 Good morning from California. My name is Will Chu. I'm the co-director of Stanford Storage X Initiative. Together with my colleague, Itui, I am pleased to welcome you to the final symposium of 2020. I think I speak for all of my colleagues here at Stanford that we are excited for this year to be over and can't wait for 2021 to begin. But before that happens, let's do one more scientific session. We have so far covered four topics on X. We had X equal fuel, X equal lithium-ion batteries at terawatt hour, X equal heat. And to top this off, today we have X equal longer duration storage, where we'll learn more about electrochemical technologies beyond lithium-ion batteries. So to get us started, E will introduce our first speaker. E, please. Well, thank you, Will. Hi, everybody. This is Itui from Material Science as well. It's my great honor to introduce George Krapche from Argonne National Lab. Krapche is so well-known. He doesn't really need that long of introduction. Let me just mention a few highlights of his career. He is a scientist. Early days working on high-temperature superconductors become so well-known in his scientific contribution. He has been elected. He was elected as a fellow of American Physical Society. He was a fellow American Academy of Arts and Sciences and also a member of National Academy of Sciences. In addition to his scientific contribution, he's also well-known for his leadership contribution and helping the science administration. He served as the associate director and director a couple of times in the Material Science Division in Argonne National Lab. He recently, not so recently anymore, eight years now, has been serving as the director of the Verkhay Hub, the Jay Caesar. And the personal level, I know George is such a great person. He's willing to spend his time with his colleagues, willing to spend his time to serve the community. Certainly many of us having the DOE research program would know George has been serving as advisors one way or the other to the Department of Energy, particularly the basic energy sciences. With that, I would like to welcome George to come to the podium to introduce to us the latest and greatest staff in Jay Caesar, George. So Yi, thank you so much for such a wonderful introduction. And I do recall very finely our relationship and friendship that goes back many years, as you said. I'd like to thank you also for the opportunity to address this symposium, really an honor. And I would like to talk about the new subject, which Will mentioned, long duration discharge energy storage, lots of challenges. I would say more challenges than opportunities at the moment. But we're working on those challenges. So what I would like to do is first introduce the little diagram on the right. That's the symbol of Jay Caesar. And I'll have a little bit more to say about that later. We believe that you should build batteries from the bottom up. So if you want a new battery with a new function, start at the atomic level and design from there up. I would like to cover the applications need for long duration storage, flow battery challenges, and some recent advances. That's what I'll do in my time. So first, some of the mega trends that are shaping the grid. And these are trends that are going to be around for the next couple of decades. Renewable wind and solar on the grid in 2019, that accounted for about 9% of the electricity. Predictions are by 2050, it may be 56%, or actually much more. Energy storage is being deployed on the grid in, I would say, prodigious quantities. It can be used for intraday firming of wind and solar on the minute or the hour time scale. Evening extension in the case of solar gets some solar electricity after sunset. And what we want to talk about today, consecutive days stabilizing. There can be two or three or even more days that are cloudy or calm. And we need to stabilize the grid against that. Storage is growing very rapidly on the grid, about 20 gigawatt hours now. And by 2050, predictions are 4,500. The third mega trend, climate change, we're all seeing it. Lots of outages in the power grid due to hurricanes, due to flooding, due to wildfires. And now, especially in California, due to wildfire risk. And I think, in fact, in Southern California, even this weekend, lots of the power is shut off because of the fear of starting a fire. And finally, decarbonization, something which has really come to the fore since 2015 with the Paris Accords. And there's some very aggressive targets to decarbonize everything by 2050. And the grid, in particular, in some countries by 2035. So there are lots of drivers for long duration energy storage. And you can see here another mega trend. And that is the cost of storage is falling dramatically. So this graph came out about a year and a half ago by Bloomberg. They plotted at the bottom. Here you see these yellow curves, the levelized cost of electricity for solar, two kinds of solar, and for wind offshore and onshore since 2009. And of course, solar is the poster child for dropping costs down by not quite a factor of 10, but getting close to it since 2010. Then they projected back the levelized cost of electricity from batteries, that's this dotted line. And remarkably, it's falling even faster than solar and wind, a dramatic price change. And if you see now that storage is actually about a factor of four more expensive than wind or solar per kilowatt hour, that will be important in planning the grid. But you can also see by extending this line with your eye that eventually they're going to meet and even cross. And if you look at comparing these prices with the cost of natural gas, here is the levelized cost of electricity from a gas peaker plant. And you see remarkably in 2020, it's about the same as the levelized cost of electricity from a battery. So you might expect the gas peaker plants will be ripe to be displaced by batteries. And even more interesting, if you plot the levelized cost of electricity from a combined cycle gas turbine, it's the cheapest way of producing electricity nowadays, it's pretty much comparable with the price from solar panels. So you might expect, again, by extrapolating these falling cost curves for storage and solar and wind, that eventually the grid, simply for economic reasons, will switch from fossil to renewable. Let's turn to long duration discharge storage. Here's an interesting graph or set of graphs from January of 2019. This information is kept by the University of Texas. And this looks only at the solar insulation measured in different ways, diffuse and direct and so on. But the point is, there are many days when that are cloudy, at least in Austin, Texas. And you see in particular here from the 21st to the 24th, you see four overcast days in a row. Well, if we're going to have a carbon free grid, we need to back up or stabilize against those days with some form of energy. And storage from batteries would be an obvious choice. The problem is that lithium ion batteries, which is the most popular battery today and the biggest one being installed on the grid, can only discharge at full rated power for four to six hours. And that's fine to extend the day past sunset a few hours, but not good enough to stabilize the grid against consecutive overcast days. So here's a graph from Paul Albertus' recent paper this year that shows the required hours of battery storage or some other kind of storage versus the percentage of wind and solar on the grid on the x-axis. And of course, it goes up rather dramatically. This is a log scale, simply because you can use batteries to replace the gas peak replants that only operate a few hours, four to six hours at a time. But there are plenty of gas peak replants that operate for a longer time. And those require longer duration storage. So this simply illustrates that if you want to have completely carbon free grid 100%, you're going to have to have somewhere between 500 and 1,000 hours of continuous storage discharge. And this, of course, is well beyond what lithium ion can do. So there are lots of candidates for this kind of long duration discharge storage. Of course, redox flow batteries, we're going to hear about today. Thermal storage, pumped hydro, compressed air, gravity storage, gas peaker with carbon capture and utilization, and a hydrogen peaker plant or fuel cell. So we'll talk about some of these. I talked about the cost of batteries and solar panels. And that introduces a very interesting question, which remains, I think, unanswered. How do you balance the ideal grid between solar wind storage and transmission? Leave hydrogen off for the moment. So because solar panels are so cheap, you might say, well, let's overbuild. And we'll have a lot of curtailment. But that'll be the cheapest option. On the other hand, batteries are coming down in price. And if you install batteries, you don't have to overbuild the solar so much. And of course, you don't have as much curtailment. And at some price point, that will become an attractive alternative. If you look at the transmission lines, which are another way to share otherwise curtail solar and wind energy, they're probably the most expensive thing to build. On the other hand, there's a lot of them already in place. And how can we make the best use of that? And that's what we want to do, of course, to deliver electricity to the distribution grid where the users are. Now you might ask, why do I have hydrogen? In the title of this, I haven't talked about hydrogen. But there is another option. And that is to use all this cheap renewable electricity to electrolyze water and produce what's known as green hydrogen. If you do that, it's interesting that it's very inexpensive to store, even for long periods, essentially, infinitely, in tanks or underground. It's very inexpensive to build pipelines compared to the cost of a transmission line. So it's a cheap way to transport energy. And using fuel cells or simply burning hydrogen in a combustion turbine, you can produce electricity and supply effectively any need in the distribution grid. So it's a shadow grid, another energy carrier for electricity that has all the versatility of electricity. And it's completely interchangeable with electricity. So it could do a lot for us. It can actually do more than this. So hydrogen has the potential to decarbonize things that electricity can't touch. Things like long haul trucking, things like maritime shipping, maybe across the ocean, things like flight, burning airbus just amounts recently. It has a five year plan to make a purely hydrogen long haul aircraft. And heavy industry, I'm thinking of cement and steel making and petrochemicals, which use a lot of fossil fuels that could be replaced by hydrogen. So there is an opportunity here to think of a new energy carrier, very complimentary to electricity with multiple options for solving problems, such as zero emissions, such as low cost, such as high reliability and fast recovery. So this is something that is worthy of a lot, I think, of attention. If we look now, let's turn to organic redox flow batteries. We all know how a lithium ion battery operates. Lithium oscillates between the anode and the cathode, storing and releasing energy. It's chemical energy, which is stored in the two electrodes, anode and cathode. The power is restricted by how fast the lithium can move in the electrodes and in the electrolyte. And the energy is restricted by the storage capacity of the anode and the cathode. So with organic redox flow batteries, it's a little bit different. The active ingredient, the analog of the lithium, is confined in a tank. It never moves. It doesn't go back and forth. There's a negalite or an aposalite, sometimes called analyte and catholite, which take the place of the electrodes. It's anions or solvent ions that move back and forth across a separator that carry out the redox reactions. The power is limited by the area of this reaction plate, separating the two tanks. And the energy is limited by the size of the tanks. So the power and the energy can be completely decoupled. This is great because you have the tank as large as you like. You can store as much energy as you like. That's great for the grid. You have liquid electrodes. So there's no strain due to volume change when you charge or discharge. That gives a long lifetime, maybe 20 years or even longer. And because they're organics, there's lots of design space. So there are thousands to millions of organic molecules you can think of that would be useful as active ingredients in the organic flow battery. So there are also challenges. Low solubility, that means you have a low energy density. You always want to get the solubility up. And crossover, unintended, of the active ions from the analyte and the catholite tanks, which ultimately degrades performance and limits the lifetime. So if you look first a little bit about a battery that is actually not organic, but is a flow battery, but made with ordinary elements. And this is a battery that was developed in JC's, or we started to look at the problem of about 2015. We were looking for batteries that were made from ultra-low-cost earth-abundant materials, things like water, things like oxygen, things like sulfur, of which there's plenty, because it happens to be a byproduct of refining petroleum. And we developed a long-duration battery based on these earth-abundant low-cost materials, which we have the IP on. In 2017, two years after we started, we spun out a company called FormEnergy. They were very aggressive in getting funding within two years, $50 million from these sources. And in 2020, just three years after we spun them out in five years after we started to look at the problem, they signed their first delivery contract with Great River Energy. That is Minnesota's second-largest utility to deliver a battery, remarkably, that can discharge at full-rated power for 150 hours, very different from the lithium-ion battery limited to four or six hours. So this is the first step. We'll see how it works in the field. Be delivered in 23. And I'm sure there will be lots of improvements along the way, but this is quite a breakthrough milestone. Here are some of the founders of FormEnergy on the left. Turning now to organic, so that was a water-based battery. Why do you want organics? Well, many compositions and structures for organic molecules. So there's a very large design space. And some of them are shown over here. Give you an example. You can go with, try to get, high-solubility. That would give you high energy density, high stability, so you would get long calendar and cycle lifetimes. Of course, organics are made from earth-abundant low-cost elements, carbon, hydrogen, oxygen, and so on. You can look for multi-electron transfer. That is two electrons transferred on each redox reaction. That would give you higher capacity and higher energy density. Lots of organic solvents that you can choose from. They may have wide working windows, wider than that of water. That gives you high voltage and, therefore, high energy density. You would co-design the membrane with the active organic molecule to prevent crossover. And you could design the battery to self-report its state of charge, its state of health, degree of degradation, and maybe even self-repair itself. And this is something that Jay Caesar is looking at. What are the challenges? Well, the biggest one, really, is that there's a wealth of molecular compositions and structures. How do you even start to assess that access, that space? And you have to find and design molecules that satisfy multiple requirements. That's a tough one, maybe as many as three or four or five requirements at once. And they often conflict with each other. Yeah, here is a way of attacking that problem of such a huge design space, so many molecules. Imagine that you want to achieve some properties for a flowback battery. Solubility, stability, voltage, crossover, multi-electron transfer, whatever it is. And imagine that you have some molecules in mind that you may design to meet those performance requirements. The traditional way is what's called direct design. We started off, of course, most of human history. Let's just think of it. Let's think of a material that might satisfy this. Then we'll make it in the laboratory and see if it works. That was dramatically changed when we started to do computational screening. So we can do high throughput simulation of hundreds, maybe thousands, of molecules on the computer a lot faster than synthesizing them in the laboratory. And then take the best, most promising candidates from that computer screening. So that speeded up things actually dramatically. There is another thing coming, sometimes called the self-driving laboratory, where you use artificial intelligence because of its knowledge of structure property relationships to start with the desired performance characteristics and find or propose some molecular structures and composition that might satisfy those performance requirements. So this is a new feature. So instead, with AI, instead of simulating hundreds or thousands of molecules, you go immediately to the most promising five or 10 and consider only those. You could then have automatic synthesis. So the laboratory using a robot would make the material, automatic characterization, run it through lots of machines, send that information back to the artificial intelligence brain here to score the material. Did it actually work? If it failed, I wonder why it failed. Let's try something else. So it actively learns from every cycle of this synthesis route. And this is what's coming to the fore. So up here in the corner, I have what we are doing in JC's to do this. This is our design loop. We start with the seed material. We use quantitative structure property relationships as our AI. We can get much more sophisticated than that. Automatic synthesis, high throughput characterization. We've got multi-objective targets, and that's critical for flow batteries. We get some data, and we learn from each experience every time you go around this loop. So here on the left are a few of the targets that are central. And you can see electrolyte conductivity, membrane conductivity, viscosity, diffusivity, solubility, rate constant. And what's shown here is the properties of various organics normalize to what a vanadium flow battery would do. That's this dotted line. And you see, in some cases, we're better. In some cases, we have a ways to go. But we've already perfected in JC's every stop on this loop. It is not yet fully automated, but nevertheless, we can carry out that loop and get the benefit. So I want to now end with three examples of satisfying multiple performance requirements. And the paper is up here in the left-hand corner, as it will be for all examples. Here, we decided to design for high operating voltage and high stability using structure property relationships and computational prediction. And interestingly, this group, it's Melanie Sanford and her collaborators at Michigan, had previously published a molecule that had high stability but only modest voltage and a second molecule that had high voltage but only modest stability. And they wanted to combine their experience from making those two molecules to make one that was high voltage, high stability, and high solubility. And of course, this is the molecule they made. Along the way, they made lots of others, which are down here, will show what they did. So here's the discharge capacity versus cycle number. So what you want is a very stable molecule that doesn't degrade as you cycle, say, 300 times. And you can see this 2 plus, the bad performer. That's this guy that they started with. The 1 plus, which was very highly stable, that's this guy. And it's a little hard to see because they almost overlap. But this fellow right here performs as well as this very best one that they started with. And all these others form intermediate between those two. What they showed is that multi-objective targets can be achieved. The derivative families must be explored carefully. So very small differences between these molecules, but very big differences in performance. So that's the first example. Now, here the idea was to get two electron transfer on each redox reaction. And that means you have three star charge states of the molecule. And of course, not sacrifice, solubility or stability. So all three of those charge states have to have equal solubility. Turns out that's a challenge. They again started with previous attempts at making molecules and made rather small changes. So this is just derivative changes, adding elements to the molecular design. And finally came with this one and this one, these two, which were their candidates. And it turned out when you add two electron transfer, you can easily lose solubility. That's what happened on this trial. And it takes additional functionalizing to restore that. The system in which it's a symmetric flow cell here for probing the three charge states of this guy. This is their best performer. And you see here plots that you are very familiar with capacity versus cycle number. These are various C rates, that is power rates. And you see reasonable performance. Here's the theoretical capacity. Here is at a constant power rate. And you see the theoretical capacity at over 140 cycles. That turned out to be 460 hours. So making some progress. Not yet good enough, I think, for the battery. Putting in a self-reporting feature into the active material. And the idea here was to use fluorescence because you can shine a light on it and see how it responds. It's pretty non-invasive. You could imagine that that would measure the state of charge, the state of health. Or in this case, the active species crossover between the analyte and the catholite tank. And it could be a potential trigger for rejuvenation. Maybe you want to drain all the electrolyte and put in new at a certain point. Or even self-repair. And these are things that we have in mind. So they tried to add the fluorescent feature, self-reporting feature, to the molecules and found that it interferes with the electrochemistry. So starting with this molecule, high solubility, high stability. Add the fluorescent feature and you lose the solubility and you get fast degradation. So that was pretty much of a disaster. Here's another version of this. And you see it's not very different. It adds the CH3 element and slightly changes some of the others. And you're back to high solubility and high stability. The very interesting experiment where they, using the fluorescence, monitored the crossover of the active material from analyte to catholite in real time. And here are the plots. This is for the same active material. And it's in acetonitrile. And you see, in both plots, there's an induction period here. That's the time it takes the active molecule to actually penetrate the separator or the membrane. And here, if you make 10 times the concentration, it actually crosses over at a slower rate. That might seem counterintuitive, very interesting. And here, if you put a different solvent, then you see that the two solvents behave very differently. So they show that indeed, in real time, you can measure the crossover. And it has some features that depends on things that you might not have expected. This is our JCZR sort of diagram. We are trying to build batteries from the bottom up. We want to start with crystals, with atoms, with molecules, put them together to make transformative materials, chemistries, and architectures by adjusting the atomic and molecular structure and use those transformative materials to build batteries that are actually designed for the application. Next click will show where we're going. We have three directions, one of which is redox-mer design here in the center. That's what we've been talking about during this talk. We also look at liquid and solid solvation. What's the atomic and molecular basis for that? Because that controls almost everything that happens in a battery. And we're also looking at multivalent ions, such as magnesium, calcium, or zinc. And those would be the inorganic analogs of, let's say, dual electron organics. And the motivations will, you can see at the bottom, I won't read them to you. We make extensive use of computational screening, both for crystals and for liquid electrolytes, which we call the electrolyte genome. And we also are now putting in machine learning, which I think is the next big phase. So bottom-up design of redox-active organic materials is certainly within reach. Here's our system. Many simultaneous design targets are required for a successful battery. Here's a few of them. It's a real challenge for conventional human imagination plus trial and error. It's almost, I would say, impossible for that conventional route to really survey the space available. Inverse design with machine learning, that's the next step, that's what we're doing. And we have this, we call our self-driving laboratory. We call it an automated experimental machine learning platform with this little abbreviation. So with that, I think I am done. And the next slide, I believe, says thank you. Well, thank you, George, for the introduction of what's happening inside this season on long duration storage. So George, let me start by asking a question of the definition of long duration first. Every time I use long duration, I'm thinking about 10 plus. Right. It's the same definition right here. Well, 10 plus is on the short side. Yeah. We saw in the Austin data that there are often four cloudy days in a row. If you look over greater time spans, it can be as much as 10 days in a row. So I think 10 hours would be on the low side. If we're serious about decarbonizing the grid, we need much longer. Even 100 hours may not be enough. Yeah. That's good. That's basically the low end is 10. So that's a similar definition. Okay. So one question, first question is, George. We know lithium ion is going down on the path and the system level get down to $150 per kilowatt hour and then also go down to a hundred probably in a few years, right? So if you look at the long duration storage, for example, about five days, let me use five days or maybe a week. So let's say every year you have about 50 cycles, 60 cycles to use for that, how many days, you know? And what will be the cost needed? I think there's a certain analysis in the literature about what's the cost target when you need to go down to per kilowatt hour of storage cost, right? The capital investment, what's the target that they see is a really light enable to meet the long duration storage? That is a great question, Yuyan and one that has been analyzed, but I think not enough attention has been paid to it. And let's take an example. Let's say you had a lithium ion battery, it could discharge for four hours and you said to yours and you cost you $100 a kilowatt hour and you said, well, I want to make it eight. The obvious solution is to buy a second battery. Then of course you can get eight hours out of it. The downside, you've doubled the cost per kilowatt hour. It now costs you twice as much as it did before because you have the second battery. And that's the problem. You need a decreasing cost for the longer storage discharge duration. So in Paul Albert's paper and many other papers, there are targets depending on the length of the discharge time. And if I'm remembering right, for a hundred hours discharge time, the cost has to be something like $3 a kilowatt hour to compete with existing technology. If you want more than a hundred hours, it has to go down proportionally. It may seem impossible to achieve those goals, but certainly with Redox flow batteries, when you just make the tank bigger, you don't have to make more batteries, it's possible that you could get the price down. And some of the estimates, the technical economic modeling that have been done, suggests that that may be a viable route. Yeah, yeah, it makes sense because it decoupled the power, the energy with the power for the Redox flow batteries. So George, the second question is, you talk about non-aquist organic Redox. I think Mike is going to talk about that as well later, maybe in a panel discussion, we can go deeper. During your talk, I keep thinking about the aqueous route versus non-aquist route, right? In aqueous route, there is Vanadium Redox flow, there is a number of other chemistry available right there. If you will analyze this from the high level, it compare aqueous versus non-aquist, right? What will be the difference in terms of opportunity can enable the long-duration storage? Yeah, and that's another great question. And there are some techno-economic papers, I can send some references, you probably know them already, then look at exactly that question. And the conclusion seems to be that the advantage of non-aquist is you have lots of choice of solvents, so you may be able to design a little bit better than you can with aqueous, which is limited to water. And you might get a higher voltage because water, of course, electrolyzes at 1.23 volts. You'd like to work at maybe four volts, and that would give you a lot more energy. And those are the advantages. The disadvantage is the cost. And it's not so much the cost of the elements that make up the organic molecules, they're cheap. It's the synthesis cost. It can be very hard to synthesize complicated organic molecules. So the cost goes up, whereas with water, you can almost say nothing could be cheaper. And so where the non-aquist batteries may have higher operating voltages and they may be able to be designed a little bit more perfectly, let's say, for the application, the aqueous batteries have much lower cost. And they have the advantage that they're already out there. You mentioned the vanadium redox flow battery, which has been commercialized for more than a decade. So there's more experience with them. And I think those are some of the trade-offs. Mike, in the panel discussion, they'd like to add to that, of course. Yeah, yeah. I probably will wait until the panel after the mic also give the top. So from the audience, there's a question, also George, by the way, and Zoom right here, we only see a limited number of audience, but it's huge audience watching on a different line right there. So not visible to us. I have David Boy right here feeding me the questions. So one question is, you mentioned for the long duration storage for this whole system level of hydrogen showing up as very attractive. We know hydrogen was intensely looked at maybe more than a decade ago in the previous administration. And it looked at the hydrogen. What are the challenges for storing large quantities of hydrogen for extended period of time? Maybe let me just add in my comment as we're over the whole hydrogen economy. We talked about more than a decade ago. Now we kind of revisit this. And what are the challenges we need to overcome? What's different right now to enable us to make it possible as a viable solution? Yeah, I think this is one of the most interesting questions. And probably for the next decade, will remain an interesting question. So you mentioned the hydrogen economy. And it was 10, 15 years ago that we looked at it seriously. We were thinking at that time that it was just fuel cells. And we were not thinking at that time climate change. So we never mentioned the word climate change and hydrogen in the same sentence. Now we do all the time. Climate change has really come to the fore. And even amongst what's called the general public, climate change, they all think it's an issue. So where's the challenge? And I think the biggest challenge is producing green hydrogen from electrolysis of water cheaply enough. Right now, it's three or four, sometimes five times as much as the hydrogen you get from steam reforming carbon dioxide. And that, of course, produces steam reforming methane, which, of course, produces carbon dioxide. So that's the challenge. We didn't have 15 years ago cheap, renewable energy that was carbon-free. We agree. It's a different time right now, yes. It's a very different time. So I think this is the major issue. And I go back to the graph I was showing of the dramatically falling costs of both solar panels and batteries. That's because of demand. So if indeed Asia and Europe, which have already signed on to hydrogen as a second energy carrier, mostly for decarbonization reasons, if they drive up the demand as expected will happen, the price could do a similar, let's say, learning curve fall. And predictions are it could come down to $2 a kilogram, maybe $1 a kilogram, which would make it completely compatible with present fuels. So although that might be 10 years off, as it was for solar panels and for batteries, it's something that we should look forward to. Yeah, Georgia agree. In some of the region now, PV electricity costs, the price, actually selling price, purchasing price, is already less than $0.02 per kilowatt hour. That's my feeling. $0.02 as renewable hydrogen becomes very interesting. $0.01 will be a large quantity generated, low cost enough. If you get to $0.01, then it's game changer completely. We are on the trajectory of doing that. I mean, it's thinking about green hydrogen is, I think, actually making sense. So with remaining time, maybe I'll just ask one more question before I invite Mike and Will to come to the podium to introduce Mike. Well, in terms of the cost of the organic greenhouse flow, I mean, green is very, very important. But certainly the membrane part, that's the device determining the power. So what's your feeling or anybody's calculation? How low cost membrane needed to be per meter square by 2 really enable, let's say, $20 per kilowatt hour, even $10 or below per kilowatt hour of the cost and for the flow batteries? Great question. I don't have a number for you, but I can tell you the following. And I didn't mention it in my talk, but we're looking not only at single molecules, monomers, but at polymers, typically maybe oligomers, three to six monomers long. And the big advantage of that is they're big. So if you size selection as your membrane and you make your polymer, let's say, three or four times the size of a monomer, suddenly the crossover problem is dramatically reduced. And it becomes reduced not because you made a better membrane, but just because you made a bigger active molecule. We are looking at that as a serious way to get the cost on. But you're absolutely right. The membrane needs to be looked at. And it needs to be looked at in conjunction with the active material because that's the thing you want to screen again. So you can use charge rejection. You can use size rejection. There are lots of ways to address that problem. And I think we're only at the beginning of starting to explore them. Yeah. So that is good. So in terms of the component, there's one audience asked questions over all for the redox flow. What's the most important feature, right, George? Right here, there's a number of things this person listed. Is the redox potential? Is it heterogeneous kinetics? Is it electrochemical reversibility? Is it solubility? Is it transport property? So what's important? I think you mentioned in your talk, there's a couple of things. I mean, stand out first, but I mean, probably just re-emphasize what are the important properties you are really looking for. That is a great question. And to give a generic answer, you have to find molecules that satisfy many of the properties at a single time. So to bring out one and ignore the others is probably a failure. But the two biggest problems in my mind are the crossover problem. And that can be affected by the membrane as well as by the molecule. But that's endemic everywhere in flow batteries. That's the first problem. And the second one actually is cost. You want to have a molecule that performs many functions, but also doesn't cost much. And that's a really tough challenge. On the other hand, let me be optimistic. There are thousands, maybe hundreds of thousands of candidates out there. We just haven't found the right one yet. Yeah. Okay, George, with that, I would like to thank you for your talk and certain remain online. I'll invite you back to the panel discussion later. Now let me pass the podium to Will Che and his young boy right there participating early career scientists. All right, Yi, thank you very much. And George, I must say your presentation was extremely exciting and entertaining that my son, who I think is the youngest viewer of this symposium was keeping great attention. So I'm also very pleased and honored to introduce our second speaker for today's long duration storage session, Professor Michael Aziz is a professor of materials and energy technologies at Harvard University. He is the co-inventor of the organic aqueous flow battery for which you received the 2019 Energy Frontier Prize from Eni. He is the fellow of the American Physical Society, the Materials Research Society and the American Association of As Mensa of Science. With that, okay, that's my cue, Mike. Go right ahead. Thank you. Thanks to both of you for the kind introduction. That's the youngest introducer that I've had ever makes me feel young at heart. All right, I'm pleased to be with you this morning to talk about our exciting new flow battery chemistries. I'm not a chemist, but I collaborate with two terrific chemists, Roy Gordon, experimental and synthetic chemist, Alana Spurugusik, theoretical and combinatorial computational chemist. And thanks to George for introducing flow batteries. I really don't need to spend any time introducing them for grid scourge, I have just one slide. I will introduce aqueous soluble organics. So it's not a choice of aqueous or organic. The big choice is aqueous or non-aqueous. And then the second big choice is soluble, aqueous soluble organics or aqueous soluble in organics. At least those are the choices I faced. And I'll talk a little more about that later. I'll show some of the performance of our most competitive chemistries now. They're major challenges and where we're going from here. As you've heard, the wind doesn't always blow. Here's three weeks of wind. The sun doesn't always shine. Three weeks of sun, the city always needs power. In fact, the cover slide is this photo here. It's Boston, a view from across the Charles River in Cambridge where I am. Obviously the sun isn't shining and the wind isn't blowing, the city's brightly lit. We're burning fossil fuels and we need to stop doing that. So the role of storage is to permit that, taking what nature chooses to give us and turning it into what we need. And what you already learned is the energy to power ratio is essential here. You need large energy to power ratios because that ratio, megawatt hours per megawatt is the number of hours over which you can discharge your storage system at its rated power before it's drained. Flow batteries are designed to get you cost effectively to large energy to power ratios, unlike in traditional enclosed batteries where the energy component and the power component are just rolled all together in one single jelly roll and you can't independently size them. In a flow battery, you have the electrochemical reactor that is going to determine your power rating, the size of the membrane, size of the electrode and then the energy rating is in the size of the tanks full of electrolytes. And so if you want high energy to power ratios, you're talking about big, dumb tanks. If the mass production cost of the electrolyte is low enough, then you can get to costs at long discharge durations that you can't get to by stacking up banks and banks of solid electrode batteries. So this has been known for some time. There are plenty of flow battery chemistries that have been out there. What you need to do is get that curve under the curve where someone's willing to pay for it. And I just want to take a couple of figures from a paper I wrote with Bill Hogan who's at the Harvard Kennedy School of Government. He's the world's foremost electricity markets expert. And we just asked the question, what would a battery, how cheap would a battery have to be to pay for itself at any node in the grid by arbitrage? That's the huge market arbitrage of by low cell high and pay for your own installation. We got data over the PJM organized electricity market. It stands for Pennsylvania, New Jersey and Maryland, but it extends through Ohio, Indiana, Chicago. It's the biggest organized electricity market in the US and hourly prices are available over the seven year period. So we got prices for every hour at every node where we had complete data. And there were two markets. There's a day ahead market where you know the prices, all the prices 24 hours in advance. So it's easy to figure out how to optimally charge and discharge your battery in order to maximize revenue. And then there's a real time market where you don't and you hope that with machine learning and so on, you can still optimize or approach optimum performance that way. But we just asked, suppose we did have perfect foresight with a battery of different energy to power ratios. We studied a variety of different discharge duration batteries here. How cheap would it have to be in order to pay for itself at each of these nodes on the grid in each of these markets? And so let's look at the real time market, for example, for a 10 hour discharge duration battery. What this says is that at about $105 a kilowatt hour of capacity, if you could install it for that, you would have been profitable at 50% of the nodes. If you could install it for about $90 a kilowatt hour, you'd have been profitable at 75% of the nodes. And if you could install it at about $75 a kilowatt hour, you'd have been profitable at all of the nodes. Now that price point under which you have to be able to install your battery goes down with increasing discharge duration because at longer discharge duration, there's fewer times when you get to completely discharge your battery without having a rest in between where you can recharge it. And therefore you get fewer pay days for fully utilizing the longer discharge duration battery. So that brings the curve down. And flow batteries are hopefully getting curves that will come down and cross below these things, whereas stacking up banks and banks of solid electrode batteries gives you some sort of horizontal curve that peters out at some number. George said for lithium ion it was six hours and that's about what I think as well. So Vanadium Redox is the most commercialized flow battery technology out there right now. It uses four different charge states of Vanadium ions in sulfuric acid separated by a proton conducting membrane. Here's an installation in Japan, a five megawatt hour installation. You can see the tanks of battery acid and the power conversion units there. That's a shipping container for scale. Ronca power in Dalian, China is near the end of completion, I believe, of an 800 megawatt hour of Vanadium Redox flow battery. That's a hundred times bigger. So that's an enormous, enormous battery. The problem with Vanadium is it's not a highly abundant element, it's not very cheap and there are enormous price fluctuations because it goes into Chinese rebar, for example. So when construction heats up in China the price of Vanadium goes up. So we've been looking for ways of getting the performance of Vanadium out of organics. The idea with organics is you have truly earth abundant elements, carbon, hydrogen, oxygen, nitrogen, sulfur. And so can we get this cost curve down and drop down into competitive ranges at shorter durations? And we've made a lot of progress on this. I wanna start our story with benzene. There's a benzene ring here. There's hydrogens on the corners, the chemists don't even draw those. But instead of hydrogens here, there's two hydroxies. So that is actually called a hydroquinone. And at a certain voltage that will spit off two electrons and the two protons will jump off and turn into this molecule here which is called the benzoquinone. So the hydrogen is still there. So that's the oxidized form. That's the reduced form. And it's a simple two electron, two proton, proton coupled electron transfer redox process. I started in this field of storage by working on inorganic ways of storing electricity like many others. But noticed that some groups were making progress using organics in fuel cells. And of course a one-way fuel cell is half the problem. Getting it go forward and in reverse is what a flow battery is all about. I started talking with the chemists about finding molecules that would work in two-way operation in flow batteries. Roy Gordon, Ted Betley, Alana Spuduguzek. We hit upon this one that's in photosystem two picking up electrons from chlorophyll in photosynthesis going between the oxidized form and the reduced form over and over and over again. That's exactly what you want to happen in a battery but this tail makes it insoluble. These methyl groups change the redox potential to a place that's not good but with the beauty of synthetic organic chemistry is the chemist can do all kinds of things about this. So first step is take the tail off, raise the number of rings and now the redox potential goes down to a good value for a negative terminal for a aqueous flow battery but that's not soluble. You can stick solubilizing groups on and this is the first molecule that worked well with us. This was the start of the field of aqueous organic flow batteries called anthoclinon disulfinate. That's a good negolite molecule. We paired it with a toxic corrosive bromine pozzolite and there's actually a European company that's licensed that and is using it for utility and industrial scale storage but we've been looking at less toxic, less corrosive ways of storing energy and that's where I really want to start the technical part of our story. We're looking for molecules that will have different redox potentials because you can put different things on the rings and shift the voltage around and you can oxidize one at the same time that you reduce another to charge it and then just reverse these reactions when you discharge it and the requirements, this is almost from the Q and A one of the questions that George got what's important will these turn out? All of these are important and if anyone in them is bad, it will kill you but we need to look at redox potential you want to be almost splitting water at high potential and low potential but not quite. The solubility needs to be high to get a reasonable energy density. The kinetics need to be fast to get you a reasonable power density. The stability needs to be long to get you a reasonable lifetime and the mass production cost needs to be low in order to make these cost effective to produce. So let's see if my clicker is still gonna go. Here we go. So here's our first alkaline flow battery from 2015, I think 2015, yeah. This molecule now with these hydroxy groups here is stable and soluble in base and we paired it up with an old fashioned ferroferry cyanide pozzolite molecule that's been around since the 1980s. It has cyanide in it, that sounds scary but the reason cyanide is lethal is it attacks the iron in the hemoglobin in your blood. Here it's saturated with iron so it's safe as long as you stay in alkaline neutral to alkaline pH conditions. If you go to acidic pH, you generate hydrogen cyanide and that is certainly lethal. But in base, and here we're at pH 14, it's stable and soluble. So here's our dihydroxyanthroquinone. There's the ferroferry cyanide. There's the single electrochemical cell and the pumps in the background there that were our first alkaline flow battery chemistry. And let's see how the performance of this chemistry looks. First polarization curve, George showed you one of these. Start at zero current density and you can see the open circuit potential at about 1.2 volts depends somewhat on state of charge through the Nernst equation as it does for many batteries. And then you draw current out of this and the voltage drops because of internal resistive losses, you want the internal resistance to be as low as possible but you can see we can draw over an amp per square centimeter of current density out of these cells. And for those of you working in non-aqueous chemistries, that's not a typo, okay? Those are not milliamps, those are amps. These tend to be like two orders of magnitude higher than what you can get in non-aqueous chemistries and that's why one of the reasons why I believe in aqueous chemistries. So we have only maybe a third of the voltage of lithium ion batteries but orders of magnitude higher currents and so we can get higher power densities that's lower area electrodes for the same amount of power rating. So the same data can be plotted as power density versus current density. Power density is simply voltage times current density. And if you do that, then instead of having to keep a whole curve in your head, you can keep a single number in your head which is the peak galvanic power density and here we're at about two thirds of a watt per square centimeter. Highest number ever published for a vanadium is 1.3. So we're a little over half of what vanadium has gotten to and they've been at it for 25 years. Of course, we learned from all their tricks and publications. The other really important figure of merit for a battery is the cycle life. And so we started cycling this little battery hundreds of times per day and we extrapolated and it looked like 10,000 cycle lifetimes and that looked terrific. It was only later we figured out that the lifetime had nothing to do with the number of times you charged and discharged the battery. The lifetime was simply calendar denominated. Homogenes chemical reactions are happening inside the electrolyte and the molecule is decomposing. And so a fade rate that might be a half life of 10,000 cycles turns into you're losing 5% per day and that's completely unacceptable. And the way we figured that out is by starting to build symmetric cells. So here we have the same exact electrolyte 50% state of charge on both sides. We call it volumetrically unbalanced, compositionally symmetric. Mark Anthony Goulet developed this technique and it works very well. You completely charge and discharge this side and any fade that you see is completely a function of what's going on in this side because it's symmetric than having things crossover or worrying about things crossover doesn't matter. And here's what you end up seeing. If you're cycling, you're losing 5% per day. Stop cycling in the discharge state, fully discharge. You lose basically nothing here and here, no cycling for a day. If you stop at 50% state of charge, you lose maybe a 1% per day. You see that here and here. If you stop at 100% state of charge, you lose of order 10% per day. You see that here and here. So it's time denominated. It is not cycle denominated and you have to deal with this. Molecules don't need to last forever in order to be interesting compared to vanadium. They need to be cheaper and in which case you can fill your tanks at the beginning with a cheaper electrolyte and get down your capital cost. And then the replacement cost gets spread out over future years and your interest rate for discounting needs to be high enough to make that worthwhile. So the trade-off comes down to this replacement cost ratio, taking the annual replacement cost and comparing it to the capital cost savings. And where you break even depends on the project life. And I just have an example here for a 20-year project. Suppose the vanadium costs 100 units, but the organic costs 30 units and loses 15% capacity per year. And this replacement cost ratio is 15% of your 30 units divided by the 70 units that you saved when you filled your tanks and that's 0.064. And so you'd go over here and say 20-year project breaks even at 2.3% interest rate. So if your interest rate for discounting is above that, then the organic looks favorable. So we've been working on in both of these directions. One is taking the low-cost molecules and increasing their lifetime. And the other is redesigning the molecule and getting truly long lifetime, very stable molecules out of that. And I'll show you a little bit of both of those directions. Here is, these are our low-cost molecules. Our first chemistry, our second chemistry, we were supported by ARPA-E at this point. So we could afford a consultant from the chemical industry to evaluate what the mass production cost would be at various scales. And when you have about a one-volt battery, then a kilo amp hour is basically a kilowatt hour. And so I think of this plot as the production cost of your chemicals per kilowatt hour, dollars per kilowatt hour versus the number of megawatt hours per year that you're gonna produce. And what you see is that small scales, this is enormous, but when you get up to the scale of one ronka power megabattery per year, which means like 180 million to build a factory just to mass-produce your chemistry, then you're down at very reasonable costs here. DHAQ and Ferrocyanide look like they'd be about there, but you've gotta get there and still that molecule doesn't last forever. So on the next slide, we'll show how we can extend the lifetime of that molecule because Marc-Antony Goulet and Lucien Tong have looked into the decomposition mechanism enough to understand it. So here's the oxidized form of the molecule, dihydroxyanthroquinone, right? But we're in pH 14, so the hydroxies have been deprotonated. The hydroxide in the base pulls the protons off and these become O minus. And here in the reduced form, OH becomes O minus as well, but it cycles between this reduced form and the oxidized form. When you charge it, you make this reduced form and at high states of charge, there's enough of this reduced form to go around that two of these will come together and disproportionate into one oxidized form and one over-reduced form. That's lost in oxygen atom. It's called anthrone. It's not redox active and so you've lost it and furthermore, these will dimerize and turn into other things that you'll never get back. But now that we know the mechanism, there are things we can do about it. First is don't charge all the way. If instead you just stop your charging at 88% of full charge, your fade rate is 1 40th of that when you go to 100% state of charge. So here you're not charging fully, you see here you're charging fully, but it's fading fast and here it's fading 40 times lower. Second thing you can do is expose it to air in just the right part of its charge discharge cycle and this molecule picks up an oxygen again and turns right back into the oxidized form of the quinone. So I called this Lazarus quinone. The students called it zombie quinone. The students won the press release called it the zombie quinone. And here's the zombie action, right? You've lost 70, you've lost a bunch of your capacity. What Mark Anthony did was he opened the reservoir, swished around in air, closed it up, got 70% of that lost capacity back. So that's really encouraging and we're working hard right now on testing the proposition that these two effects are multiplicative and that we, by combining them, we can actually get down to hundreds of a percent loss per day. And at that point, we're closing in on that. And if we can do that with this very cheap chemistry, I think we have a viable candidate for a spin out. In the other direction, we've designed extremely long lifetime molecules. And on one slide, I'm going to attempt to cover about three years of development here. The dihydroxyanthroquinone has OH here and decomposes, but David Quaby and Kai Shang Lin put these arms on this molecule. This becomes what's called an ether linkage. The solubilizing group is now further from the ring and the ring is much more stable that way. And the capacity fade rate went down by two orders of magnitude, down to 0.04% per day. But still the ether linkage is vulnerable to nucleophilic attack by this or this. Yunlong G brought the fade rate down by another factor of about three by replacing the carboxylate with a phosphonate group here. And the two reasons phosphonate helps is first, it's a weaker nucleophile than carboxylate. So phosphonate doesn't attack that as readily. But second, it's a better solubilizing group. And that means we can bring the pH all the way down to nine and still have good solubility. With this one, pH had to be at 12 for decent solubility. Now we're down at nine. And so there's less hydroxide around to attack your vulnerable ether linkage. And finally, just this year, Min Wu and Yan Jing, whoops, before this year. No, this happened this year too. These molecules have been licensed by two specialty chemical suppliers. So potential collaborators who asked us, can we send them a kilogram of this stuff so they can try it out? And we've had to say, I'm sorry, that would take a graduate student two years to make. Now you can go and buy it. So you can do your own experiments on these. But this year, we've developed an even more stable class of chemicals. Min Wu and Yan Jing have done this by removing the vulnerable ether linkage completely and having just carbon linkage arms. These arms are truly bulletproof. You can see the fade rate now is in the hundreds of percent, thousands of percent per day that extrapolates to 3% per year for that one. And our current champion, dipheavolic acid and perquinone with a more complicated arm here, gets you down to less than 1% per year. And so I'll just show the performance of this one on the next slide to give you an idea of what's going on. Here it is at pH 12 against ferroferricyanide. Open circuit potentially shown here versus state of charge about one volt. Here are the polarization curves. Here's the power density. So we have about half the power density of DHAQ. It's still respectable. And when we cycle this, here's what happens. So here's the black curve is the discharge capacity versus time. And what we see is we're losing about 5% per year when we discharge fully cycle at pH 12. What Min did here was he stopped and exposed to air and we recovered capacity, same mechanism as DHAQ and dropped some KOH pellets into the negolite and got the pH up to 14. And now we have this world record capacity. So as far as I understand, this is the lowest capacity fade rate in the absence of rebalancing techniques for any flow battery ever published. Organic, inorganic, aqueous, non-aqueous, monomer, oligomer, polymer, whatever. So why does that work? Well, we have the same mechanism going on, this disproportionation mechanism. I haven't even drawn the arms anymore because they're inert. They're solubilizing but inert. All the action is in the central ring where this is the reduced form and two of those come together to form the oxidized, sorry, the oxidized form and the over reduced form that loses its oxygen. So air exposure will give you back your oxygen, that's this, but this reaction involves water and produces hydroxide. And that means if you bring up the pH now up to 14, you increase the hydroxide concentration by two orders of magnitude and drive this back to the left. So with the ether linkage, high pH breaks your arms, but with the carbon linkage, high pH actually helps. All right, oh, this is different on your computer than it is on mine, but I think it'll be okay. We've got the water stability window here and just an overview of where the aqueous organic chemistries look to me right now. The ones that are bold are stable enough to be commercializable. And then it's a matter of production cost and some other things. The ones that are not bold are just not stable enough. And what we've seen in neutral pH is about three years ago, we discovered a very, very stable combination of a functionalized ferrocene and a functionalized virologin, but the voltage between them is pretty low. And with other ferrocenes, with higher potential, the stability isn't there. With other virologins, with lower potential, the stability wasn't there. This year we published a lower potential virologin that is truly stable. So maybe there's hope there for breaking that inverse correlation between stability and voltage. In acid, we don't have anything that's quite stable enough. In base, ferroferry cyanide is pretty good. It's cheaper than vanadium. Not a whole heck of a lot cheaper than vanadium. It'd be nice to replace it, but it is cheaper. And down here we have our Methuselah chemistries and our zombie chemistry. And there's a new molecule I didn't mention before, phenazines, which were introduced by Pacific Northwest National Lab. They have lower potential than anything, but they're not stable enough. But there's a preprint now by Yunlong Ji, former postdoc with us now in his independent career. Looks like he may have developed a very stable phenazine. So that's really interesting. With these, it's a matter of what's the synthetic cost at mass production scale. And if they fade, the ones that are not in bold, how much can we extend the lifetime? So we're working hard on that right now. So let me end by acknowledging our sources of support and pointing out Roy Gordon in charge of the synthetic chemistry, Alana Spuruguzik, whose combinatorial theoretical team has calculated the properties of over a million different molecules for us. And then the main players in the work I told you about today are Minwu, Mark Anthony Goulet, Kaixiang Lin, David Kwabi, Yan Jing, and Eric Fell and Yunlong Ji. So with that, I'll thank you for your attention. Be happy to answer questions. All right, Mike, thank you very much for your excellent talk. So I'm getting a little bit of distractions here. So E, I might ask you to step in if things gets worse. Hopefully all the parents can appreciate this. So we have a number of questions. So let me maybe start off with the chemistry side of things. So Mike, you showed a tremendous progress on improving the caliber of life. This is really wonderful progress. As you go to those molecules with much better stability against decomposition, are there trade-offs? Are you losing something, for example, in the kinetics or solubility that you have to consider as well? Yeah, are there trade-offs? What we haven't seen trade-offs in solubility, we've been managing to solubilize those pretty well. The kinetics seem to get a little more sluggish. Our power density is down by a factor of two from DHAQ. We don't have systematic understanding of that. So I don't know if it's just a chance or if it's really something significant and systematic. The main trade-off is in synthetic cost because now you, these molecules, you start with DHAQ and then you add other organics to give you the arms. And so the cost goes up. And it looks like right now that the cost of the truly stable molecules we've developed is too high even at mass production scale to be competitive. Whereas for DHAQ it looks low enough. So if we can extend the life of that adequately then that's still a win. We continue to look at the lessons learned in order to see whether we can develop infinite life molecules that are low in mass production cost. And so for example, we now have one synthetic method that starts with anthracine instead of anthroquinone. Anthracine is a cheaper, more abundant starting material than anthroquinone. Great, Mike. Thanks so much. It's truly exciting to see the progress there. The next set of questions has to do with the membrane-liquid interface. You didn't have too much time to get into that. Could you speak to a bit about any potential opportunities and challenges when it comes to the electrode-electrolyte interface or the membrane interface that you're working on? So the membrane needs to, in our aqueous organic chemistries it needs to have high ion conductivity, excuse me, have high ionic conductivity for the monatomic ion that you're trying to pass and block the organics. And the job is a bit easier when the organics are bigger. In many of our chemistries, ferrocyanide is the fastest crossing species that you don't want to cross. And so we look for membranes that have really good selectivity against crossing the crossing of ferrocyanide. Our DHAQ original chemistry used a napion 212 membrane about ferrocyanide does cross through that. More recently, we moved to a hydrocarbon membrane that at mass production scale should be really cheap, like less than $25 per square meter. And it does a much better job of blocking a ferrocyanide. It's very thin, so then there are trade-offs about robustness and durability. And we may need to get, if this were to be commercialized, you might need to make it thicker and live with a little bit more voltage loss in order to have the robustness you need. That's to be determined. Great, Mike, so related to that point, you mentioned that the degradation can be mostly regenerated by oxidation or other chemical rays. Does this mean that there's also no significant increase in the impedance at the various interfaces as the batteries operated? Yeah, we've seen no significant change in the impedance. I should say maybe our longest runs have been maybe six weeks. We'd love to run for a year, but our students are in a hurry to publish papers and graduate and our postdocs need to move on to their next positions. And usually in the academic scale, low battery, a piece of tubing bursts or something goes wrong with the program and the potential status or something. So I don't think we have any runs that have gone longer than six weeks, but at that level, we haven't seen substantial increases in impedance. We haven't found, so the decomposing molecules when we lose them, they're not like polymerizing on the electrode and making it less active, for example. We haven't figured out where they're going and we need to do that. And certainly over years, if you're gonna accumulate schmutz from decomposing molecules, you need to understand where it goes or control where it goes. It may just go into a little particulates at the bottom of the tank. It may turn into CO2. We don't know and certainly to do any of this at a real commercial scale, you'd need to find out. So far, no red lights. That's great to hear. Let's take two more questions. So the next one is on your final slide, which shows the stability window of water. And there's been a lot of excitement these days in water and salt electrolytes and other related chemistry. What are some of the opportunities for employing higher stability window solvent for your chemistries? I am a firm believer in aqueous solvent. I have not seen any evidence that non-aqueous solvents can have nearly the conductivity or nearly the low cost that would be required to make a flow battery chemistry competitive. Now, I'm not a chemist. So I have limited vision, but in the range of my limited vision, that's what I see and that's why I stuck to aqueous chemistries. I should say, a lead acid battery gives you two volts and that's aqueous. So there's a thermodynamic stability window, but there are many things we use in our society that are not stable thermodynamically, they are metastable. A PN junction is metastable, but that doesn't stop us from building our entire information technology on it. So the 1.23 volts isn't, it's a true thermodynamic limit if you want something to last forever, but there's certainly plenty of opportunity for metastability and it's not crazy to imagine a two-volt aqueous battery if you can do it from lead acid. I think it's not crazy at all. Lithium-ion batteries are also metastable and so is the primary alkaline batteries as well. So no, I think you make a great point there, Mike. So the final question is on the technical and economics since you mentioned cost and I'm very pleased to see that both you and George are really thinking about it very carefully and integrating that with your innovation and the technology side. So there's one specific question and a more general question. The specific question was on your analysis of the vanadium redox flow battery. This is from Steve living at ExxonMobil. Have you, in your cost analysis, did you consider the residual value of vanadium? The residual value of vanadium would turn if you'll recall my slide on the trade-off, the replacement cost ratio of replacing your organic versus vanadium, which lasts forever. So if you want to just take the vanadium and recycle it and put it back into your battery, that's the equivalent of turning your project life into an infinite life project. So the lowest curve of the family curves that I presented then represents an infinite life vanadium project. And then that's the trade-off you'd look at there. You'd also need to know about the relative reconditioning cost of your system when you recondition organic. And I understand vanadium flow battery companies right now they plan to recondition once every 10 years or so. So there's some unknown difference in reconditioning costs that could be put into that analysis as well. Great, Mike. So the very final question, again, on technical economics, is the balance of systems. So one of the distinguishing feature between, say, lithium ion and redox flow, batteries is a substantial difference in the energy density, volumetrically as well as graphometrically. We're beginning to see large deployments of lithium-ion battery for our storage and we'll touch more on this in the panel discussion. But the similar deployment size for flow batteries is considerably larger. I know there's been a lot of analysis on this already but I wonder if you can share your own thoughts on sort of the chemistry costs and the balance of system costs when it comes to determine the total cost of storage and the importance of energy density in determining that ratio. Yeah, that's a very reasonable question. I'm not an expert in that. I have not studied the balance of system costs much. I know that flow batteries are out there and that they're competing and that you're not gonna drive them. There was a car company trying to create a bunch of PR about a flow battery powered car and I was very skeptical of that. But when you're talking about size, then it's gonna matter where you wanna put these things. If you wanna store them, if you wanna store energy in the middle of Manhattan, then space constraints are gonna be different than if you wanna store them at the base of your wind turbine or at the edge of your wind farm. So important consideration. I just don't have numbers for you. Well, maybe we can brainstorm this further. Let me thank you, Mike, again for the discussion and the presentation and invite George and of course, E, back to the podium. And we have about 20 minutes left and I hope we can have a spirited discussion on long duration storage. E, would you like to kick us off? Yeah, sure. I'm listening to both Mike's and George's talk is just fantastic and think deeper about the long duration, what could be the solution of choice right there. So I have one question. You're really coming from, Mike, I appreciate your study a lot. I mean, this calendar life, actually you can have good cycle life but calendar life can kill you if you don't have a stable molecules right there and your progress just amazing, the idea you put in. So this is for both Mike and George. If I look at the past history of lithium ion, since 1991 when lithium ion coming out, the cycle life, 200 cycles, calendar life may be about a year at most two years, that's it. And fast forward in 2005, that's about in up to 15 years. When I joined Stanford faculty, that's where I used 2005. Well, the cycle life is about 500 for lithium. Calendar life from the cell phone, you see probably three years, four years that that's about it. Now until now, right? That's about close to 30 years of lithium ion batteries. We have cycle life about 1,000 for most of the most batteries, maybe a little bit more, some of the chemistry go to 2,000, 3,000. And the 30 years learning, I mean, the calendar life of seven, eight years, maybe 10, depending on how you use it. So what I mean is right here about this lifetime issue, both cycle life and calendar life. What's your thought? How do we speed up this process and the development? And we don't want to wait another 20, 30 years to see that's the chemistry. We can have sufficient lifetime relevant to the grid scale storage. How do we do that? George, you mentioned your self-driving lab, well intriguing. So can we do something about it? That's a wonderful question, Yiyin. You put it actually very, very well. So when you talk about the cost of the battery, you have to include in your thinking, what's the lifetime? If suddenly it were twice as, the lifetime were twice as much, you might say, well, the cost has been reduced by a factor of two. And thinking like that, I think it's gonna come more and more to the four. So when we always try to get the cost down and we do that through the supply chain, we do that through better technology, through better manufacturing, there's always a learning curve, which right now for batteries is probably 18%, which is quite remarkable. It's gonna, that's gonna flatten out. And you'll be left with other issues. So you'll, and lifetime is one of them. And I think with, there are many degradation pathways in lithium ion, in flow batteries as well. Most of them are parasitic side reactions. Reactions that have nothing to do with storing or releasing energy, but use up the active material. So lithium reacts with something it's not supposed to and it's taken out of the charge discharge cycle. We can do a lot of that. And in my sense, that is one place where R&D ought to go. Another, and I have two more, one, second one would be recycling. Everybody knows we recycle lead acid batteries at nearly a hundred percent and we hardly recycle lithium ion at all. That's partly because they're complicated and it's hard to get. You don't wanna come down to the elements in a lithium ion battery. It's too expensive. And then you're faced with reorganizing those elements in the new lithium ion battery. So you wanna take out the cathode and reuse it without bringing it down to elements. Well, that's a promising direction. It's tough because it's a complicated system. I think with flow batteries, it's actually a lot easier. And it's easier to just drain the fluid and replace it. You could never do that with a lithium ion battery. Jay Caesar is looking and we're not very far in this direction yet, but we're looking at the idea of self-healing which is one of the reasons we like to work with oligomers. So you could imagine there's a bad monomer in an oligomer of let's say five or eight monomers, break the backbone, take out the bad monomer, put in a new one and find a way to have this happen automatically. So the battery will self-report time for a fix and somebody presses a button or maybe it just happens chemically and the polymer chains break and are replenished. So I think that's actually an easier problem. And then you mentioned, this is the last one, fascinating idea of the self-driving laboratory which we're using to discover organic molecules but you could use it to discover anything. And there are lots of other materials you'd like to apply that to within, let's say lithium ion batteries but you can apply it also to processes. So one thing we're looking at is can you simplify the synthesis of an organic molecule by trying lots of synthesis routes and learning automatically with in principle, no human intervention, a synthesis route which is either cheaper or faster or whatever your goal is, you could apply that to lifetime. Let's look at the degradation. Let's look at many ways of addressing the degradation and start ranking them and learn on every cycle how to do it better. My feeling, it's a very powerful technique. Remarkably, you don't have to understand why it works. You just have to know that it does work and I think artificial intelligence, machine learning, very good at finding things that work without understanding why. So it adds a new dimension I think to the R and D space. Mike, do you have comment on this question? Yeah, a couple of comments. One thing we need to do is find accelerated ways of evaluating lifetime because we're looking for lifetimes of years to decades and so you just can't make rapid progress by setting something up and letting it run for that long. So we're building now to do high temperature experiments where we systematically vary the temperature. There you have to understand the mechanism. So if you get shorter lifetime at high temperature, you need to understand the mechanism well enough to know if that mechanism is still what's going to be limiting your lifetime at room temperature or operating temperature, which might be a little bit higher than that. So this is a tedious kind of set of experiments to do, but we're setting up to do that now. And in terms of machine learning, I think there is a lot of potential there, but I should say when we started in this with Alain's team, we wrote proposals about machine learning and how the combinatorial theoretical chemistry was going to lead the way. Every single significant advance that we've made came from the creativity of a benchtop researcher with gloved hands. And once that insight was there, the theory was great for fleshing out all the associated possibilities and optimizing around that and understanding what's going on. But the insight in every case was a creative student or postdoc, experimentalist who had an idea and made something. So sooner or later, the machines are going to lead the way in that, but in our experience, it hasn't happened yet. Thank you, Mike. Will, pass back to you. Yeah, thank you very much. I want to actually expand on this discussion. This is something that's very near and dear to my heart on accelerating R&D. And Mike, I think you described it perfectly. For electrochemical energy storage technology, one of the biggest problems is that they last so long and it makes assessment and evaluation very challenging. So if I were to combine what George and Mike, you said, George, you talked about how to navigate design spaces. So the idea of using a robot really can help there by making thousands of compounds. And Mike, you talked about forecasting. So if you can forecast the behavior of the molecule into the future and take the two methods together, you can really decrease the R&D time. So I think this sounds extremely attractive. I wanted to see if I can probe a little bit deeper to understand in the area of flow battery, what is the bottleneck right now? Is it the synthesis time to make the molecules? Is it making the cell someone assembling the cell? Or as Mike, you said, is it just putting on the test bench and degrading the battery? What is the time controlling part of the R&D process right now? Well, let me offer my point of view on that. The synthesis, the conceptualization can happen quickly. Getting the synthesis to work and getting to adequate purity that you can get high enough yield and adequate purity that you can do some experiments with it and claim that the results you see are the results of a single molecule, that takes time. I mean, that can take a couple months for a student or a post-doc to do something like that. And once that happens, then there's the battery of tests we do and that can take less time. That can be pretty quick unless you're trying to do the long-term stability experiments and then those can stretch out as well. But I guess the big bottleneck in the research then is the synthesis. And especially because we're building up to having accelerated ways of doing testing. In terms of going out and making the world a better place, the bottleneck is mass production because we're coming up with these specialty chemicals and they're just the chemicals can be $1,000 a kilowatt hour at the specialty chemicals price point. So there there has to be a way of getting over the barrier to getting to cheap mass production. And our indications are example for a DHAQ factory here, you're talking about $180 million or something like that. And that's not an easy thing to come up with. Money is always rate limiting, right? Okay, so that's a good point. George? Yeah, those are great comments, Mike. And I think the question of where's the bottleneck is really the question we should ask. And hopefully we'll address one bottleneck, maybe remove it or reduce it and then go on to the next bottleneck because it's never just one. But to me, and I might agree with your statements about it's the creativity of a bench researcher which so far has made all the difference. And I think we ought to learn from that. And the lesson I would learn from that is just trying a lot of random things because I can think of them isn't the way to go. And some of our machine learning, it's beyond that point but it hasn't yet got to the point where it competes with a brilliant or let's say creative bench worker. I'm thinking though historically about computers playing games, they could never beat chess masters for the longest time. Now chess masters cannot beat a machine, period. And so we're not there yet with self-driving labs for sure but it's a very versatile idea. But the bottleneck, getting back to the bottleneck, to me it is the vast variety of organic molecules which are around. So even within one family and I showed a couple of examples of making 10 different derivatives of a basic molecule and they all had different either, well, they all had different electrochemical properties or solubility properties or the list goes on. And how to predict in advance even for a creative bench top scientist. That one's gonna work and that one's not. It's a real tough one. So there's, as you said, Mike, there's still this college synthesis or maybe it's just trying enough things even if the synthesis is easy. And to me that's the challenge. So I'd like to have either a good enough knowledge intuitive of what works and what doesn't work or just have a database that has that kind of information hidden in it and use machine learning and artificial intelligence to bring out the patterns and realize what works and what doesn't work. I think that's, I don't wanna say it's a bigger bottleneck than synthesis. Probably it isn't. But when it comes to synthesis, I remember the learning curves of solar panels and batteries takes about 10 years to get the price down to where it's really competitive. And I could imagine we're gonna have to look at that 10 years for some of the materials in flow batteries, the complicated organic ones, but it is possible to get the price down. And in the practical world, it really is the price that counts. We can talk about wanting to address climate change and reduce emissions. We should do that. But until it's comparably priced, either by policy or just by the... It's gonna be a tough one, I think. George, thank you very much. Maybe I can, since we're talking about cost, maybe I can get the both of you to comment on I think two flagship projects on the flow battery side and on the lithium-ion battery side for stationary storage in terms of energy stored. So, Mike, you already mentioned one, the vanadium flow battery in Dalian, China by a lonker and another one that comes to mind. It's a local one here in Monterey, California. So PG&E, Tesla and Vesta Energy are putting in a nearly one gigawatt hour system at Moss Landing. And I think first it's just tremendous to see that we're deploying at gigawatt hour now today. I was wondering if the both of you could discuss these real-life scenarios and examples of deployment for both the flow battery and the lithium-ion battery side and talk about what do you think we can learn from them? How would this drive the R&D? So this Moss Landing system is vanadium or lithium-ion? Lithium-ion. So this is Tesla conventional lithium-ion. One gigawatt, about one gigawatt hour. How many hours? Four hours. Four hours. So as George pointed out, lithium-ion is really good and at short discharge durations, it can reach four hours now. And the line George drew was six hours. That line may move up a bit over time. It depends on what happens to the trajectory of lithium-ion battery prices. So if people have asked whether lithium-ion batteries will become the Chinese single crystal, silicon photovoltaics of storage and just crush everything else because of the learning curve it's been able to come down. And for discharge durations up to some single digit number of hours, that's probably looks like it's probably gonna be the case. And what we have to deal with is what happens beyond that? What happens in the tens of hours to hundreds to seasonal storage? Yeah, I would agree, Mike. And a couple of comments on the learning curve which I think is critical and often overlooked. I remember 2017 and again it was Tesla installed 100 megawatt hour battery in South Australia. A big gamble that was the biggest installation at that time. And everybody was waiting to see how it works. And after a year, and I think remarkably, it seemed to be working pretty well. No, there were no problems. And that gave everyone a lot of confidence. Well, let's try some bigger batteries. Now we're up to a gigawatt. So and must landing will be another test point. Each time we break another sort of size limit, we wait and see what's gonna go wrong with it and how can we fix it? But it happened remarkably fast. So from 2017 to 2020, three years, we went from 100 megawatts to a gigawatt installation. Which means that we have a lot of confidence that it's really gonna work. And I can see, as Mike pointed out, if it's less than four hours, hard to beat lithium ion. It's gonna get cheaper. It's gonna get better. Nothing else is close to it now. So that's our battery. But when it comes to other applications, there's certain things that lithium ion just can't do. I mean, one that comes to mind is aviation takes about 800 watt hours per kilogram to fly a plane, a regional plane 600 miles full of say 50 or 100 people. And lithium ion is never gonna do that. So you've gotta look at other things like lithium oxygen, for example, it gets a lot of attention. And we've been talking about long duration. I don't think you're gonna make a lithium ion battery discharge for 150 hours, like form energies. It's a flowback water inorganic flow battery, like it does. So you may look at that battery, form energy's battery, and say maybe that's the next lithium ion quote unquote for long duration because it's got a head start. There's nothing else in that space at the moment. It's definitely gonna get better. And this one installation in Minnesota with great river energy will tell us, can it work? Is it gonna fail? Did we overlook something? How can we make it better? I think that learning curve, it's not a cost learning curve, it's just a technology learning curve. It's gonna be really fast for that battery and other long duration batteries. So I would be optimistic that my feeling is there are more applications that lithium ion can't do than there are applications it can do, and there's a lot of space out there waiting. Lithium ion is gonna cost three revolutions. Personal electronics, it was absolutely responsible for that. Electric vehicles, it's on the way at the personal car level, not necessarily at the heavy duty freight truck level or at the maritime shipping level, but it's on the way. And the grid is coming too, but not everything in the grid can be addressed with lithium ion, just as not everything in transportation can be addressed with lithium ion. So I think there's plenty of room for flow batteries and other technologies beyond lithium ion to fill that space. Can we add just a little bit to that? I might go ahead. Yeah, there's a huge space between say six hours where lithium ion may become just not competitive and 100 plus hours where form energy is targeting. In the tens of hours there, there's enormous space, enormous market and a lot of opportunities. And the second thing I wanted to say is whenever you assemble a very large amount of energy in one place, you have to worry about sudden accidental catastrophic release of that energy. And we've seen fires from lithium ion battery installations as well as sodium sulfur battery installations. And so safety becomes more and more important as you get to a larger scale. Yeah, let me inject a little bit on this topic. If I look at the levelized cost of storage per kilowatt hour, one third, it's only about one third, that it's the capex is your lithium ion batteries or the flow batteries, the installation. And then there could be a third coming from maintenance, air conditioning and things you put that you need to maintain. And then this third is actually financing. So, and from that perspective, that I could change the analysis, the equation. We mostly talk about the capex and then this all packs, there's financing costs. I think lithium ion is going up because of people's confidence is getting high with all this experimental project going. That will help the financing one, will help the financing, the lifetimes, confidence and so on. So would there be anything, let's say for the flow batteries that could have impact on how people think about operational costs and as well as the financing costs, to speed up the flow batteries, whether it's a quiz or non-a quiz solution. I'm thinking about it just for brainstorming. If I can comment, I think that's a great question to address and we know so little about flow batteries, not that we haven't studied them in the lab, we have, but out in the real world, the applications world. And I think we're gonna learn a lot. I've heard people who are lithium ion battery fans say, why would you want an organic flow battery? It's basically a chemical plant with all the safety hazards and other hazards that you have to deal with that nobody knows what to do with because there's not very much experience out there. But as the experience comes, I think we're gonna become comfortable and it's as much psychological as it is technical. After we become technically comfortable, we become psychologically comfortable. I always come back to the example of filling my car with gasoline. And if I think how much energy is running through my hand as I hold the hose and what it could do if it exploded, it's huge. And yet we all, everyone in the general public does that all the time and never thinks about it. We're psychologically very, very comfortable with that. I think we will get to that point with flow batteries. We'd have to get them out there. And I would, I'm not a policymaker, but I would think one of the things all government should do is find ways to encourage the actual deployment of technologies like flow batteries which are basically ready to be out there in sort of pilot form to see how they work. And it's a very good investment in the sense if it doesn't work, you'd like to know that early and you can go on to something else. And, but we're at that point in the energy transition, things are happening quickly and new opportunities are arising. So I believe that they ought to be tried in demonstration form. Yeah. Thanks, George. Well, we're circulating back to you. Yeah, I think we're at the end of our time here. So I'm just actually reading from the Q&A. It looks like our great colleague, Stan Winterham, would like to have the final word. So I will just read his comment verbatim. So Stan wants to make a comment on the four hour duration for lithium ion batteries. So he said, presently storage is limited to four hours because that is all the Federal Energy Regulatory Commission allows for cost recovery. So I think the regulatory consideration is significant even for lithium ion. So thank you, Stan, very much for that comment as well. So I would like to thank you both again, George and Mike for really great discussion. I'm sorry that there's not more time to do it but I think one of these days we'll all get together in person and do this much more extensively. So if I can, thank you, Mike. Thank you, George. And if I could have the slide please. We're done for this year. I wish everyone a very happy holiday, safe, state well. So returning January 15th, we will have two distinguished speakers to kick us off in 2021, which I'm very hopeful it's a better year. Jeff Don from Dalhousie University who was an expert on lithium ion batteries and Professor Yachal Horn at MIT who has been developing chemistry and fundamentals of various energy storage and conversion technology. So thank you again all for joining us today and have a safe remainder 2020.