example, at sulfur batteries, we can go to what we call tender x-rays, which is a little harder of an x-ray, a little higher x-ray energy. And then finally, if we want to look at, for example, the transition metals in the cathode, we can go to even harder x-rays or higher x-ray energy. And so by changing our x-ray energy, we can change the chemistries that we're sensitive to. And so the fact that we can change what we're sensitive to and x-rays can penetrate a lot of materials that electrons or visible light cannot penetrate. And we have a number of different techniques that span the length scales that are relevant. We have a really powerful tool at a synchrotron to study batteries. And so for the next half hour, I'm going to give you two examples of just two different types of tools that we've used on two different battery types. So the first example I'm going to talk about is the transmission x-ray microscopy or TXM, and that will allow us to see morphology changes. And I'm going to specifically target porous alloying anodes. So how can we make an anode that's better than our current anodes but doesn't have this degradation problem of silicon and other alloying anodes where they crack and break apart after the first few cycles. The second example I'm going to give is to use x-ray diffraction. And x-ray diffraction is a common technique that is used across the field. Many people can do this in their home laboratories. But we're going to do this by narrowing down our x-ray beam and using diffraction as a way of mapping a full cell and to get the microstructure across a pouch cell. And so we're going to be looking at specifically the degradation that has been caused by fast charging. So for the first example, alloying anodes, as we all likely know, alloying anodes such as silicon, tin are much higher capacity, gravimetric capacity than our current anodes which are made out of carbon. They don't quite get up to lithium metal, but they do solve a lot of problems that lithium metal still have. And so with this large capacity though comes this large volume change. So we can get to 300-400% volume change as we're inserting lithium ions and alloying with the metal. And this large volume change especially repeated over thousands of cycles causes cracking and fracturing. It also has a continuously breaking of the solid electrolyte interface. So we need to create an interface that's flexible enough to either go with this volume change or we need to mitigate the volume change. And so this is severely limited, for example, silicon batteries. So my collaborators and I from UCLA, they developed a way of doing nano porosity very simply by selective dealloy. And so you have a parent alloy and you selectively dissolve one of the metals and you come with a result in a coarse structure. And so they've done this for tin in this example, and they have an internal porosity of about 25% and pores that are on 30 to 175 nanometers. And what they found when they cycled over hundreds of cycles is their nano porous tin was very stable over hundreds of cycles, whereas if they cycled dense tin particles, there was no stability at all and it couldn't even cycle. More than 10 cycles reliably. And so we took this nano porous tin and stuck it in our transmission x-ray microscope. And we found that we could see the pores, and we decided that this porous network was visible in 2D. And so we wanted to study it, not just after cycling it, but while we're cycling it. And so we took transmission x-ray microscope images during the first lithiation and de-lithiation cycle of dense tin, which is on this top row, and compared it to this nano porous tin. And I give you just two examples of one particle per cell, but we actually looked at multiple cells and multiple locations across the cells. And this technique is very fast compared to the cycling times. We can actually look at more than 10 areas at one time. And so what we found, if we traced out the area, and we can only say something about the aerial expansion rather than the volume expansion because these are 2D images. And we can track that in time with voltage. We see this large burst expansion in the area in the dense tin at the very end of the lithiation cycle. And you can see that here. So we have this dense tin that hardly changes in volume at all until the very end. And at this very low voltage, we get this large expansion and a crack. The nano porous tin, on the other hand, didn't show much of a total aerial expansion if you track it over time or over voltage. But we did see this morphology change, which was kind of disappointing. And so during the de-lithiation cycle, looking at those same two particles, we see that as we de-insert lithium, we don't return to the original volume or shape of the dense tin. And you can see this in the aerial expansion plot. And we do return with the nano porous tin. But again, we have this morphology change. And so what we found is dense tin is irreversibly deformed. It does not return to its original area. Porous tin, on the other hand, returns to its same size, but it's not in the same shape. So there's this irreversible morphology change that we want to get rid of. And so we took it a non-step further. And so we hypothesized that if we have an inter-metallic with both metallics, both metallics being active alloys to lithium, that lithiate at different potentials, then they can act as the stabilizing agent in the structure while the other alloy is alloying. And so we have this tin antimony alloy. And while we're, for example, lithiating tin, the antimony stabilizes the structure and vice versa when we're lithiating the antimony. And again, we can take this tin antimony parent alloy and dealloy the tin so that we have an equal mix of tin and antimony in a porous network. And so now we have this porous structure. It's got a little smaller of pores, about 20 to 50 nanometer pores. And it is a tin antimony alloy, as we can see from the diffraction, and it cycles very well. And so we took this and we wanted to see how this cycled under the TXM. And so here are just one example again of a porous tin antimony alloy. In lithiation, we see that the particle does grow. There's actually cracks that form. Here's a crack that interestingly forms, and then actually disappears as the particle continues to grow. But other cracks form as well. So there's a few cracks that are forming. But there's a lot less morphology change than we saw in just the porous tin. And during de-lithiation, the particle shrinks again, but the cracks remain. So if we just compare a few snapshots, this is nano porous tin, another example that I showed before, but it shows the same large morphology change, even if we don't have an aerial expansion change. And then that nano porous tin antimony on the bottom, which shows an aerial expansion returning to its original area, and no morphology change other than a few cracks. We can also plot this. The nano porous tin antimony is this brown, and it follows very closely the aerial expansion of the nano porous tin. And so they both have an increase in area, and then during de-lithiation return to merely their same original area, where if you see again, this is just the bulk tin plotted again, we can see that the bulk tin is obviously not returning to its original state. So we have a similar aerial expansion tin antimony alloy that's nano porous, but it's a much more stable particle morphology. So we wanted to look a little closer on what exactly was happening with the nano porous that made it a more stable morphology. And so if we just looked at a small section at the side of the particles, the particle is then thin enough that we can actually see the porous structure. If we look at the center of the particle, because we only have 2D images at the moment, it's really hard to see the porous structure because it's so thick of a particle. But if we look at the sides, we can see this nano porous structure. This is the nano porous tin on top. Below you can see the smaller pores of the nano porous tin antimony. And this is before cycling, after lithiation and then after delithiation. And you can see the pores in the lithiated nano porous tin have expanded. The poor walls actually break. And so it's this nano structure in the nano porous tin that the nano pores that are breaking apart, that's causing the large morphology to change. And so if you plot the average pore area over in lots of different, lots of pores in this region, you can see that as you lithiate the nano porous tin, the pore walls break so you get a larger pore area. The pores are expanding. And then they become unstable as the pores collapse and during delithiation there's just further damage and you do not return to this nano porous structure. The nano porous tin antimony on the other hand is more stable at the nanometer level. And so the poor walls expand and actually get filled. And so we have a contraction of the poor area. And you can see this in this image, the poor walls actually get smaller as the nano porous tin antimony expands into the pores. And then they contract and you see that you retain that nano porous structure after that first cycle. So we have a very stable nano porous structure with this tin antimony alloy. We wanted to see how stable that was. And so because of the way that synchrotron radiation works and how you get beam time on a synchrotron instrument, it's not possible to take the same cell and look at it over hundreds of cycles. And so we took a different cell and at UCLA, we cycled it for 35 cycles. And you can see it's stabilizing. And we wanted to look at, okay, what's the 36 cycle look like? Does it behave the same as it did in the first cycle? And so this is our nano porous tin antimony alloy during the 36th lithiation cycle and then delithiation cycle. So it's a different particle than we saw before. It's a different cell than we saw before. But you can see that, again, we have a very stable morphology. There's very little change happening during the delithiation, I mean, lithiation and delithiation, even after 35 cycles. And if we plot the aerial expansion in the open circles of the 36th cycle, you can see, again, we have this increase in volume or area and then a decrease. But it doesn't increase as much as we did in the first cycle, which is not surprising because every cycle, it has the slow, small amount of not recovering to its original area. So there's less expansion than the first cycle, but we still have a very stable morphology on the macro scale. If we look at the pores, we also have very stable pore structure still. And so this again is just looking at the open circuit voltage before the 36th cycle. You can see the pore structure is still visible. This is fully lithiated as the pores have expanded. I mean, collapsed because the tin antimony is expanding into the pores. And then they reappear as you did delithiate. And so if you plot, again, the 36th, the average pore area with voltage versus the first cycle, you can see the behavior is very similar. There's a similar decrease in pore size and then recovery. And so we found a very stable nano porous alloy using tin antimony. In a side project that we had or parallel project in the same program, which was an energy frontier research center called Scaler, we were exploring an alternate strategy, and that was to do interfacial engineering. And so here we again have this tin antimony alloy structure, but we've engineered an interface that is a business rich and it's liquid like so it allows some slippage between the tin antimony grains. And so you can see this in cryo stem, you can see this business rich grain boundary. And we found that using x-ray tomography using again our transmission x-ray microscope, we can look in 3D this time but it's exit show so we just harvested these particles. We can look at the structure with the business interfacial engineering scheme, and then without it just with the tin antimony alloy. And if we take 2D slices from these 3D volumes you can see there's a few cracks in the business example, but there's significantly more cracks without this enriched grain boundary. And so after 20 cycles, these are after 20 cycles, we get fewer cracks with this interfacial engineering. And so our next step in this project is can we combine the strategies, can we get tin antimony alloy with nano porosity, and this interfacial engineering where we have these liquid like grain boundaries that allow slippage between the tin antimony grains. And so that's what we're working on next in this project. So for the remaining part of this talk I want to switch directions and talk about diffraction and use diffraction to map out the microstructure of cells after fast charging and look at what the degradation mechanism of fast charging is in these cells. So fast charging I'm going to define that as any charging that's faster than 15 minutes. That's about equivalent of about a 4C charge. And so the benefit of fast charge or the need for fast charging is I actually heard this recently at a talk is, you know, people when they buy or they're thinking about buying an electric vehicle. They care about range, but once you own an electric vehicle you care about fast charging. So I really want to be able to go to a charging station and in a pinch charge your battery to 80% in a net matter of a few minutes. But how do we design batteries that work well for fast charging that don't compromise the normal charging rates. So it's believed that lithium plating can dominate the degradation mechanisms in fast charging and reduce the capacity and this is done through a loss of lithium inventory. And so we wanted to use our x-ray techniques to quantify the amount of lithium plating and really make that connection to the capacity loss. Normally, if you're testing cells and you want to see if you've plated lithium, the typical thing to do is to disassemble yourself and look at it visibly. So if you just take an image with your cell phone camera of disassembled anodes, you can see in, for example, a 10 minute charging you have lots of lithium plating on your graphite, whereas in a 15 minute charging there's only a little bit of lithium plating that's visible. And that's not a very scientific quantitative measure of the amount of lithium. That's basically looking at how shiny is our anode. And so we wanted to be a little more quantitative. And so we turned to x-ray diffraction and we thought to do x-ray diffraction because lithium metals crystalline and x-ray diffraction is sensitive to any crystalline material, even low Z materials that are transparent to x-rays like lithium metal. So we take a small x-ray beam and pass it through our pouch cell that we haven't disassembled and look at different spots. And we have some spots looking at the diffraction pattern in 2D, which are then rings. If we integrate that over Q, you can get some small, some spots that have strong peaks at the lithium Q, and that indicates a lot of lithium. And then some medium peaks and then some very weak or no peaks at all. And so we can map this and pseudo color it as a heat map to give us a spatial map of any of the crystalline species over the pouch cell. And here is a typical diffraction pattern of the pouch cells we used. These were graphite anodes and 832 NMC, so NMC 532 cathodes. And so we can see the NMC peaks, they're labeled in green, they're very strong, and we can see the graphite peaks. As graphite lithiates, staged graphite peaks appear and these graphite peaks disappear. And as we plate lithium, we should be able to see a lithium peak, although it is very small compared to these large peaks that dominate the diffraction pattern. And the real benefit of this is it's going to be significantly more quantitative than just visually inspecting whether or not you've plated lithium. And we can do this without disassembling the cell. And so we took seven different cells. Here is the fast charging capacity loss of the seven cells. There were two cells that were cycled at 4C, three cells cycled at 6TC. And here is sort of the details of exactly their charging protocols. All of them were discharged at a slow rate at C over 2. And then we have two cells that were cycled at 9C. And you can see there's a range of capacity loss across these cells after 450 cycles. And it's spread out even looking at just the 9C cells. You get a small capacity loss in one of them and a significantly larger capacity loss in the other one. So there's a large spread in how these identical batteries behave even over similar or identical cycling protocols. And so we took all of these seven cells and I've put them in boxes for their different charging. So looking at the 4C, we mapped them with X-ray diffraction. This is a heat map of lithium. And then after we fully mapped them, then we disassembled them just to prove our point that we could in fact map lithium. So here in the 4C, you can see there's two bright spots of lithium and they correlate very well with what you can visibly see on the electrode. If you plot, if you plate more lithium, these are the 10 minute charging, you get more inflated lithium, but again it correlates, our heat maps correlate with what we visibly see. So we also have the 9C charging and here you can see these are the two 9C cells that charged or had very large capacity fade differences. So we have one that's 10C capacity. It doesn't seem to actually have a lot of lithium plated. We have one with almost 30% capacity fade, which has a lot of lithium. So just anecdotally just looking at these cells right now, it seems to match that the more XFC capacity loss you have, the more lithium plating you can see. So in the next slide, I'm just going to focus down on these two 9C cells. So here again are those heat maps of lithium metal and where it is on the anode. We can also because graphite is crystalline, we can strap, we can map with the same data, what the graphite heat map looks like. And you can see where there's less graphite. And that's a little confusing because you're like, well, where's the graphite going? Well, it's not going anywhere. It's actually no longer graphite because it's staged graphites. It's LIC6. So if we map the LIC6 peaks, we can see they correlate with the lithium. And so these two nominally identical cells look very different, both where the lithium is plated as metal, but also where the lithium is trapped in the graphite. And these are both mapped at 0% state of charge. So 100% of the lithium should be in the cathode. And so what we see, first of all, there's more lithium with more capacity fade, significantly more lithium with more capacity fade, even in these nominally identical cells. And also, the lithium that's plated is co-located with lithium that's trapped in the graphite. And so something's going on that's either causing the lithium to plate where there's degradation in the graphite, or it's the lithium in the graphite not able to escape as you de-lithiate the anode because the lithium metal is plated and blocking. And so this is all just a reminder. This is what I call dead lithium because we're at a 0% state of charge. So we are unable to de-lithiate and strip this lithium metal out. So if we again look at all of those seven cells and we sum all of the total lithium plated, so taking each of those spots in the heat map and sum up the total amount of lithium, we can plot the loss of lithium inventory from the known amount of lithium that we start with before fast charging. And subtracted from the amount that we see plated with capacity loss. And then the loss of lithium inventory that is due to it being, the lithium being trapped in the graphite as LIC6 versus capacity loss. As you can see for the amount of lithium that's trapped in the graphite, that is not dependent on the percent of the capacity loss. So that is independent of capacity. And so we think that is dominated at least by the capacity that, sorry, that the amount of lithium that's trapped just from the formation cycles before we do fast charging. And on the other hand, the loss of lithium inventory from plated lithium is linear with capacity loss. And so the loss of lithium inventory does scale linearly with the fast charging capacity loss. And so that goes hence towards loss of lithium inventory as plated lithium is dead lithium is the driving mechanism for capacity fade in fast charging batteries. Because we have not disassembled the cell before we took these heat maps with defraction, we also are sensitive to the crystalline cathode material. So we can track the state of charge, we can map the state of charge in the cathode as well. And so going back to those two nine C cells. This again is the lithium heat map, looking at the plated lithium. And so we can look at looking at the unit cell volume of the nmc crystalline structure. So as it expands and contracts, as you let the end elitiated, that can be mapped to state of charge. And we also plotted the peak width, because we noticed the peak width of the nmc 003 peak changes. So two things that we found. The changes in the nmc peak width correlate very well with lithium plating. And so this is that second plot here, you can see it plots correlates very well with where we've plated lithium. So we just found that the lower unit cell volume. That means there's less, less lithium in the nmc, likely more lithium on where it shouldn't be on the anode side correlates to more dead lithium. And so here the darker regions here correlate with the lithium intensity of the plated lithium. They're a little more spread out than you have. But there is definitely talk between there's interaction between the lithium that's being plated on the anode, and what's happening on the cathode as a result of it. And so finally, we decided to plot the occupancy of the cathode. And so going to the cathode side we can plot just like we did the amount of lithium on the anode, we can plot the lithium in the cathode. And we found that the lithium occupancy in the cathode is also linear with the capacity that's retained in the cell. And so therefore, all of the lithium in the cathode is active. And that's because capacity fade is the loss of lithium inventory, plus the loss of active cathode material and active anode material. And if we can explain all of the capacity fade from the loss of lithium inventory, then we can say that the lithium in the cathode is active lithium. And the loss of lithium inventory can be a completely accounted for by the capacity fade during, can completely account for the capacity fade during fast charging. So this is just another sort of nail in the coffin, confirming that for fast charging. The loss of lithium inventory due to dead lithium from lithium plating and not being able to strip again is really the driving factor in the capacity fade during fast charging. So to summarize, I give you just two examples of the types of x-ray tools that we have at the synchrotron. We use the transmission x-ray microscope to look at morphology changes. We found that there was stability in being both nano-porous and also having this dual alloy of tin antimony. And we also started to explore interfacial engineering with this binsmith layer in between grain boundaries. And now we're trying to combine those two techniques to get even more stable alloying anode materials. And then we used diffraction to map out the microstructure and we can find that the loss of lithium inventory through lithium plating is the dominant reason for capacity fade during fast charging. And we can look without taking part the cell at all, we can look at where the lithium is in both the anode and also in the cathode. So finally I want to thank my research group. They're the ones who really did all of the hard work. The anode work was started by Jesse Coe, who's now at Johns Hopkins. It's been taken up by David Ajman Budu, who's currently in my group. And the lithium mapping with diffraction was done by Partha Paul, who's now at the European Synchrotron Radiation Facility. I also want to thank my collaborators and thank the funding sources. The tin antimony alloy work was done by an energy frontier research center called Scaler. And the other work was done by the Excel program, which was a collaboration between Slack, Argon, Idaho National Lab. And then finally, I just wanted to spend two seconds to plug a virtual field trip that we're doing between Berkeley and Slack. Please visit our website to register for the next workshop, next field trip. You can also see past shows there. Our next virtual field trip is going to be zero carbon housing. Thank you very much. Thank you so much, John. The field trip looks very exciting. Well, thanks for the wonderful talk. Wow, two really challenging problems from the material side of an allowing, allowing, electrode and overcharging. So we have a number of questions, but I thought I will start with, with an overarching one. So you mentioned in the second part of the talk. A lot of analysis based on tracking where the lithium goes in terms of between NO cathode and lost lithium. What have you learned from the images about where the lithium plating occurs? So we learned, similar to what people have observed in the past, that the lithium plates near the edges, but not at the edges of the anode. And they actually plate further from the edge than, for example, the overlap between the anode and cathode. There's a midge match between the anode and the cathode. So it's a little further in than that. But it is knowing where the lithium is going to plate first is still unknown, I would say. That seems to be very random between the cells. If you just have a little bit of lithium, where is it going to be going first? And so some follow up work that's happening now is really to track in the first few cycles, rather than looking at, you know, already dead batteries. Where is that lithium plating first? And how is the cathode, the state of the cathode and the health of the cathode related to that or not related at all? Right, John. So just to clarify, so you're seeing these sort of large scale features, like you said, a bit away from the edge, but not quite at the edge. Were there also unexpected, like a hot spot? You know, all you need is really one plate of lithium to short the cell. Were there any sort of randomly occurring spots in the cell in your study? Not in our study that we discovered. And I think that goes to the fact that we were looking at cells that had been cycled for 450 cycles. So they were very well cycled. And so there was, at least in the 60 and the 90 cells, there was a lot of lithium plating. And so we didn't see, we didn't look at any cells that had actually been shorted by lithium dendrites, for example, they all survived to 450 cycles, even though they had severe capacity loss. So we weren't looking at that sort of very detrimental but rare event of lithium shorting the battering. So here's another related question to that. So I think most of the images you showed are all planner images. Have you done any, this is a question from one of our audience, have you looked at any cross sections about, you know, is it happening closer to the separator closer to the current collector? Is it the lithium plating work? Right. Yeah. No, we haven't yet. And I think that is a tricky thing to do in the cells that we had. We have looked at cross sectional cells, but they're usually sort of designed cells that aren't your standard cells. We do have some related work trying to use both neutron and x-ray imaging in coin cells, and that will, and with tomography, and so we will be able to see in all three dimensions where the lithium is plating. And so that work is pending. Wow, that is really exciting. So the goal is to have a 3D resolved plated lithium. Yeah, yeah. And the reason, yeah, yeah. And the reason we use neutron imaging in addition to x-ray imaging is they're both sensitive to very different things and so neutrons are very sensitive to lithium metal. Awesome. Well, we have time for one final question. So you also showed this really nice results of the alloying electrode. I think that you've also work on conversion electrode. Do you generally find that x-ray imaging has sufficient resolution to reveal these what we think to be very nanoscale effects in these very high volume changing electrodes? It's always a struggle. So as a microscopist, you have this sort of tug, pull and tug between wanting to get the highest resolution image possible and wanting to see a statistically relevant amount of the battery. And so I can either with this transmission x-ray microscope, I can get 30 nanometer resolution on a single particle. I can do this in 3D and get maybe 50 nanometer resolution, but I'm spending a lot of time on one particle. I could also go to the micro CT and get a larger field of view 3D image, but with only maybe a micron or half a micron resolution. So it's always a question of what do you want to see? What's the question you're asking and then picking the right tool for that question? If I want to go to even higher resolution, then I would always go to an electron microscope. But of course, then you have to design your sample so that it fits within the electron microscope. So you can look at a lot of particles and get a relevant survey of what memory particles are doing, or you can go to high resolution and figure out what's happening on the nanoscale on a single particle. And so I would say we need to do everything. Sounds like a big opportunity. We have a lot of additional questions in the chat. So I think, Johanna, if you have a few moments, feel free to answer them via the chat. And thank you so much for sharing that. We'll have you back for a discussion after Eging's talk. So following Johanna's excellent introduction of X-ray imaging and diffraction, Eging Liu, who is also a lead scientist at Slack National Accelerate Laboratory and at the Stanford Synchrotron, will continue this thing and talk about 3D imaging from very small scales to very large scales. And I also have the pleasure of knowing Eging for about 10 years. I still remember one of the early talks he gave. I forgot exactly what material was on 3D tomography of some very complex system. And I thought, wow, this is so much you can learn just by visualizing things with such exquisite detail. Something I want to say about Eging is that not only is he developing the measurement methods, but he's also developing and pioneering the interpretation methods. These data sets that you obtain, it's very, very large terabytes in size. And he's one of the pioneers in applying advanced methods such as machine learning and computer vision to really understand and interpret these giant data sets. And Eging has been leading the fields of understanding cathode materials and developing cathode microstructure, which is highly relevant to lithium-ion batteries. So Eging, we're greatly looking forward to your many of the beautiful images and movie you will show in the next 30 minutes. Eging, go ahead please. Okay. Thank you, Will, for the very kind introduction and thank you for the opportunity to present my work. As you can see from this title, I will highlight the multi-scale aspect of the research for lithium-ion batteries. Now, before I talk about batteries, talk about the technology that developed over the years, I want to show you one old photo taken about about 10 years ago. You can see that this photo was taken in the year 2013, and you can see Johanna and myself together with our colleagues, Joy and Darius at the BIM line at Cesario. You have met Johanna just now. She has not changed at all over the years. On the other hand, you see me right now, I age a lot. Together, we not only study batteries, but also provide useful data to, you know, reveal the degradation or the aging mechanism for the battery scientist as well. Now, a little bit introduction of my research group. I work in a highly interdisciplinary area. We play with the materials, we do the characterization, we make sense of the data, and then we do all of this for the desired functionality. Well, the cartoon is simple, but when you look at the real sample or real data we are interfacing with, you realize that there are a lot of problems here and there. The materials can be very messy. The device is full of defects. The experiment we do is highly delicate. Many things can go wrong. And the data we acquire in there is very noisy. And that's why we really need to tackle this problem in a very systematic fashion in which we need to develop not just the experimental methods, but also the computational tools to help us to have this information from the data. Now let's talk about lithium-ion batteries. I'm sure that the people in the audience are very familiar with this. The lithium-ion battery has two electrodes and the lithium-ion goes in back and forth as the cell is operated. Now, if you look at the real thing, the real physical device, you realize that it's highly complicated through many different scales. As I said, you know, if you zoom in to look at the internal structure of a cylindrical cell, you see this gery roll structure. If you further zoom in, you look at the electrodes, they are made of thousands and millions of particles. And these particles, they are grown together in different fashion. They have different size, they have different shape. They have complicated internal structures. If you look into the subparticle structure, you'll see the domains, grays. And of course, if you further zoom in, look at the atomic scale, they are local phase transformations and local lattice distortions as well. Now, because I didn't study batteries, I didn't study electrochemistry in my grad school, that's why when I look at the complicated systems, I need to develop my own understanding of it. This is what I did. If we think about what happens when you charge a lithium-ion battery, let's look at the bottom row here first. What happened in the charging process is that the lithium-ion and the electron would diffuse from the bulk of the cathode to the surface of the cathode. If you look at the electron and the lithium-ion, they will go through different pathways, and hopefully they will meet on the other side, on the surface of the anode, and then they get together, they then integrate into the anode structure. Now that reminds me of the traveling experience, which everybody is familiar with, right? What you have to do is you have to diffuse through a traffic, you get to the airport. If you look close to the airplane already, you might even see the airplane through the window, but there's a little bit of inconvenience standing in the way. After that, the passengers and the check luggage, they go through different pathways, but hopefully you will meet on the other side, then you can go through another diffusion to the destination, which is a hotel. Now the problem here is that these two charge carriers, they have different resistance, lithium-ion slow, electron is fast, is similar to this traveling scenario, where the passengers usually arrive, has no problem, but the luggage can be missing. So that's why you potentially get stuck because of that. Now for this complicated system, we care about the morphology, we care about the lattice structure, we care about the oxidation state. And utilizing the S-ray based technology can help us to understand many of these different aspects of the material at these learning scales, and can offer very valuable information for us to gain in-depth understanding of the system. Now let's show, let me show you a few examples. We can start with a small scale, looking at the lattice distortion within a single primary grain. Now in this particular case, what we do is we utilize a nano-focus S-ray beam and illuminate a lithium-cobra oxide single grain. Now when we talk about a single primary grain, we usually think it's a single crystal, but is it really a single crystal? The answer is not. And what we do is that when we have the nano-focus S-ray beam illuminated on a sample, it will generate a bread diffraction in a certain angle. Now if the sample is really a single crystal, then when you conduct the last scan of the beam over the particle, the bread peak should not change at all. You know, long-range structure alluring. That's the definition of single crystals. But in reality, what we observe is that as we conduct the last scanning, this bread peak will deform, it will move. So that carries information about the lattice distortion within this single primary grain. And by doing this mapping, we can then reveal how this lattice is deformed. It's de-spacing in homogeneity that's twisting of the lattice and also the bending depending on how the bread peak is deformed. Now that's what happened inside a single crystal. It's already very complicated. What happened beyond the single primary grain? What happened across the grain boundary? We show here that a 3D imaging of one single particle that has two grains attached to each other. As we can see from the morphological data on the top here, you know, this grain boundary actually has mechanical consequences. If there's a crack that's formed, it's going to likely propagate along this grain boundary causing these two grains to detach from each other. In addition to the mechanical degradation, the existence of this grain boundary also has chemical consequences. If you look at these images on the bottom, which is conducted using a spectral microscopy that is sensitive to the oscillation state of the element of interest, in this case, you can see that these two domains exhibit different oscillation states, in different cobalt oscillation states, which fingerprints the local state of charge. So the existence of the grain boundary actually hinders the free lithium migration across the grain boundary. These two domains show different state of charge. Apparently, this state of charge heterogeneity at this level is not good for the battery performance. So what we're doing, what do we do? We would like to modify the property of this interface. We want to address this problem by conducting this engineering of a buried interface. The approach we adopted is this trace element doping method. So what we do is we conduct very low concentration, about 0.1 weight percent, co-doping of titanium magnesium aluminum for this material. Now it's interesting to show here that through the experimental result, the different doping actually has different spatial distribution inside this particle. For example, the aluminum is more or less everywhere. On the other hand, you can see from this image that the titanium naturally segregates onto the grain boundary. So it naturally modifies the property at the grain boundary at interfaces, which will then contribute to the improved performance in different mechanisms. These acetyl calutirizations allows us to study the material at different electrochemical states when it's harvest from a cell. As Johanna has highlighted utilizing the S-ray as tools, we can see these materials under operating conditions. In this particular case, we designed a partial cell geometry, which allows us to follow one single lithium-cobal oxide particle over many cycles. And interestingly, we observe that even for this one single particle, if you operate the cell under different C-rate, this particle will respond differently. Now, at the beginning, we were charging in one C, discharging in one C, you see that only 50% of this particle returned to a discharged state after the first cycle. Now if we do this even faster, in a few minutes charging, a few minutes discharging, you see a smaller percentage of the particle is returned to a discharged state, even though the entire cell shows that it's at a discharged voltage. Now if we do this for a longer time, five hours charge, five hours discharge, then you see a significant portion of the particle can be recovered. And of course, if you do the long term cycling, even under the same condition, there will be some degradation over time. So that shows that, you know, there are so much complexity already from the single primary grain level to single particle level. And that understanding is very useful for us to design the materials for better performing better. The degradation of the battery materials not only happens as we electrochemicalist cycle the cell. Here I show you another example in which we first conduct about 20 cycles of the cell and then we just store the cell under low temperature. There's nothing under the temperature except for just exposing to the low temperature conditions. And after that, we recover the cell back to low temperature for further electrochemical cycle. And as you can see, from this prop here, the low temperature storage actually induce some irreversible degradation, where it's not very obvious at the beginning. So after you recover to low temperature, but as your cycle for thanks to hundreds of cycles, you can see this degradation effect already. And then what happened, right? So we can conduct imaging of this material from the electrical level to a particle level and the low temperature. And what we observe is that there's a irreversible structural degradation. You can see there's more crack form under low temperature. And the formation of a crack is not going to be here when you recover to low temperature. That's why it will use the fast charging performance for the following cycles. Now, if we further zoom out, you know, we have showed I've shown you so many examples at a single particle level, but if you further zoom out, you realize that these particles, as I said, agglomerate together, and they are embedded in this porous carbon and bondometrics. And collectively, you know, this whole system actually delivers the desired functionality. So able to conduct high resolution imaging of relatively large volume of the electric, you will realize that the degradation pattern is highly complicated. For example, in this particular experimental result, we found out that these two cluster of particles, they are only about 100 microns apart, but a degree of damage is very different. And what's causing that? And what's the consequence of this phenomenon? That's the questions that my group and my colleagues have been asking ourselves and have been making efforts to address. Now, we can, you know, we have the capability of conducting this experiment and getting the data, but how do we analyze the data in a statistically meaningful fashion, you know, with good efficiency. So we have been looking into the development of machine learning algorithms to help us to do this, to achieve this goal. I'm showing you here one example, right, so this is the one two dimensional slice through a 3D volume we imaged. Now we do have good image contrast in this particular case, but the problem is that the segmentation, the identification of different particles inside this image is not trivial. For example, if I simply utilize conventional methods for image segmentation, I will likely identify these small fragments as individual particles. Now, if I take this data as input for my statistical analysis, I will end up with some impression that there are many small particles and these particles have different, what do we achieve? Right, and now this is not true because, you know, with our eyes, plus our brain, we know that these smaller pieces, they belong to the same particle. And we know that a property conducted segmentation should result in something like that on the right hand side. So we would like to do this for many, many images of this and we would like to do this automatically and we don't want to have too much of human labor in this process. Now, the, in order to do this, that reminds me, it's conceptually similar to the phase recognition program that every cell phone has nowadays. So I did a little bit experiment. This is a screenshot screen recording for myself. Okay, I was trying to take a group photo of these brilliant people. Okay. And you see that my cell phone is trying to be smart. It's identifying the face in the field of view and put a yellow box on the face that is recognized. Now, you can see that it's never perfect. It never identified all the faces in the image. And to me, you know, it never recognized ever Einstein sitting in the center of the group. I think that's unacceptable. Now, we were to do better than this. And how do we do that, right? And I think, you know, here I'm just showing you this one where it's simple but yet very effective approach. Now the difference we have here is that we have a three dimensional data and what we can do is we can take the 2D slice of the three dimensional data in different orientation, different depths. And then once the slices is shredded, we can apply the traditional method utilizing the previous our group previously developed machine learning method for the 2D particle identification. And the previous developed machine learning method is already a big step but it will have errors here and there in my miss a certain particle in a certain orientation in my miss the same particle in a certain depth. But because of the fact that we have a three dimensional data and by adding this data fusion step dramatically improve the fidelity of the identification process. So in our case, there are many, many particles, thousands of particles in this development allowed us to vary efficiently and automatically and accurately determine every single particle in the image volume. And that data set serve as a good input for the follow up analysis, which I illustrate here systematically. So what we do is, we take the data, identify particles, and for every single particle we extract the structural and chemical and, you know, different characteristics of the single particles. So take this data as input, we build another model to conduct a prediction of the electric damage. And the question we were asking is that, what is the critical characteristics that will determine the final damage degree of the corresponding particle. And the fact that we were able to harvest thousands of particles provide sufficient amount of input data for us to conduct this statistical analysis of this machine learning. So long story short, what we conclude in the end is that in the early stage, you see here I'm comparing the two cycle in actual and not a 50 cycle electric. And the horizontal axis is the different characteristics of the particles and the actuals in the vertical axis is their contribution score to the final degree of damage. Now this comparison shows that in the early, early cycles, the features on the left play a more significant role for the later cycles, the features on the right actually becomes more significant. Now if you read this text here on a horizontal axis with more details, you realize that this is a very interesting pattern, the features on the left is more relevant to individual particles characteristics. On the other hand, the feature on the right is more relevant to the particle to particle interaction. So that's why we think this reveals this individualism was in work in the early cycles, maybe, you know, the individual particles they behave very differently because they are different from the very beginning but in the later cycles, in order to really make it last for longer cycles, making, you know, a prolonged cycle life, we really need to look into how to put these particles together. Okay. So, let me further zoom out. Now we talk about the electric. Now, what do we see when we further zoom out, we see the cells, right. And the cells, there are many different form factors, people are making the cells for different applications. For example, in this particular case, we were studying a commercial cell, this is a cylindrical cell that failed the quality inspection because of the self discharging effect. So the question we were asking is that what was wrong, what's happening inside and if you take a low resolution but you know covering the entire cell tomography of this cylindrical cell, you realize that, you know, it doesn't look bad. It's actually pretty decent structure. It has all the structure that's anticipated in the system. Now, the key is to look a little bit more into our details and look at the different locations. So for example, if we carefully investigate these, you know, current collectors, you know, positive terminal we see the deformation of the current collectors. And if we go through this electric size at different depths, you can then identify, for example, there's the lamination, there are voids, cracks, here and there. And in addition to that, we also see this impurity particles inside this cylindrical cell. For example, you see this white bright spots here, indicating that the layer, there could be some impurity particles inside. Now, having this 3D data reconstructed, allow us to identify the exact location when we unroll the journey road structure. And once we do that, then we can harvest the electrical corresponding location for further analysis. So what we did is that we cut this small piece out, and we conduct a number of different follow up characterization. For example, we can conduct for rest of mapping over this particle for interest. And what we see is that this different impurity particles they actually have different composition. For example, many of them has chromium and iron that could come from the stainless steel. So, which may come in from the machine operation. And some of the particles actually has zirconium in it and haphazone in it. So these particles could well come from the ball mating process. So knowing this composition of these impurity particles will provide useful information for us to go back to the manufacturing pipeline to identify the problem. In addition, we can conduct spectroscopy measurement over this region of interest. So for example, we can conduct the surface ray spectroscopy in different modalities and using hot as well to achieve different probing depths. So it shows that the existence of this impurity particles could further induce surface reconstruction. And you could also affect the subsurface and any bulk redox reactions inside this. Now, this is, you know, cylindrical cell. It's actually very important and broadly utilized form of the battery. But if we look into the other cell formats, for example, looking at the pouch cell format, the conventional tomography method actually face a lot of problems here because just because of the shape of the cell so that you cannot really rotate the entire cell for the 3D measurement. So what we did recently is that we developed a Laminography approach, which, you know, shows the experimental geometry in here, I won't go into details. But what I want to highlight here is that using this method, we can actually not only see a cathode with good resolution and good fidelity, but also see the graph anode with good contrast actually. And this is done in situ so we can monitor the evolution of the cathode and any anode as we operate a cell under different conditions. So for example, in this case, I'm showing you here that as we charge the cell, we can see that there are some hotspots developed inside of cathode particle. So that shows the intraparticle heterogeneity. On the other hand, for the image on the bottom here, you can see that there are two small particles on the left, the intensity is more or less the same, but on the right hand side, you see one because brighter than the other. So that shows this intraparticle heterogeneity, both of which can be induced as we cycle the cell or we abuse the cell electrochemical. The same thing can be done to the anode side, the graph anode as I show you in this image here. There are void formed upon charging, you know, cracks can be found in certain particles, anode particles, and we can also observe some of the debounding cracking at the electrical level as well. More interestingly, we also observed lithium plating in this cell. So what we did is that, you know, because of the fact that we have a 3D data of this pulse cell, so we can select the same depths. So this is a slice near the anode and the separator. You can see that at the beginning is what it's doing. There's no structure whatsoever. But as we overcharge the cell, you start to see this interesting feature that's developed and we associate this observation of this feature with lithium plating because we can disassemble the cell and again we can visually inspect the location. So which confirms that this is indeed the lithium features. So it actually was a little bit surprising to me because originally everybody was thinking that this low concentration, you know, low Z element is very transparent to the heart as race. But we are showing here that with the proper experimental configuration, this can also be observed. So that really opens up a lot of opportunity for us to study this complicated degradation of lithium ion battery and the operating conditions. And quite like here that there are many different degradation mechanisms and the fact that we are able to visualize different components in the system really offers opportunity for a in depth study in this field. So I talked about the complexity of the battery from the single primary primary grain to secondary particles to the actual to the cell to big cells. Eventually, the battery cells has to be integrated into a system to power some device. So this is a very recent study that we were doing. It's very interesting. I have this wireless earbuds is abused by my son. It's not working anymore. That's why we decided to put it into the S3 and take a look what's happening inside. It's very interesting that, you know, in addition to observing the damage inside a cylindrical cell. We can also reveal the structural defects at a device level, how the cell is connected, how it's integrated into a system, how it interface with, you know, for example, battery management unit. And all of this, you know, because more industry relevant and, you know, the existing tools actually provides good opportunity for research in this direction but in my opinion, significant improvement in the throughput is still desired. So here my group is also looking into the methodology developments to improve the efficiency and hopefully we will provide a better solution for the future efforts in this field. So finally, let me conclude. I talked about the multi-scale structural and chemical complexity in the lithium-ion battery. I also want to highlight that the manufacturing of the lithium-ion battery actually has many, many different steps in it and they are dedicated. So many things can go wrong and utilizing the S3 tools with different modality, covering different landscapes came all for opportunity for us to not just to get the fundamental insights, but also to conduct the failure analysis to get the valuable information to inform the production process. So with that, I would like to thank my collaborators in all these works. As I show in this slide here, we work with material scientists who create materials and we utilize a suite of S3 techniques, not just a single term but also the laboratory S3 tools. We conduct computational developments to help us to make sense of the data and we work with theorists to understand the insight and to provide feedback to guide the next generation, next iteration. And with that, I would like to thank you for your attention. Thank you so much, Ijin, for that beautiful talk linking all the different length scales, largely using one technique. I think this really speaks to the power of the approach. So we have a number of questions here and maybe we can start with a high level one. So Ijin, you showed a lot of beautiful images and tomograms. Can you comment a little bit on how it is connected to the macroscopic behavior of the battery in terms of the electrochemical measurements? Are you able to correlate the two? So that's a good question, thank you, Will. There are two things I want to highlight. For example, when we are working on high resolution imaging, looking at single particles, we often design special cells. For example, the capillary cell, some people even do single particle cells. So for that, on the one hand, the electrochemical data is directly relevant to the particle you're observing. But on the other hand, the electrochemical data is not the real cell, the electrochemical data. So there's a trade-off over there. But on the other hand, for the large-scale industry-relevant or maybe even commercial cells, I think the statistics is really key. So in addition to just imaging one cell, or maybe a few cells, I think the high throughput catalyzation of many, many cells and linkless two-layer electrochemical data in a statistically relevant fashion. I think that would be the right thing to do. And I believe that a lot of industry is doing this. And our expertise can apply and also I would say that as I highlighted at the end of my talk, methodology development for even higher throughput to conduct a structural measurement will be key to really address this. Great, Ejin. There's also now just another question on imaging even lighter and dilute species, specifically the electrolyte. Can you comment on some of the opportunities for imaging electrolyte, salt depletion, wedding non-wedding and so forth using x-ray tomography? So there are challenges in general when we are talking about dilute species and also low concentration and also low Z in general using x-rays. So I think this is a common sense. But as I show in this recent development, it actually surprised me that utilizing a laminography approach can actually show you the lithium-prating inside. And with fairly good resolution, you know, it is one micron resolution. So I see the opportunity over there. On the other hand, I also want to highlight that, for example, if you have some electrolyte additives in the cell, actually, you can also image a bulk of the material as an indirect measurement of the consequences of the presence of the additives. For example, we have a very recent paper in Nature Energy together with the group from Brookhaven starting these electrolyte additives. At the very beginning, we were utilizing soft x-ray tools to really focus on the surface chemistry. But then we started brainstorming. I was asking, you know, in the end, the lithium has to integrate into the bulk, right? It's the bulk that store the energy. So when we tune the surface chemistry, in the end, it should reflect into the bulk heterogeneity as well. So, as I said, then we conduct, again, similar approach in chemical imaging, you know, with machine learning statistics, we realize that indeed a little bit of a semi-chemistry modulation can actually affect what's happening in the bulk. So you're not directly measuring what's happening on the surface using hot additives, but it will help you to understand the consequences as a whole. Great. Thank you so much, Yijing. And then maybe we can take one final question before inviting Johanna back to the panel discussion. So in several of the papers you mentioned, you're identifying these very small changes in the composition or microstructure of cathode active materials. So have you been able to derive some unifying rules about how to design the microstructure of these cathodes based on your understanding of the degradation pathways? Yes, that's a very good comment. So one thing that we propose in a recent science paper is that we observe the polarization effect, which is kind of unavoidable, particularly if you are going to seek electrodes. And we propose that, you know, by doing this structural engineering at the actual level, for example, we can tune the packing density as a function of the depths, then, you know, it could potentially mitigate some of this polarization effect. But in the lateral direction, we think that it's not just to make the individual particles better, but to make the consistency of different particles so that everybody will work together. It's a very interesting, you know, common made by my colleague, Ke Jie from Purdue University, he said it's very similar to managing a big group of researchers, right? You have 20 students, five of them are super good. So what you do, right, you can use them, you can, you know, let them work very hard. At the beginning, then they get tired out. Then you start to look into the second tier. So this is one scenario. The other scenario is that if you engineer the system such that these individual particles, you know, this group, they more work together as a whole as a team. Then maybe you reach to a equilibrium at that stage, all the particles are still in good shape. So, so it's not just making the materials, but also to put the materials together in coherent fashion. I think that's a broader comment to be made about things in life in general. Awesome. Thank you so much again for the wonderful talk and the beautiful images. So I think we have just a little bit under 30 minutes for a discussion with the both of you. So if I can ask Johanna also to rejoin, welcome back Johanna. So, you know, in this part of our seminar, we sort of go all over the place and discuss things a little bit more broadly. So I thought I would maybe start the discussion by asking the following question. I think this may be the questions on a lot of people's mind. So this seminar is attended considerably by those working in industry. And in industry, iteration time is very important. People are trying to, you know, design their next best batteries tomorrow. And you both show these very extensive and in depth studies of materials that took, you know, many months to measure and sometime years to fully interpret. So there's a little bit of disconnect right so this is basic research and method development using very advanced and not widespread available tools. So that's where we are today. You've made the case very clearly that this can really advance battery engineering and development, both of you. But how do we make these tools to really work on the time scale of industry as people are trying to, you know, really go through thousands or tens of thousands of iterations to get a better battery. Maybe I can ask Johanna to weigh in on this I think this is kind of the crucial question on my mind. Yeah, well, I would almost say that's the wrong question to ask. And I'll explain myself. So I think you're right these tools are very advanced and they take a long time to both gets the data and analyze the data and we're working on speeding those sort of things up. The flexibility of the instruments is an issue. And so I think there's, there's two things that we can do with a synchrotron the first is to really use the synchrotron to get some fundamental understanding of how these class of materials work, and with that understanding, be able to validate models that could perhaps expand that into similar but you know not identical cases, either different materials different chemistries scaling it up things like that. And then the second way is, we can also think about maybe developing other in situ monitoring techniques. And I'm sort of, I haven't really thought about this in detail yet, where, maybe we can get an understanding of of signatures, for example, there's been some work in in using sound sound waves to to probe batteries and and and understand degradation that's happening in a battery. But it's really hard to interpret the data. So we could do x ray imaging or other characterization with x rays at a synchrotron simultaneous to a probe that would be more accessible to industry, maybe even accessible to monitoring electrodes as you are making them or cycling batteries at a larger scale. So I think those two connections connecting to the fundamental research to sort of broader scale up issues with modeling and also then monitoring techniques and understanding the signatures from different monitoring techniques that could be used at the industrial level. Yes, yes. So, I agree with everything Joanna said. Now, maybe in my perspective, it's useful to look back in the history of s rays. Now, before singleton even exist. All this s ray characterization tools was available in the laboratory, people were doing this in the laboratory is broadly available, you know, if you have a decent budget, you can buy some tools for your lab to conduct some studies. But then we have the singleton development which shows, you know, great improvement in the throughput because of flowers and all that. So people start to, you know, come back and come to utilize singleton as a, you know, let's say advanced tools. But after so many years, with all the developments in different technology as resource as a detector as three optics, and all that. Now we are actually seeing a trend that we can bring some of these capabilities back into the lab. You know, it doesn't have to be as bright as a singleton, which is billion dollars too much. But you know, if you say, you know, one out of many to lower, or maybe a little more than that lower, you know, it becomes start to become available. And for this, you know, the industry application availability, and the accessibility is so important. So I think at this point, it was to look into some, you know, effort to really make it stand alone technique as well. So I think there are a lot of efforts in this field already, and I'm also looking into this in the future. So Jean, if I merge your thoughts, I think what you're saying here is, yes, it will be the throughput for these advanced measurement is indeed much slower than other lower fidelity measurement like lab tomography and extra imaging. But the insight is much greater. So if you can somehow transfer the insight backward. That would be very helpful. But I think you're also saying that there may be opportunities in the coming years or decades to bridge the gap a little bit more maybe bring more intense x-rays to the laboratory. And that has already happened, just not as far. So it sounds like there's a both way traffic going, going. And are you both envisioning a future where, you know, these synchrotron x-ray imaging and laboratory x-ray imaging will all be sort of together in a very giant iteration loop and then just sort of a down selection process of, you know, picking the most important samples that would deliver the most insights, and then put that in front of a synchrotron x-ray. I think for the public science side of things, yes. I would also say that we should, as a battery community, think about how the pharmaceutical company uses the synchrotron. And so the pharmaceutical companies use synchrotrons all of the time. They do proprietary research. They actually use the same beam line as their competitors. And they have really solved the accessibility issue by simplifying the data that they want and making a pipeline. John, maybe you can speak a little bit more to this pipeline. Like I said, I think many of our lessons are from industry. Can you briefly describe how companies might be able to employ these resources for their internal R&D efforts and maybe highlight collaborations or two in which this has happened? Yeah, yeah. So I think it's at its infancy in the battery field for sure. But so let me first explain how protein crystallography works with the pharmaceutical company. And so there they have essentially simplified the synchrotron question that they're asking. And so they want the crystal structure, the protein structure of whatever drug they're creating. And they have used computers to automatically look at the data as it's coming in, reject anything that isn't of high enough quality. And the scientists from the pharmaceutical companies don't even need to really know much about analyzing the data because it's all been automated. And so I think that's the extreme case that we could maybe aim for. I would maybe argue also that maybe our problems are a little more complicated than just knowing where atoms are located with relative to each other. But I think we can take a lot of those lessons for eliminating the red tape or creating a very smooth path for industry to make these partnerships happen with staff scientists doing proprietary research. And also, I think we have learned a lot from the past two years of operating a synchrotron during a pandemic, which has made us leverage remote access far more than we ever were comfortable before. And it has allowed us to think about problems of automatic data transfer, trying to collect data remotely, automatic sample changing. We now have robots on many of our diffraction beam lines to swap out samples. And so we're working towards that. And I think it'll work in many different cases, but there will always be opportunities for these harder experiments with the in situ or the multi scale degradation problems. So I think it's a combination of both. So you asked for an example of companies. So I've actually worked with a few different companies, both in a larger scale like we had a collaboration with Bosch, for example, and also startups where it was more trying to get do we funding like be through an SBIR or something like that small business grant. So we work on both levels, because Eugene and I are, you know, scientists by trade, we get fed by publications so there's less of an incentive for us to do non for us to do proprietary work, but there are mechanisms at the syncretron that allow that to happen. And I would say that syncretron, the cost of using a syncretron is very cheap, because you're not paying for human labor, you're just paying for instrument time and and so compared to other things. It's not an expensive. Yeah, so maybe I can add that to really facilitate collaboration between large facility and the industry. I think for the industry, it will be beneficial if they have a clear definition of the question that they're asking. So for example, some of the questions, more engineering questions, maybe you really need to have a lot of not just measure one sample 10 sample maybe you need to pay 100 samples to really get an idea with statistics, then that kind of question. You know, you need to consider and there may not be a good question to ask from, you know, singletron, you know, experiments. On the other hand, if you have a, you know, a very difficult but very important cutting edge, you know, material problem. Maybe you don't have a choice. You have to use singletron. So, so I guess, depending on the need, then we can work out different arrangements for for collaborations. I will also say that, you know, we mostly utilize the singletron but we also utilizing the laboratory systems and, you know, to study the industry relevant samples like sales and commercial sales, the syringes about sales and we also have let's say expertise to offer in that domain as well, just a different domain. I think the National Lab generally encourage this and we can work together to explore. I think both of you made it very clear today it's getting the result the data is just half of the problem and then interpreting analyzing interpreting it is actually could be even more challenging so I think it highlights the need for both. So, so maybe coming back a little bit to the science and the engineering side of things. Both of you talked about sort of two extreme of microstructure right. One extreme is when the microstructure changes very large, right. Like the tin alloy. Electra that Johanna showed. The other extreme, I think you're talking about a tiny crack that happens. So there's both the small microstructure change and big microstructure change in the battery. Can you talk a little bit about sort of what you think the future is in terms of microstructure engineering. There is now this increased notion that we need to go to these very microstructure changing materials like conversion electrodes to get the high energy density. But there's also this feeling that will be small microstructure changes really what kills you at the end. So maybe we can you know zoom out from the methodology itself but to the sort of the science and engineering of how do we manage microstructure going forward right so that the limit of no microstructure changes is long gone right we don't do that anymore. You know, that doesn't give you the type of energy density needed for for battery technologies. So I think this is something that's being very sort of on the top of my mind as well as we think about, you know, how do we manage microstructure. So how do we manage people. You want to go first you want me to go first. Okay, um, so I think it goes up at least for the alloying anodes. What we found is. So the manner of porosity is somewhat like you're saying is is a little bit of of trying to prevent the microstructural change or maybe just make minimize it. But I think what we've also found is that many of these alloys go through either crystalline structural changes or amorphous crystal crystalline changes. And so we've found that I think there was there was one anode that we looked at that went through crystalline structure change if we lithiated it, but if we so the aided it, it was all amorphous and then ended crystalline and allowing the anode to go through an amorphous change to heal the cracks that we created early on. And so this was allowed us to sort of manage the large morphology changes because we healed the cracks as we formed them because because it was it was no longer just crystalline phase changes. I think the the other thing to think about is whenever we're managing these these microstructural changes is, you know, if are we developing systems that are scalable. You know, if we create this is perfect anode that can cycle for thousands of cycles, but you can only develop it in a small beaker and you'll never be able to scale it to something that's industry relevant then. And I think it's a scientifically interesting finding, but it's not necessarily going to to develop the battery community. I think, you know, a lot of this study provides insight into the magnetic insights and that, you know, we, what I say, so so we take this man is insights in a way that, you know, is not just a literal, you know, finding you know it's just just not just by itself but we need to extrapolate a little bit. So for example, when we talk about the polarization effect, then we see okay maybe the gradient will be a viable approach and how do we achieve a gradient. We can do, you know, somehow feel guided or some for example the 3d printing, you know, they becomes like open field right so the inside we got is maybe it's a desire for the gradient structure. But how do we achieve it, then it's open field everybody can come up with new ideas to achieve that so so that's one example. Right. And also, for example the the consistency of different particles. Right. And we say it's important. We don't necessarily need to make like perfect spheres right or consider perfect sphere we can have some elongation for example if that's better for our system perspective. But if we do have the elongation, then, you know, do we align them on, you know, in a natural structuring efforts, or do we prefer not to have them aligned to prefer random orientation. So the many magnetic study can help us to get that information to figure out what's desirable and then, you know, and more engineering effort can be proposed to achieve it. Eging, thank you so much for that. Maybe building off what you said, and I think when when we report energy density in academia we often say you know per kilogram right graph and metric. But often in industry, the interest is more on the volume metrics side, especially for electric vehicles to keep the battery small. And, you know, I, if you look at some of these problems that you have reported on today, you know, decreasing the volume metric density is a good way to solve the problem maybe by adding more carbon network or by providing more space for the particles expand and shrink. So what do you think about this sort of what seems to be a fundamental trade off between graph and metric and volume metric energy density when it comes to a sort of volume changing material. Is there any way to break that relationship. So you know, you know how can we pack the electrodes much more closely, but still allowing it to last many thousands of cycles. I don't have a good idea for that. But I mean to me, I agree with you. It seems to me it's like a fundamental conflict, we are trying to resolve, you know, you know, I can make a ping-pong ball so robust and it bounces around it doesn't crack. But, you know, the density is so low it's not going to be useful for the SDSR. But I guess, after all, it's an application-driven design. I guess it's better to start from the end application and then define the need and, you know, define, you know, what the metric density is desirable and all that. I mean, work from there, backward to conduct the material design. I think that will be the rational approach. Johanna? Yeah, and I agree with what Eugene said, and I think designing electrodes, not just designing the materials to have high capacity both graph and metric and volume metric, but designing the electrodes as well. And I think there's some problem that we found in the fast charging world is you can design a battery that fast charges. That's not the problem. But can you design a battery that's high capacity that fast charges? And so I'm, you know, I'm less worried about, and maybe I should be, I'm less worried about capacity, graph and metric and volume metric, because I feel like the battery industry has done that. We have really great electric vehicles. And I'm worried about looking into the future and looking at do we have the material resources to meet this demand that we've created because we've made electric vehicles so desirable. And we've sold the idea that we're going to electrify everything and everything's going to have a battery. Can we actually do that with lithium? Can we do that with NMC? You know, so, so I think, at least for my interest, going towards those, those novel electrode materials that meet what we have graph and metric and volume metric right now, but with lower cost, more sustainable materials. I think that that's at least my interest in where the community is going. Sounds like an awesome vision, Johanna. Thank you for that. As we're coming to the end of the hour, you know, I thought I would ask more of a forward looking question. So I find it's very useful to look backward. So, you know, if you take yourself back to, you know, 2005 2010, I don't know what you thought, but I, you know, if I would say when 2022, what will we achieve? My guess is that many of the things you presented today is like, oh, that's probably going to be very, very hard to do all of this stuff to see lithium to see carbon, you know, to see 3D microstructures intensive nanometer level to do batteries, you know, cycling operando. I don't know, maybe you're optimistic that you thought that was very easy, but I certainly thought it was very challenging. So I would like to ask you to sort of think ahead for another 10 years. Where do you think we will be at in terms of advanced material analytics. In terms of characterization, what do you think we can do 10 years from now that we won't be able to do today as a sort of for looking sort of set the goalposts for for the community. I'll give you a few seconds to think about. Yeah, yeah, I think one challenge that we have yet to tackle and we've been, I would say we've been wanting to tackle it sort of from day one is the true combination of techniques. And so one technique will never answer all of your questions. And, and right now we often take one sample and go to different instruments and or even similar samples and look at them in an array of instruments. The ultimate goal would be to take all of those instruments and put them into one, one mega instrument and be able to characterize your battery, while it's operating across many different probes, and really get a fundamental understanding of the skills on on the same battery, and really make those connections and maybe that's a complete pipe dream but I think as a as a as a characterization person that that's the future that I get excited about. Well this video is recorded so we can play it back 10 years now to see if this is realized but it sounds so awesome. Well he likes to give pictures of me 10 years ago. I mean, for me, what I want to do in the next 10 years is to bring all these techniques we talked about today, you know, it's so cutting edge so fancy so high resolution. So bring it available so that we can plug it into the manufacturing, you know, lines, you know so that you will provide sufficient throughput efficiency and sufficient degree of automation. So that's really utilized in the industry, instead of just utilize a single job. So that's, that's my my dream. And how am I going to do that. I don't know. But but I think efforts are needed to push this direction. Both of you on record stating these really I think really bold goals but you know, like I said, you know 12 years ago 2010 you probably. Yeah, what I thought today is not quite possible but yet here we are. But brighter notes. I mean there are a huge amount of resources being invested a huge amount of interest. And this is a critical link to the energy transition and not just for batteries before other material driven technology as well so I'm very hopeful that we can make that happen. And the other that the entire community working on this. So, thank you both very much for sharing that bold vision. I really look forward to talking to you in 10 years, or and also between them as well. So with that, let me join our audience and thank you both again for taking the morning to talk so so number of exciting things happening in the coming weeks here at Stanford. We will have our first big in person seminar. So for those of you tuning in from the Bay Area. We will be hosting Professor Martin winter from the University of Moonser. So he directs the meat battery center one of the largest academic industrial complex in Europe, and he will be joining us next Tuesday. So if you're interested in attending, please register now and the spots are very limited. And it's a one of the great opportunity to meet our European colleagues here in the Bay Area. And then, following that, we will have another exciting storage X seminar in this is next week. So our schedules a little bit out of sequence so on May 27 so one Friday one week from today. We're going to have a discussion on what is a very interesting concept of cell to pack. And Professor Chaoyang Wang and Mooji Ejas CEO of next energy. Our next energy will be discussing batteries at the link skill between cells and packs. And this is a very interesting technology to increase the performance of batteries not at the materials level nor the cell level, but at the pack level. So this will be a very interesting length bridging discussion. And then two weeks after that on June 10. We are very pleased to feature Kevin Wujek, who is the chief technical officer at blue current. He will be a very promising solid state battery startup here in the area with strong connection to Stanford to talk about their latest polymer based solid state batteries, and then he will be joined by my colleague, Professor of Urva saleo, who will also talk about next generation recyclable batteries based on polymers and organic molecules. And just to remind everyone, please connect with us on social media. We announced a lot of our events on LinkedIn. They're going to be more and more and more thanks to the easing COVID situations. And we really look forward in in sending many of you. And then for those of you very interested in education Stanford also has a professional education program. So take a look at this link. There are multiple courses were offering including a recently launched one on electric vehicles and electric grid being being taught by experts from Stanford and my colleagues. So with that, I'd like to thank everyone once again for joining us today morning here at Stanford. And I hope all of you have a great weekend. Thank you everyone.