 Good afternoon and welcome to today's energy seminar. Today we have a student award-winning panel. And to introduce and moderate the panel discussion, I'd like to introduce Dr. Richard Sassoon, who's the executive director of the Precourt Strategic Energy Alliance right here at Stanford. Richard's been here since 2003 as a PhD in physical chemistry and is an expert on physical and analytic chemistry, as well as by now, energy systems of all types. And he's a great friend of mine and a linchpin for many, many initiatives that have taken place since he's been here at Stanford. Richard, take it away. Thank you very much, John. Thank you. It's always a pleasure to be here. So welcome to another session of the energy seminar, where we showcase some of our top student researchers. Some of you may have been at previous energy seminars, and you would have heard some outstanding student presentations in the past. And this year should be no exception, so no pressure, guys. OK, so each year we conduct an annual Stanford Energy Student Lecture Series every summer. And the purpose is to not only showcase some of the cutting-edge energy research conducted by our students, but also help students better communicate key takeaway messages about their energy-related research to a broad technical audience. And so this past summer, we had 17 students give talks. And the judging panel selected the top three, and they became distinguished student lecturers. And we'll hear from them today. But before introducing them, let me just thank some of the organizers of the program. Juno Lee was the student seminar manager who helped with logistics. And Ms. Bernardo helped with recording the lectures and providing feedback sessions to the speakers. Joyce Lee coordinated the series. And Jenny Mill joined me in the judging panel. So let me just now go ahead and introduce the three presenters. The way we're planning it is we'll each give their talks one after another, and then we'll hold off on Q&A until the end, after the last talk. And then we'll have all three speakers take questions from the audience. So first, we will be hearing from Bralee Bourgeois. Bralee is a 50-year PhD candidate in the material science and engineering department and at the mentorship of Professor Gen Dion, where he studies plasmonic photochemistry. Prior to coming to Stanford, Bralee received his bachelor's degree in engineering physics from Tulane University. After Bralee, we'll hear from Sonja, Sonja Martin. Sonja is a fourth-year PhD candidate in the Stanford Sustainable Systems Lab. Her research centers around controlling various battery systems to maximize their decarbonization potential. Sonja obtained her MS from Stanford in 2022 and BS from Berkeley in 2020, both in mechanical engineering. And then, last but not least, we have Dhruv Sury. Dhruv is a first-year PhD student in the Department of Energy Science and Engineering under the mentorship of Professor Ines Azevedo, where he assesses pathways for decarbonizing global electricity production. Prior to coming to Stanford, he graduated from the Manipal Institute of Technology in India with a bachelor's degree in aeronautical engineering. So now I'm going to hand it over to the first presenter that will be Bralee, and he'll give his presentation on Plasmonic Photochemistry. Thank you. All right, thank you so much for coming to the talk today. I want to give a glimpse of the work that I do in my PhD in this talk, kind of broken into two parts. So what we work on in our group is trying to understand Plasmonic photocatalysis, a type of photochemistry, where we're hoping to use light to drive chemical reactions instead of heat to achieve both more interesting chemical selectivities and use electrified or renewable resources to drive chemistry instead of fossil fuel-based sources. So in the first part of my talk, I'm going to describe how we do that at the benchtop, how we look at chemical reactions, the chemicals that come out from these different gas phase processes we do. And then the second half, I'm going to talk about how we use advanced in situ electron microscopy to try to look at the atomic scale of these materials, understand how the atomic structure changes in reaction environments, and correlate this with the chemical activity we see, and try to understand how do these processes we see really happen at fundamental length scales. So to begin with this, why are we interested in chemistry in the first place? We've made these huge leaps and bounds in how we generate electricity through wind and solar. There's a lot of great research in how you store that energy through batteries. But a really important thing going forward is how do we create the materials we rely on in sustainable ways? How can we switch from fossil fuel-based resources for boat heating, as well as the inputs that go into these chemical reactions? So you can see here the carbon output of various different industries broken down by sector. And you can see chemicals and plastics right up here at the top. It's a really massive-scaled problem. And to put this into a different perspective, I'm showing the different fuels, plastics, fertilizers that we use per person on a yearly basis. So you can see there's this massive amount of chemicals that we rely on. And the support industry at this scale, we need facilities that look like what you see to the left. You have these almost city-sized factories filled with miles of stainless steel tubes and pipes, these building-sized reactors. But ultimately, at the heart of these chemical plants are these small, metallic nanoparticles called catalysts that help to drive chemical reactions faster and thereby using less energy. So what we want to study is these catalysts, how we can design them at these fundamental scales. But we don't only want to study these from the perspective of normal thermal-driven catalysts. We want to understand if we can use light as one of our inputs instead. We study a particular type of photochemistry called plasmonic photocatalysis. So what is a plasmon? A plasmon is a type of optical interaction that manifests when you structure materials around the lens scale of light itself. So typically, when we think of optical property, we think of something intrinsic. We think gold is gold, silver is silver. But it turns out that if you structure a material at these very small lens scales, you can engineer those optical properties just based on geometry. So here I'm showing silver nanoparticles. They're all the same material. They're just different shapes and sizes. And as you change the shape, you can get it to scatter blue, green, red light. You can engineer the frequencies of light, the wavelengths of light you interact with just based on geometry. Not only can you engineer what wavelengths you interact with, you can engineer how you interact with those wavelengths of light. So here I'm showing a technique that we do in our lab, a technique that's been used by many others called electron spectroscopy. And we're essentially mapping out how you can focus light at sub-defraction limits on the edges, the tips. You can engineer how you interact with this light as well. And in the past 10 or 15 years, scientists have been trying to understand, can you harness the energy from these nanostructured engineered interactions to drive chemical reactions further? So here I'm showing CO oxidation on a silver catalyst. As you turn light on and off, you can modulate how this reaction takes place. So when I started my PhD, we weren't really doing chemistry at that point in time. We were just trying to understand how hydrogen can interact with these plasmonic photocatalysts on really fundamental scales. But I wanted to understand, could we drive chemistry with this? Could we see it at the benchtop? Could we show some application for the thing that we've been studying on fundamental levels? We had to pick some chemical reaction to work on and we settled on ethylene production. Ethylene's the most mass produced organic compound on the planet. It's used to make polyethylene, which is the building block for most plastics. It's petroleum derived. When you produce ethylene, you always have this byproduct component, acetylene, that is created in that process. Acetylene acts as a catalyst poison downstream when you make your polymers. So you need to remove acetylene to very, very low concentrations in order to be able to make plastics. So the reaction we focus on is taking this acetylene component and turning it into ethylene. The challenges in this is that most catalysts will go too far. They'll create this product ethane. It's essentially a waste product. If you form this, you have to put in a ton of energy to be able to convert it back into anything useful. In industry, the way that they accomplish this is they use these palladium-based alloyed catalysts. Palladium is a very good hydrogenation catalyst. It's good for driving this reaction, but it's not very selective. It typically goes too far into the reaction. So their solution is they add in these more inert materials, silver and gold, to try to improve the selectivity of their catalyst. Luckily for us, silver and gold are inherently very good optical materials. So they have these strong visible resonances in blue, green kind of frequency. So you can see with the silver and the gold, it's about 450, 530 nanometer optical resonances, very much indivisible. So our question was, can we take this industrial catalyst, try to utilize this already-present plasmonic property and see what happens when we drive photo reactions on this instead? So like I said, when I started my PhD, we really weren't working on this, so we built this entire reactor setup. This is a schematic of kind of some of the hardware that we have into the system. We have these computer-controlled gas flow systems, lasers, temperature controllers. We use gas chromatography for a product analysis. There's a lot of different parameters that we vary here. So part of my PhD work has been developing this open-source Python library to be able to automate both the carrying out these automated photo reactions and the analysis afterwards. One of the things we've been doing recently, is trying to demonstrate how useful this can be. So we're attempting to work on this kind of prototypical reaction that's been done in the literature. It takes three years for kind of like the use case that people have shown before. We try to do this automatically. So this is a subset of the data that we generate. We're doing this measurement called the activation barrier measurement. We're trying to look at how sensitive is our chemistry to heat? So we set some wavelength. So here we're on our plasma in resonance. We vary our laser power. We change the temperature of our catalyst and we look at how much does it respond. In this case, we find that the more power we put in, the less sensitive it becomes to heating. We look at the same thing in a fixed power. We do different wavelengths. We find that on resonance, as we go towards this 530, 480 nanometer wavelength, we increase the barrier even further and eventually we can build up enough data that starts to take each of these slopes, translate this into an activation barrier on the right. We build up these maps of what is the barrier for this reaction? How hard is this reaction to run under different optical conditions? Usually a map like this takes several years of collecting all the data. These are really long and hard experiments, but we can do this in about a week of just purely autonomous reactions. So we're really excited about how much time we can save from automating experiments like this, but when we collected this data, we were disappointed to see that elsewhere, usually the barriers go down. You illuminate your system. You try to run this reaction. It gets easier whenever you shine light on it. For us, we found the opposite trend. The reaction becomes more difficult. We're really disappointed in this. We thought, oh no, this is terrible. But in our chemistry, it's not just about the reaction rates that matter. It's not just about how fast you can drive this chemistry. It's about what your products become. So even though we're slowing down the rate for this overall reaction, what we found is that we can actually target this selective kind of middle product, this ethylene over ethane, as we illuminate our system, we can get higher and higher selectivity. So we're creating more selective catalysts by using light to illuminate it instead. We wanna understand how this happens. So one of the experiments we can do is what's called a rate order experiment. We vary one component that we're bringing into our reaction. So here we have hydrogen and acetylene going in. We vary the amount of hydrogen. We look at how does the reaction respond? How sensitive is it to how much hydrogen we put in? When we do light on and light off, we see it becomes less sensitive to hydrogen. Something's happening with hydrogen whenever we put it in. If we put in acetylene and we vary how much we put in, we see there's really no effect to how sensitive it is when we use light or if we don't use light. So this is the question is, what is hydrogen doing? How is hydrogen interacting with our system, controlling what kind of products we get out? There's some precedence for this that hydrogen matters from the thermal catalysis literature. So in this experiment, these researchers use X-ray techniques to look at what is the structure of their catalyst? What does their catalyst look like as they're running a similar chemical reaction? And they find that if they use very low hydrogen pressures, they get very high selectivities for this reaction that we're running. If they increase the hydrogen pressure, they eventually get very bad selectivities. They look at the structure of their material and they find that they create something called palladium hydride. Hydrogen is entering into the atomic lattice of one of these crystals, is creating essentially a new material that behaves very differently, it has worse properties. So as known already, structure impacts chemical selectivity. We wanna understand, is there some structure change happening in our system? So we use a environmental electron microscope at Stanford with a specially designed optically cupped sample holder. We can shine light into the system through fiber optics while we flow gas in. And what I'm gonna be showing in the next couple of slides is we're watching as hydrogen moves in and out of this lattice by essentially tracking atomic motion and atomic expansion. So this is a schematic of what our experiment looks like. Here we're using these nanotriangle type catalyst. We shine the electron beam, we shine light, we flow gas. This is showing that we have these tip modes. We're shining light essentially on the tips of these catalysts. We're gonna excite this resonance and see what happens. So here I'm showing the thermally driven case and you can see this kind of contrast going across the particle. And what's happening is hydrogen leaves from out the particle. You're growing a pure palladium metal phase. Hydrogen is leaving from it, is exiting out through one of these tips and you're transforming what the composition of this material is. You're transforming a structure going from that unselected phase to a selected phase. If we use light to drive this instead, we can see first that we change the location where this transformation starts. And second, it happens much, much faster. We can drive this process much faster when we use light. And we can try to understand this geometry change by looking at electromagnetic simulations. Usually we find that these particles they're not perfectly symmetric. One tips usually sharper than the rest. We find that corresponds to where we have the highest light concentration in one of these tips. We look statistically at where are we getting these transformations whenever we drive this process with light and without light. So we find without light, it doesn't really matter. One of these tips is gonna be where hydrogen is leaving. We're kind of even statistics. If we use light to drive this instead, what we find is that there's a strong preference for this transformation to happen in the sharpest tips where we're concentrating light the most. So plasmas and light enhancement seem to be affecting how we can drive hydrogen out from these catalysts, how we can go through these atomic structure transformations. And ultimately what we're trying to show is that plasmas can impact the atomic structure. It's known that the structure impacts the chemical performance. And we're seeing that light can impact the chemical performance as well. So we wanna understand as there's some trifecta of events here in which all three of these are kind of corresponding properties. So with that, thank you so much for listening to my talk. And I'll be happy to answer questions at the end. Thanks very much. Thanks very much, Brani. And so next up is Sonja, who's going to talk on electric vehicle green charging with marginal emissions signals. Great, thank you for the introduction. As we've said, my name is Sonja. I'm a fourth year PhD candidate in mechanical engineering advised by Professor Ram Rajakalpall. And I'm excited to talk to you today about EV green charging with marginal emissions signals. So to start, we know that overall EV adoption will reduce emissions. Their EVs are better than conventional internal combustion vehicles, but because they have to charge, they will still release some carbon dioxide emissions from the grid via charging. And as you can see in this plot, EV sales shares in California and across the country in the world are growing significantly in California. There's legislation that says by 2035, all new vehicles sold in California should be EVs. But we also have legislation that says by 2045, the grid should be powered by 100% clean electricity. And as you can see, these are not the same years. So hopefully we'll have a lot of EVs on the road, but the grid will not be 100% clean. And so this brings us to the question, can we control EV charging to reduce these emissions generated from charging? And to do this, we implement what we call green charging, which is emissions-based charging control. And so to explain this, we have to kind of zoom out and look at the grid overall. So to power the grid, we might have some clean sources such as solar or some dirty sources such as fossil fuels, natural gas, coal, et cetera. This power goes through the transmission and distribution lines to get to some sort of charging infrastructure. And if you're an EV driver, you might choose to charge, maybe you need to charge based on price, based on convenience, et cetera. But we wanna think about what if we choose to charge based on the times when the emissions are lowest? So looking in this diagram, we're kind of highlighting these times when the grid is powered by either clean sources or natural gas generators instead of the dirty or coal generators. And so our research goal is to identify a signal that can inform driver behavior based on emissions rather than price or other factors. And it turns out this is a pretty complicated question which helps for my research. So to start looking at some results, we're gonna look at this plot. And the x-axis is looking at the number of added EVs to the grid. So starting with 100,000 on the left and the y-axis is added emissions. So we're first looking at uncontrolled charging. So just how drivers charge, no charging control. And this gives us a certain amount of added emissions. We also are looking at a few different simple charging control methods, the marginal emissions factor control and average emissions factor control which I'll explain in later slides. But we want to compare those control methods with the uncontrolled method. So for 100,000 added EVs, these simple methods work pretty well. You get some emissions reductions, but when you add more EVs up to a million, it's a little bit worse, but two million is significantly worse. And in this case for marginal emissions, controlling charging actually is worse than just leaving drivers as they were. So that poses a problem clearly. And so it brings us to our takeaway that controlling large amounts of EVs in a single group with these MEF and AEF signals does not reduce emissions and instead can accidentally increase emissions. And to understand kind of why this happens, I'm gonna have to back up a little bit and talk about the modeling that we used. So we start with a grid dispatch model which helps provide us the marginal emissions factor signals. So in our grid, we're dispatching resources based on cost. So we have on the x-axis generation capacity, we dispatch all the renewables and then the generators to meet the electricity demand based on cost. So cheapest ones first all the way up to the most expensive ones. Each of these generators also has a corresponding emissions rate, as you can see directly below. And this is not ordered as neatly as the cost. It's very noisy. And so if we're looking at an example, let's say we're at 65 gigawatts of capacity. We have a cost and we have a corresponding emissions rate. But then let's say you plug in your EV or 300 people plug in their EV and we need to dispatch the next generator. That has brought up the cost just a little bit but the emissions are significantly more. And this kind of brings us to the idea of the marginal emissions factor, which is what is the emissions rate when the next generator of the next generator that comes online when demand increases marginally. And if you plot this value over time, you get a plot that looks like this. So the green curve is the short run marginal emissions factor. You can see very noisy. The purple curve is average emissions factor, which is the ratio of the total emissions to the total demand. And this helps explain why controlling such a large group of electric vehicles is not effective because the signal is so noisy. It's very sensitive. If you're charging during these times where the signal is at its low point and you accidentally charge a little bit extra longer and now you're charging at a place where the signal is at its high point, that's going to inadvertently increase emissions. So ultimately we need to come up with a better solution. That requires some modeling. So kind of go through the methodology for how we get these modeling and results. So we need both an electric vehicle and a grid model to make this work. So for the EVs, we need some sort of added demand profile, which is the charging power profile. We also need some information about the electricity grid, which here we have the generator merit order that we saw on a previous slide. Both of these inputs go into our grid dispatch model and we output some sort of emission signal and we're using that as signals for charging control. And in my research, we're looking at a case study of WEC, which is the Western interconnection and ultimately these results reveal the need for more sophisticated charging control. So we also have zoom in picture on WEC and in the case study we looked at both historical data from 2020 and also future simulated data from 2030 in both January and July to capture seasonality. And in the case study, as you saw in the plots before, we're comparing different control schemes that use linear optimization with the different types of emissions factors. So for our EV data, we have data from Volkswagen, specifically Audi e-trons and 748 of them in California. And this data set has mileage, driving information, but most importantly, it tells us when, where and how long the vehicles are charging for and that helps us determine flexibility when we can implement control. And to implement the control, we're looking at analysis with the grid dispatch model and we have to run the model three times. And the reason for this is we want to compare the uncontrolled charging emissions with the controlled charging emissions. That's our major comparison point. And so to do this, we have to start with running the historical baseline demand through the grid model to get our first dispatch. Then we run the uncontrolled charging demand, which we sample some EVs. We have that power profile added to the baseline and we get our second dispatch. This gives us a marginal emissions or average emissions factor signal which we can use to run the charging optimization and output a controlled charging demand which we add to the baseline. And that gives us our third dispatch. And by subtracting off the reference or baseline emissions, we can compare again the uncontrolled charging emissions to the controlled charging emissions. And of course we want the controlled charging emissions to be lower. So now this brings us back to this plot that I showed earlier. Hopefully it makes a little bit more sense now. We have different numbers of added EVs in our simulation month of January, 2020. And again, with only 100,000 EVs, the control simple marginal emissions or average emissions factor works pretty well. But when you scale this up to 2 million EVs, you have a lot worse performance and might inadvertently increase emissions. And as I mentioned before, this is because the marginal emissions factor is pretty noisy. And so our solution to this is controlling EVs in smaller groups. And this brings us to our novel method called the cascading short run MEF method. Once again, controls the EVs in smaller groups. So I'm just gonna walk through a few results for that, starting with January, 2020. So this plot is looking at the information we saw before, just in a slightly different way. On the x-axis, we still have the different numbers of added EVs. We're looking at comparing the uncontrolled case with the two simple control cases, marginal emissions factor and average emissions factor. And you can see when we get to up to 2 million vehicles, the marginal emissions factor control does worse than the uncontrolled case. But here we're adding our novel method, the cascading method, which reduces emissions by in this graph, like over 10%, probably 15 or 20%. So, and it's a scalable method, as you can see, it still reduces emissions with 2 million vehicles. We also wanted to see, well, this is January, 2020, what's happening in the future grid because the grid power mix is changing. So here's just January of 2030. So the uncontrolled case already has lower emissions to begin with, which is a good thing, but still the MEF and AEF aren't working super well. The cascading method, on the other hand, still reduces emissions by 10% in this scenario too. And lastly, we'll look at the summer because of course in the US, or especially in WEC, we have a different grid mix in the summer. So in July, 2020, we have pretty similar results across the board, except for the average emissions factor, which is performing better in the summer. And when we look at 2030, we see that when you get to 2 million vehicles, the average emissions factor control actually is performing better than the cascading method. However, unless you want to use a different signal depending on the time of the year, we definitely recommend that the cascading method is used because that's the only signal that works in all seasons and in 2020 and in 2030. So this kind of brings me to our conclusion, which is that EV charging control needs to be implemented carefully. So we looked at different data, EV data, and our grid model with the e-tron data set, and this enabled us to run our charging control based on the actual driver charging patterns. And we first looked at simple charging control, MEF and AEF, which showed us that if you're looking at 2 million vehicles added to the grid, you might inadvertently increase emissions. So we developed this cascading method, which is controlling EVs in smaller groups, and this leads to at least 10 to 20% reductions in grid emissions across the board. And so we definitely recommend that EV operators and grid operators are looking at this control carefully, not using a simple scheme and controlling EVs in smaller groups. So with that, I'd like to thank my collaborators and mentors, Siobhan, Ram, Max, Mohawk, and Volkswagen who sponsored this project but also gave us this great data set that allowed for all of this analysis. So thanks for listening. Feel free to contact me and we can take questions at the end. Thank you very much, Sonja. Thank you. Okay, our final speaker is Dhruv, and he will talk on pathways to decarbonizing global electricity production. Thank you, Rishit. Hi, everyone. My name is Dhruv, and as Richard mentioned, I'm a first-year PhD student working with Inesh's video. And the title of my presentation is a question that we're gonna answer together in the next 10 to 15 minutes. How do we develop scalable pathways to decarbonize global electricity production? I'm very inspired by Vice President Al Gore. So following in his footsteps, if I was to give you a one-slide slideshow, it would be the slide. That's the troposphere, the lowest part of the atmosphere. And it's blue because of the oxygen that refracts the blue light. And the top of the troposphere isn't really far away. If you were to flip Stanford's roads around, it would take you about 45 minutes to walk from the engineering quad to reach the top of the troposphere. So volumetrically, it's not that much. And in this narrow belt is where all of the world's fossil fuel plants emit their carbon dioxide, sulphur dioxide, nitrogen oxide. It's in this narrow belt that we have all of the world's fossil fuel emissions. So now, the problem is not that this is happening, but rather how do we get from this scenario to a case where we shutter each and every single fossil fuel plant in the world? And that's exactly the question that we're solving for. We've developed a first-of-a-kind system where we analyze every single fossil fuel plant in the world. We construct an inventory of every thermal, coal, and gas plant. And then we determine its current cost of operation, which includes the capital cost as well as operations and maintenance cost. And in that very location, we consider how much would it cost to put wind or solar at that plant location and then compare the cost metrics? And is it cost-effective to do so? And so even for a second, if we leave out the carbon impact of changing these plants, if the capital cost of a new wind or solar plant is less than the capital and operational cost of an existing thermal plant, then it doesn't make economic sense to be operating that plant in the first place. And so using this framework, we determine the share of every single country's generation mix that can be swapped out from coal and gas to wind and solar. Before I show you some results, let me give you some context. This framework is really important because of two key growth rates. The first is the growth rate and primary energy demand. How fast is energy demand growing in the world? Or rather, because it's so heterogeneous, how fast is energy demand growing in every country of the world? The second key growth rate is the generation from renewables. And again, this is very heterogeneous by country. So these two growth rates are very important because in order for us to drive a sustainable transition, the second growth rate should be far more than the first or rather one-for-one displacement. But in the actuality, it's not. And even though there's a strong economic case for displacing thermal generation, it's not just happening in all parts of the world. So let's take two countries an example. In India, for instance, primary energy consumption increased by close to 9%, whereas in the same year, generation from renewables increased by 6.9%. So although a 2% difference may not seem too much for most of you sitting in the audience, 2% of India's annual electricity production is equivalent to the total energy consumption in New Zealand, Denmark, and Hong Kong combined. At the same time, let's consider China. China's primary energy consumption grew by close to 6%, whereas its renewables grew by 8%. So that seemed to be doing well. But as of last month, the country generated 14% more energy from coal. So it's already reversing these trends. So instead of just setting fossil fuel retirement targets or displacement targets, what if we consider a scenario where we take every single fossil fuel power plant in the world, analyze its cost effectiveness, and determine, is it actually economically and environmentally feasible to switch to solar or wind? And so as an example, if we consider a 500 gigawatt gas plant, in that very location, what is the capacity factor for wind? What is the capacity factor for solar? And what would it cost to generate the same amount of electricity as the original plant was generating? And then we determined the replacement size, the cost of the replacement, as well as the emissions abated. Now, one caveat here, a 200 megawatt turbine generating for one hour is not equivalent to 200 megawatt hours. The output is actually very non-linear just because of the intermittency of wind. The same is applicable to solar. And so when we consider the replacement size, usually wind and solar replacing a thermal installation are much more, as we'll come to an example. Just come into a methodology. This methodology has three data layers. The first is a global database of power plants. The second is the underlying techno-economic assumptions. And third is the regional wind and solar potential. So we first construct a global database of power plants. We draw on global as well as regional databases. And this consists of coal, oil, gas, hydro, all kinds of fossil fuel as well as renewable plants. The second key database is how much does it cost to, one, the capital installation of the plant, and two, how much does it cost to operate the plant? This graph, you see the levelized cost of electricity production from a range of different technologies. And then we overlay these two data sets to see how much does it cost to actually operate this plant on an annual basis. The last data set is capacity factor for wind and solar. The graph on the top shows the capacity factor for wind. The one on the bottom shows the direct irradiation potential for solar. And in both, the more yellow it is, the better it is. So in India, for instance, there's a lot of solar potential, not much for wind, apart from offshore wind. Let's look at this framework for a single gas plant. The one I picked is one in the east of India. It's a 135 megawatt gas plant that was commissioned in 2004. And in 2018, it generated about 600 gigawatt hours of electricity. From the EIA, we know that the capital cost of a gas plant is about $700 per kilowatt hour. And assuming it recovers this capital cost in a 30-year lifetime, the cost that it's yet to recover in the next 11 years is close to 57 million. We use discounted cash flow analysis to annualize this cost. And we determine for every single unit of electricity that it produces, the cost of keeping this plant online is $7.3 per megawatt hour. At the same time, from the IEA, we know that the natural gas price is close to $45 per megawatt. So adding these two, the total cost of capital plus operational is close to $52.3 per megawatt hour. Now in that very location, if we consider wind, wind has a capacity factor of 0.1, solar has a capacity factor of 0.16. And so to generate the same amount of electricity in that location, the size of wind and solar is sizeably more. Wind is close to $93 per megawatt hour. Annualized cost for solar is $38 per megawatt hour. So what we see is it's actually cheaper to decommission this particular gas plant, put solar in its location, and then operate it. In fact, the operator would earn about $9 million every year if they transitioned from gas to solar today. The last thing that I want to say is that we compute the ratio of the cost, the delta and the cost to the delta and emissions. And we calculate this metric. In this case, it's minus 14.25, which means by replacing the plant, you also obeyed emissions and it's economically feasible to do so. So what does this MAC curve look like for India and China? That red line represents the plant that we just evaluated. It's again negative because it's actually cost effective to do the replacement. The width of the bar represents how much cumulative CO2 emissions is abated by making that replacement. And if we do that for every single plant in the country, we see that about 90% of India's generation mix can be transitioned to low carbon or wind and solar at a negative cost, which means it's cost effective to do so, and the same for China. India has a lot of only replacements of solar, whereas China has replacements to both solar and wind. The significance of this MAC number is that there are a lot of technologies that will cost a significant amount just for a replacement, whereas if we consider a MAC number, it's gonna be negative, that means it's economically feasible to make that replacement in the first place. There are, of course, limitations to this study. There are three key ones in this. One is we consider firm versus non-firm power. We're considering a replacement of a gas or a coal plant that can always keep cranking with one that's intermittent, so we have to right size storage along with solar and wind. The second one is an in-situ replacement. Instead of a direct one-for-one replacement, what if the new wind or solar plant is located a few kilometers away? And the last limitation is, we're limited on data and generation emissions, so the more data we get available, the more accurate these assumptions can be. So I just wanna conclude that with this framework, we've shown that it's indeed very possible to decarbonize country-level electricity production and that it's economically feasible to do so, not just from a carbon imperative. Thank you so much. Thank you. Thank you, Dhruv. Do you want to stay up here? And maybe Brahli and Sonia can come up. And we'll open it up to questions. Does anybody have any questions? I think we start with students first. Yes. Hi, I have a question for Dhruv. It's for very interesting results. So I guess I got two questions for you. So when you were talking about replacing a factory with solar, is that the same size? Or just to re-put using the, so it is the same size? No, no, I don't have the same size. Oh, okay. And second question is that you're talking about scaling up those renewables like a lot. Do you think it's environmentally feasible to do so? Really good question. So on the first point, it's not the exact same size. So when we're replacing say a gas plant with solar in that location, we determine what is the capacity factor? And so always solar will be much high. The size of the plant will be much higher. On top of that, we consider coupling solar plus storage so that instead of just installing solar, it's not firm power. We consider, okay, over 24 hour period, what if we had to consider the same amount of generation? So it's not the same size. It's usually 1.5 to 2x more than the original plant. The second one is, I mean, the best thing to do here would be a LCA study, which is we do a lifecycle assessment. We consider what are the environmental implications? What are the carbon implications of going renewable? But that's a really good question. Are there questions? Maybe I should ask a question. I can ask one. Again, it's probably for Drew. We've had a couple of talks on the actual cost of building renewable plants in developing countries from World Bank, et cetera. So how do you think about the fact that the actual required rate of return is three or four times higher right at this moment? Is that part of your calculus? That's a good question, John. One thing we consider is many different discount rate assumptions. And obviously, the higher the discount rate, the more the MAC curve shifts. But as the risk index changes, I guess the MAC curve would shift. Great. The other tie in here is the sustainable finance people are actually working on getting the required rate of return lower. So that's a nice synergy. I don't know if you've talked to them. They might be able to give you some tips on how to kind of make that almost immediately usable by people in India in that regard. Other questions, comments? Well, I can ask one more. Have you actually talked to people in India or in WAC about these proposals? Yeah. It's just recent research on these two projects. So I spent time at COP28 in Dubai, and this is all I did. Just rally with policy makers to see can this actually be implementable? And obviously, the number of assumptions that go in have to be relaxed. And that's a big critical step that needs to be made. Sonia, in WAC, have you pitched the disaggregated margin of cost? Yeah, I haven't talked to people in WAC so far. And I think definitely would want to get information. I think one thing the model does not have is information about transmission congestion and seeing how that would play a role into this disaggregation. But I think once you kind of are modeling it with a little bit higher accuracy or at least checking those models, then that can make sense in the future as well. Great, great. Yeah, lots of synergies there. As you know, we actually have had student mentors go work for WAC for summers and whatnot. Another question for Sonia. So in your model, you've used up to two million EVs sort of in the market results. I believe California in 2023 is at about a million already. By 2030, I think we're projecting seven million or so. So how does that model work when we actually scale it up to those types of levels? Are there similar results? Yeah, one of the reasons that we did not model more than that was because in terms of WAC projections for power capacity, there is like right now not enough power plants coming online to do more than, honestly a lot more than two million. Also in the future, you're gonna need a lot more storage and there's not enough storage right now. From what we saw though, the results are similar. If you have, it depends on how many groups of EVs you're using. So here we modeled 20 groups, but if you keep the group size smaller, if you look at the results, if you have group sizes of a hundred thousand vehicles or less, that tends to work well scaled with the cascading method. So presumably if there's enough capacity and enough storage in the future, if you're using this group size of a hundred thousand, this can scale up to well over two million EVs. I have one for Bryder. So since there's all this talk about economics and what not, have you looked to the point where you have a sense where the economic impact of your research might not encourage a specific project? Recipient projects and how much they might cost? Yeah, there's maybe one startup company right now that's doing this particular type of photo catalysis. One of the things that people often kind of criticize is a lot of these are, all the research and this is LED based, laser based. We've got like in theory, you can make these kind of like black body absorbers, but usually people study systems that kind of absorb a narrow range of light. So maybe not best applied to like solar. So when we think about like scaling up, a lot of people are thinking about can you have like high efficiency LEDs? So there's a company trying to do that for ammonia decomposition to make hydrogen. And they're thinking about some other like methane reforming for hydrogen. So a lot of it is geared right now towards hydrogen production and starting to come online as startup companies with some decent funding. But yeah, it's kind of the name of the game for I think a lot of the chemistry stuff right now is there's a lot of stuff you can do if you have hydrogen, where do you get hydrogen from? So on the same note for Bradley, have you looked at like the only one hydrogenation process? Does the light help with that? The olefin hydrogenation? Yeah. With CO2, underwater. Oh, you mean like CO2 hydrogenation into... Yeah, so we have two projects. One is on CO2 and hydrogen starting with molecular hydrogen as one of the feedstocks and doing the same exact procedure with that. And then we have a new thing that we're kicking off that's using water as a feedstock. Those are really hard reactions, right? Usually you need to go to a lot higher temperatures to be able to drive something like that. There's a little bit of work in the field that has shown some positive results. A lot of it is using UV light. I think mostly it's governed by higher energy absorption, like these high-bang-out materials, not really like plasmonics. But yeah, there's some progress in it. It's a very difficult procedure. We're trying to combine it right now with electrochemistry. So you can do electrochemical CO2 and H2O to make some hydrocarbon output and then try to use plasmonics to guide how selective it is for what hydrocarbon you produce. Any last questions? Richard, do you have any final words or questions that you want to ask? I can ask you all the questions. So your model was very interesting. What you didn't take into account was the timeframe it takes to do all these replacements. You said, if you replace one plant with another, you can get an improvement here. But if you were going to extend your model towards time and now it's going to take decades to do this, could you comment on that? That's a really good question, Richard. And something that we do is we consider, in the US, we have the energy information administration. They come up with an overnight capital cost. So if we were to shut this down and put up a new plant, what would happen? But if we consider the temporal context, what would be most cost-effective is if we went from most negative to positive in the order of Mach number. And during the time that it takes to deploy those plants, keep recomputing based on the discount rate or the actual construction cost, the levelized cost of energy. But there can be also an optimization coupled into this framework as well, but that could be another step. Okay, thank you. Great, before we shut down, this is our last Energy 7R site for this quarter. We will start up again in two weeks on April 1st. Another note is on April 8th, we will have Roland Horn, the director of the Precord Institute, talk about geothermal. And after that, we'll have a celebration of the 400th Energy Seminar, believe it or not. Think I've been involved in almost half of those. So be there or be square, as they say. So with that, I'd like to sincerely thank Brylee, Sonya, Drew, and Richard for great presentations and the audience as usual for great questions. Thank you very much. Thank you. Great job guys.