 Okay, I'd like to welcome everyone to the Purdue Engineering Distinguished Lecture Series. We're very pleased to have today Professor Ted Sargent of Northwestern University. I'm Sang Kim, Head of Chemical Engineering School, and this distinguished lecture is co-hosted by the School of Chemical Engineering and the College of Engineering. And with that, I'd like to first announce, I've been asked to announce that there is an overflow room in the Forney G140, the large auditorium, so if you would like to be seated while you're watching the lecture, you can also do that from Forney Hall as well. So with that, I'd like to introduce Interim Dean of Engineering, Mark Lundstrom, to the stage. Mark. Thank you. Thank you. Thank you, Sang. And good afternoon, everyone, and welcome to our first Purdue Distinguished Lecture of Fall 2022. So since 2018, the Purdue Engineering Distinguished Lecture Series has brought thought leaders in academia, in business, to campus to discuss the grand challenges and opportunities in their field, and today we are pleased to welcome Professor Ted Sargent. Professor Sargent's lecture, How can research in solar energy harvesting and electrified chemical synthesis contribute to defossilization? This lecture will set the stage for a panel discussion. Professor Sargent is the Lynn Hopkins Davis and Greg Davis Professor at Northwestern University with appointments in the Department of Chemistry and Department of Electrical and Computer Engineering. Before joining Northwestern about a year ago, he was University Professor at the University of Toronto and also held the Canada Research Chair in Nanotechnology. Professor Sargent's work to unite chemistry, physics, and engineering has attracted wide attention. As one example, his publications on these topics have been cited more than 100,000 times. As an electrical engineer myself, who stayed in his own discipline but who has collaborated with several chemists and chemical engineers, I know what a special person it takes to work across these disciplines and I know how special their contributions are. Would you please join me in welcoming Professor Ted Sargent? Well thank you. It is wonderful to be at Purdue and it's really gratifying to see such a nice crowd gathered today. I'm so pleased that it's such a nice audience with us today. You know, as you were preparing for today's event and were tweeting about the fact that I was coming, I was really gratified to see that I got maybe 60 or 70 or 80 likes and retweets on the tweet that LaTan sent out and that might have been the largest number of tweets and retweets I'd ever seen regarding anything I was involved in and then just yesterday I saw a tweet of a man on a bicycle with a baby Capybara in his little container on the front of the bicycle in Berlin and just in a single day he had accumulated something like 8,000 tweets or something like that. So it gave me a sense of perspective which is always helpful. Yeah, so I'm thrilled to be here and I'm going to talk in kind of two chapters. The first about finding ways to generate more low carbon electricity and trying to really accelerate our path towards more low carbon electricity and then the second one I'm going to ask what should we do with it? So what should we do with what I hope will be an abundance of available and low-cost electricity that has a low carbon footprint? So on my first topic and I suppose this kind of covers both I'll first just acknowledge that while we're trying to reduce our carbon footprint we're going to do it in the context of a huge amount of economic growth and inevitably with economic growth comes energy consumption growth. So this isn't just a problem of finding ways to generate the same amount of electricity in novel ways. This is actually a challenge of producing considerably more energy to meet growing demand and to eventually zero out our net carbon emissions. So this is a really big grand challenge. In fact the ethical imperative of doing this and doing this well is even more obvious when you look at the fact that the growth is coming from non-OECD nations. So it's coming in parts of the world where if we succeed we'll be continuing to help in the advancement of quality of life which again is necessarily accompanied by growth and energy consumption. So this is actually a really important societal challenge that we're embarking with and we need to meet the societal obligation in a way that is coherent with obviously our climate obligations as a people. You of course know well about where some of these climate imperatives come from. It's the fact that with CO2 now exceeding 400 ppm in the atmosphere we are making this significant contribution to global warming and that is having all sorts of impacts and it's even having greater impacts on the people who can least afford to suffer from those impacts. Now when you look at our plans and our pledges you'll see that the world community has made progress in articulating pathways to reduce net CO2 emissions. My vertical axis here is gigatons of CO2 equivalent so you'll see that where we are now is around 40 and we aspire to moving towards a net zero scenario by 2050 but existing pledges and available technologies don't get us there. Now you could think of that as frightening and I would encourage you to do so but if you're somebody who's doing an undergraduate degree in engineering or a graduate degree or a postdoc in engineering or the physical sciences you could think of it as bracing that we're at the heart of solving this huge societal and global challenge. In fact if you're somebody who's interested in making money you could also think of this as a big challenge because when you think of the size of the global energy industry today roughly five trillion dollars depending on how you count this is inevitably going to be displaced by low carbon solutions over the coming two decades which actually isn't all that long it's maybe the first half of some of your professional careers so we will have figured out this problem and there's five trillion dollars a year in new classes of energy revenue to be created per year five trillion per year so it is a big problem but it's going to have huge potential a huge potential for individuals like folks in this room but also if you think of for nations if certain nations get there first and are able to be right at the vanguard of the science and the engineering and the application of the translation of these technologies then their economic benefits will be tremendous as well. So specifically where do we need to make impacts in order to move in the direction of net zero? Well the simple answer is everywhere and there are probably things on this list that aren't the things that you're thinking about like behavior and avoided demand you may not be thinking about the fact that we can create incentives and we can use behavioral economics in order to try to drive down a component of demand where that demand can be reduced. Energy efficiency there's technology opportunities here there's also utilization opportunities as well. Hydrogen has to be a major part of this equation industrial hydrogen is already a big component associated with CO2 emissions and we have reasonably mature technologies to address this and now we need to scale those and make them even more mature. We need to bring online a huge amount this is the green the light green downward arrows of wind and solar and in order to do that we need to reduce any friction related to the adoption of these technologies. These frictions can include the fact that in parts of the world it's still too expensive to use these sources that everywhere in the world these intermittent sources need storage if they're really going to become part of a robust solution. There are rare materials inside some of these technologies and so there are opportunities to think about where we will find those materials or where we can replace them with alternatives and then the last part of my talk today is going to concern the orange rectangle which is carbon capture and utilization which is going to be a requirement given that some of these sources of CO2 emission are hard to obey it's going to be a requirement to get us the rest of the way to net zero. But I'm going to start with some of the interesting challenges in the area of producing more low carbon electricity from the sun and here I just want to mention first that the sun is a very very generous resource if you were to compare the amount of solar energy hitting the earth in a year which is my big yellow cube it is huge compared to the total proven reserves of coal. So we have no excuses from a sort of fundamental perspective in terms of the energy flux that's arriving on the earth. But obviously this is a distributed resource it's spread out an area and so our challenge is to find ways to be efficient enough in capturing what energy there is and cost effective enough in order to start to take on a significant portion of our energy supply from this kind of source. Now there are very ambitious projections for how much this sector needs to scale but in order for this sector to scale in the ways that meet even just its portion of the decarbonization challenge there's a huge amount of work that has to be done to further reduce the energy intensity the capital intensity and the cost of producing electricity from solar sources. Now what's amazing is that empirically there's actually been an impressive downward tick in the cost of solar cells and of solar modules and this uptick in the growth of utilization of solar that's just happened in the last five or ten years is directly connected with that. But if you look underneath it it looks like this may be a downward tick that we can only get once it's related to a learning curve traced to the manufacturing technology of these solar solutions and so you see here that there's been this dramatic decline in what's called the per piece learning. So this is a learning curve component that is related to better understanding the scaled manufacturing of these solar cells and it's been great it's had a huge and rapid impact but the impact coming from improving energy efficiency has been more limited and many of us have come to believe that we need another step change in the efficiency of these solar cells to make a further striking downward tick to further increase adoption of solar technologies. Now this is an eye chart and I apologize for it and you don't need to absorb at all everything on this eye chart. The general trend here that we're looking at is that over decades solar technologies from their inception took a while to improve in performance there's a lot of deep science of deep engineering semiconductors interfaces crystal growth materials processing but that these have substantially saturated they're coming to within about 80% of the thermodynamic limit of how well a solar cell based on one absorber so based on one choice of semiconductor composition can do. Now then something happened about a decade ago that really took the world by storm and solar which is a new class of materials perovskites perovskite crystals have been around and understood for a long time of a particular subclass of this crystal class called organometall halide perovskites but for obvious reasons I'm just going to call them perovskites for the rest of the talk. These came on the scene in a decade they caught up in performance to what had taken five decades to achieve and then more strikingly recently when you combine them with silicon they started to overtake the legacy technologies so this has really gotten people interesting so what are these perovskites up to? Well the crystal class is ABX3 broadly speaking and when I said organometall halide perovskites it was that the X is a halide sometimes it's a mixture of iodide and bromide or sometimes even some chloride and these materials are very versatile that's one striking property so you can see here I'm showing that they can emit colors of light that span the full rainbow and that even go into the infrared that ends up being a very powerful idea in making solar cells that can more fully harvest the full solar spectrum. Another thing when we in our group got early into this field this is about seven years ago is that these materials seem to like to be pure they seem to like to not have sites defect states within their band gaps where electrons and holes can come together and can recombine and lose energy but instead they seem to like to clean their own band gaps and ensure that between the conduction band edge and the valence band edge there is a very very low density of what are called deep electronic trap states and we studied this using some really kind of basic materials growth and photo physics it's actually the kind of stuff you might have seen at Bell Labs in the 1960s in the 1970s where we were growing single crystals of these semiconductors in the liquid phase and using classical spectroscopies to study how clean their band gaps were and got results that even those these even though these materials were being made around room temperature they were being made in an open lab in these single crystals the trap state density was in the same ballpark as in the semiconductors that we make in clean rooms at high temperatures with incredible purity of processing so there's this kind of hint that what was helping these perovskites succeed in solar was that even though they were made under these rather familiar kind of room ambient conditions they were producing very clean sets of densities of states but the other thing they were doing is because they were processable from the liquid phase you could make this semiconductor into an ink well you could deploy it in all sorts of ways as well you could spray it you could do spin coating which you may not want to do for solar cells but if you want to integrate a new class of optoelectronic material onto a silicon wafer you might well want to spin code it if you want something patterned you could inkjet print it and if you wanted to figure out how to make a solar cell kind of the way you'd print a newspaper you could doctor blade or roll to roll print it and so this capacity to make from an ink look like and still looks like a really powerful way to really scale these things up in fact when you envision scale you know a midget imagine an old-fashioned newspaper printing process many of you out there believe that newspapers only exist on iPads but long long ago before you were born there were these pieces of paper that could roll very very quickly and could transfer ink to them and so this is a very scaled very low cost per area technology and actually very perfect printing technology now what gets really interesting from the efficiency perspective is when you couple these into layered structures where you're no longer limited by that idea that I mentioned before that only one semiconductor junction is active but where two or three are active and so the solar cell starts to become a kind of rainbow cell on its own and then from techno economic projections this is where things get interesting from a driving greater supply of lower cost low carbon electricity perspective folks have modeled how much can we reduce the equivalent cost of electricity from solar when we move to these technologies that are more and more efficient you see going from the left to the right across here higher and higher efficiencies and then in the final one a reasonably high efficiency but combined with a really low manufacturing cost and you'll see here a projection for a further reduction relative to the existing kind of state-of-art silicon technologies being achieved by these projected perovskite performance levels and one of the challenges that's really needed the attention of applied physicists and chemists and engineers of all sorts over the last couple years is the fact that these perovskites while promising for performance are not yet robust enough and so you can see the problem being depicted on the graph on the left which is that you start this thing operating and then after 350 hours what looks like a very promising initial efficiency has declined and what matters if you're going to achieve these and these electricity cost targets what matters is the average deficiency over a 25 year lifetime and also what matters is the predictability of that 25 year lifetime which the solar people call bankability the ability that you will be able to invest in a solar asset and make money off of it in a robust predictable way so this kind of decline in performance is not okay and so that has led to a lot of activity and a lot of effort in the perovskite field in making more and more robust solar cells it's also led to a lot of scientific investigation of where these limitations and performance are coming from so here I'll just highlight some of our work from earlier this year we made a device that actually flipped upside down the perovskite solar cell architecture that allowed us to deposit it onto a completely inorganic whole transport material and nickel oxide whole transport material and that allowed us to pass one of these reliability tests there's a bunch of standards that have been established in the solar sector to pass the one that's an accelerated lifetime test one that occurs at 65 degrees C and to report the absence of degradation over a thousand hours now you may say thousand hours is in twenty five years and indeed it's not but these accelerated lifetime tests allow us to extrapolate or to project forward for how reliable these technologies will be when we do operate them under more normal conditions for more extended periods of times in fact what we really need to do is to build reliability physics models using these accelerated lifetime tests to be able to project and extrapolate and that's what the field is now starting to do I've mentioned a couple of times this idea that solar performance can be enhanced when we no longer rely on just a single band gap and I want to explain where that comes from and then where the opportunity to make multiple colors of band gaps derives from so if you want to look at the solar spectrum which is the very broadband orange distribution here spanning as you can see the visible the near infrared and the short wavelength infrared if you place the choice of band gap let's say at 620 nanometers at about two electron volts this has been done here but we'll be able to absorb those colors above that band gap so to the blue of that so to the left on this spectrum but we'll leave aside those that are under unabsorbed you can see a lot of potential current density that's being left on the table when we place the band gap in this position and so you say well let's move it all the way over to the right let's absorb all of these photons but then the voltage that the solar cell can deliver will be limited by the value of that energy gap another way to say that is that all these juicy energetic photons in the visible are going to very quickly have much of their energy lost to thermalization inside the semiconductor and so this leads to a compromise if you make a device using a single junction approach and it leads to this upper bound of a little above 30% for the solar power conversion efficiency but if you add just even one additional band gap and then you have to renormalize the first one so you need to choose these two band gaps optimally you can extend this limit from 43% from 33% to 43% and so there's this very striking possibility that becomes available when you do more selective photon harvesting in a tandem device so in some of our work we've looked at doing this with silicon solar cells now silicon solar cells actually have this great advantage which is the beauty of the fact that we can etch silicon along certain crystal planes to make pyramids which function as anti-reflection coatings but if you're somebody trying to put a perovskite active layer onto these one of these rough topologies it actually comes as a challenge and so in this study we worked with colleagues to first figure out how could we keep the pyramids very optically effective at their anti-reflection job but also make them compatible with the solution processing of the perovskite active layer so that was the first optimization was to find a suitable pyramid size and then the next step was to build an integrated device with this technology that evenly apportion the photon current between the silicon which is what I show on the top right here in the blue and the perovskites and it needed to current match because these were being connected together in series we had to equalize their currents and optimize the choice of band gap in the perovskite layer and as a result of that we're able to advance the efficiency of these solar cells by combining the two active layers together and as you can see here also achieve an improvement in the sustained performance so the continuous reliable operating performance of the devices one of the areas this is going to now is bifacial solar cells the legacy approach the silicon approach is now not just absorbing the light that's coming in from the top directly from sunlight such as what's shown on the left image here but they're also using reflection and backscattering could be off of especially chosen material or it even could be off of sand or could be off of snow and they're taking in that light from the backside as well and when they do this they can achieve an increase in the overall energy harvesting yield that's accomplished and so the field has moved towards how can we now make tandems of silicon and perovskites or perovskites and perovskites with one another that leave the backside open and transparent and that can still achieve the requisite current matching requirement even as the angle of the sun varies and the spectrum of the sun varies so a really interesting challenge for engineers but ultimately what these can do I talked about yield which is my vertical axis here ultimately what these can do is they can achieve a relative increase of as much as 30% in the yield to solar energy flux that's accomplished in these tandem solar cells by levering the scattering and reflective properties of the underlying surface that the solar cells are installed upon through these transparent back contacts and through the reoptimization of the solar cell itself and so it's these kinds of strategies that again are bringing us back towards these higher and higher efficiency cells that are lowering the cost and so for example in recent periods we've now started to move towards demonstrating you can play this tandem game just with perovskites just with ink based materials so if you're concerned about the capital cost of building a conventional solar cell factory based on silicon or if you're trying to start a solar industry in a place that doesn't have one at scale you know now you've got this printable technology that allows you to go create a next generation solar industry on a much more capital efficient model so recently we've demonstrated that in collaboration with colleagues this is with professor hi-ran tan that we could achieve the 26% solar power conversion efficiency threshold with one of these two-term and turtle mental two terminal tandem technologies so to end the first of two chapters in my talk today I'll just summarize by saying that the I think the most interesting challenge that still needs our primary attention in the solar materials and technologies field is the efficiency challenge I think we still have more work to do and the ways to tackle the challenge are are many in number they certainly involve architectures maybe the regime of electrical engineers and applied physicists traditionally but they involve new materials designs and synthesis and materials processing and so they engage for sure the chemical engineer but very much also the materials scientists even the mechanical engineer and as I'll say a little later in my talk to try to accelerate the discovery of these materials they engage the computer scientist and the computer engineer and the data scientist as well so it's really quite a broad spectrum of research and applied engineering research and r&d tech topics reliability science is going to be a really important frontier area for this field and I am choosing my words carefully here I mean maybe reliability science sort of sounds better than reliability or durability but what I really intend is that the only way to get these technologies to have market and climate impact quickly is for us to understand the underlying mechanisms of degradation so we can go remedy them or so when when a solar cell dies in the field we can go tackle the problem we can tackle it swiftly and we have a basis in science of how to do that and for sure we need to work on scale and pass to further scale by lowering materials and processing costs all right so I'm now going to take a sip and change tracks and move over to the question of what can we do with what we hope will be this ever-increasing abundance of low-carbon electricity now first I'm going to say you know when I kind of got into this area you can imagine because I've been working in solar and opt electronics for a while you can imagine I might say well how can I use my expertise in dealing with photons and semiconductors to go tackle this other problem which is of taking carbon dioxide and turning that into say fuels but instead at the time I really felt strongly that if I was going to go tackle that I would actually be quite agnostic as to where that electricity came from if I was really committed to let's say making hydrogen or making chemicals or making fuels using this electricity all I would want is more of it I'd want it to be low in its carbon footprint and I'd want it to be as cheap as possible and so I kind of decoupled these two halves of my research activity and said here I would just focus on once we have this electricity what can we do with it to try to accelerate climate impact and so as I was introducing the topic I already pointed out that being able to synthesize using electrical power hydrogen is going to have a huge impact and being able to capture and utilize CO2 will also be very relevant and in particular I had shown in my net zero plot that there are certain sectors that are going to be very hard to decarbonize one example that's easy to understand is that I guess we don't know for sure yet but it may be hard to figure out how to make airplanes that run either on electricity and batteries or that run on hydrogen or they are at least accepted is running on those technologies especially on the hydrogen so it may be that aviation is going to be a very hard to decarbonize sector and of those you know fairly limited now we hope CO2 emissions that remain some of those will be kind of eaten up by the things where we don't have alternatives and so aviation may be an example of that and so it's interesting then to ask are there ways to offset if those are kind of inevitable emissions are there ways in which we can offset them so one potential strategy to do that would actually be to consume CO2 in the course of synthesizing those fuels so this is that CC us acronym that I mentioned before we have lots of sources of carbon some of the ones that are maybe a little bit more short-term imaginable are say the emissions from existing factories that have a CO2 footprint and that are at a point source emitting CO2 so we know where it is we know where I need to attach our carbon capture infrastructure to it it's at a reasonably high concentration which makes it a bit more of a short-term approach long-term and I would even say the subject of controversy at least the subject of some speculation as to whether we can solve all the challenges is direct to our capture so CO2 is too concentrated in the atmosphere relative to what we would like from a climate point of view but it's not very concentrated at 400 ppm so this is really fishing for quite rare molecules and at the moment it is a very energetically expensive approach but this idea of direct air capture has a huge amount of interest because you can imagine we could actually it's easier to picture how you could actually once you captured that CO2 in a cost effective and an energy efficient and a naturally low carbon way once you did that and if you chose to sequester or you chose to build this into building materials which is if they're long-lived is another form of sequestration you can really start to imagine how you could have a beneficial impact on the net CO2 in the atmosphere so this motivated us to pursue an approach to our research that would again take CO2 from some stream would combine it with renewable low carbon electricity from one of the sources I mentioned today or another from hydro from nuclear and water or proton source and focus on the technology to electrochemically reduce the CO2 into fuel thereby closing a loop or potentially into a chemical that could make it into a long-lived material in order to fully lever this increasing abundance of electricity the technologies to do this are and of course the basics are known these are electrolyzers and the field of water splitting electrolyzers to make hydrogen has seen tremendous advancements it's got more work to do and Scott remaining challenges but it's certainly way ahead and it's shown some very compelling results we would focus instead on CO2 electrolyzers where in a system with you know fairly strong analogies we would perform on a cathode electrochemical reduction of CO2 to say a fuel or a fuel precursor and on the anodic side initially we probably produce oxygen through the oxygen evolution reaction maybe more interestingly down the road there's a lot of reactions that occur in industry that desire oxidation a lot of ethylene becomes ethylene oxide and a lot of propylene becomes propylene oxide or propylene glycol so there are a lot of oxidation reactions that we could also imagine seeking to pair this with now one of the challenges that we in the whole community have immediately faced is that there are many options so CO2 can be reduced to many different possible products ranging from C1 like carbon monoxide and formic acid to multi-carbon like C2's ethylene and ethanol and many others and the pathways through which all these reactions happen are not completely understood and how certain catalysts and choice of electrolytes in the overall environment influence these processes are not fully understood especially when they are electrically powered but what we did know is that we would need to make a lot of real engineering progress we would need the Faraday efficiency which is the yield on electrons so if I put a hundred electrons in how many of them go to making propanol if that's the business I'm in so we would need to improve that yield on electrons we would need to improve the intensity because without further improving the intensity towards near the levels that water splitting electrolyzers work at the capital cost of these systems would be too large we would certainly need to improve energy efficiency a component of which comes from this electron efficiency the Faraday efficiency but another component of which comes from the voltage efficiency we would have to scale down the voltage to much closer to what thermodynamics mandates than where we are today and then eventually we'd have to engage in the same crucial area of durability science to bring durable products to the market now those are challenging problems and it was a tall task at the same time to look on the right side these are problems in material science that require multi component materials engineering and the field of material science and nano materials had already given us incredible diversity of nanostructures and placing materials on supports in fact the field of thermal catalysis had and and it continues to lead in this area with also with understanding these effects and the effects of how a nanoparticle on a support behaves differently than in the wild shape effects the effects of facets and confinement alloys and cores and shells so there's incredible diversity of degrees of freedom available to us but perhaps what was limiting was more predictability how would the synthesis of a particular composition of a multi component material actually drive performance so actually the first step that we took in the field was was to try to kind of get a stable baseline just to be able to make a co2 in this case to ethylene one of our kind of model molecules a co2 to ethylene electrolyzer that was stable just for a couple of hundred hours that achieved constant performance where we could make these reproducibly every time and only once we did that would we be in a position to then start to systematically vary the catalyst composition so this we did on a gas diffusion electrode something that allowed the gas phase co2 in through the back allowed it to come into contact with a solid heterogeneous catalyst this one's showing copper nanoparticles and for that to be in contact with a liquid electrolyte so really a three-phase boundary being formed between the gas and the solid in the liquid phases and then we started to explore so I'll give an example of one of the things we've been exploring this idea came really from working with colleagues our co-authors here are Teo Agape and Jonas Peters at Caltech and Teo and Jonas had already done really brilliant work in understanding how to improve heterogeneous electric catalysis and how to use molecular strategies to try to augment the efficiency of these kinds of processes in an early computational studies together we predicted that the way in which the carbon monoxide intermediate co star intermediate on one of these catalysts the way in which it was bound was it bound in one of these two possible configurations the atop or the bridge it looked like this could really influence the carbon-carbon coupling step so this key step along the way to the formation of a multicarbon to see two intermediate that's along the path to ethylene and then we asked you know how could we influence that how could we influence that physical chemistry and with our colleagues realized that if we could put a layer of organics of molecules right near the surface of the catalyst near where the CO was bound we could potentially tune the energetics so we could really tune the environment to the interaction between the CO ads or bait and the catalyst kind of from above now Teo and Jonas and their colleagues already had remarkable abilities to do this they had a family of N-aryl pyridinium salts where they could really systematically tune the beta charge on the nitrogen group and so we had a library at our disposal that we could go explore systematically and what we found was that the Faraday efficiency that electron efficiency that I described was very strongly tuned by the selection of these molecules and it was very strongly correlated with the experimentally measured ratio of the atop to the bound ads or baits on the catalyst and so it kind of fulfilled this idea that a sort of a tweaking of the environment in which the ads or bait saw itself from above was a feasible strategy and that the molecular approach one that of course is now widely adopted and studied in homogeneous catalysis could really be united with the heterogeneous approach used in electro catalysis many electro catalysis works so this we optimized together we're able to oligomerize the best of these molecules to make a more stable platform and we're able to achieve about a hundred ninety hours of continuous operation in this enhanced performance co2 death lean system so I spoken about what you do if you already have the co2 captured but another really interesting area that many groups are interested in is trying to figure out ways in which you could make the capture of the co2 either from the industrial flu or from the factory that's producing ethanol or perhaps eventually from direct air capture how you could make that proceed more efficiently both from a energy cost perspective and also from a capital cost perspective and the approach that we've taken is one known as reactive capture so first I need to just highlight the legacy approach that that many groups are pursuing and is now growing in its adoption and its performance which is first to take your gas be it air or be it flue gas contact it to either a liquid or a solid zorben but then use typically thermal swing or pressure swing or vacuum swing to release that co2 and feed that into one of these electrolyzers I've been speaking about but that comes with quite a few inefficiencies and so recently this community of folks in reactive capture started to ask could we use the liquid that captures the co2 and which of course now has a high concentration of the co2 within it could we use the liquid and feed that directly into our electrolyzer so it's called reactive capture because you're doing both steps here you're doing the capture and release step and you're actually taking the co2 down the line closer to the final product such as co or syngas and so recently we built systems to do this I'm a little sensitive to the fact that I'm a running a little on time so I might skip one or two maybe the engineering drawings we can skip and we implemented it first with amines liquid amines that are used already in co2 capture especially in the flue gas application and found that when we engineered the electrolyte we could produce carbon monoxide which can be then further reacted downstream we could produce that with a 72% faraday efficiency in this direct fashion electrically so this field of electrochemical co2 conversion also has many fascinating challenges this is great news for the undergraduate master's doctoral student in the postdoc out there these systems do require further advantages advances in their energy efficiency and their carbon efficiency they also require work on durability science and there's a very interesting regime in which we can explore how to make them produce more diverse chemicals the chemical sector is diverse there are many products we'd like to produce directly I'll use what remains of my two minutes to just skip through a few areas we're excited about for the future one is that both I and my students love a good translational problem everybody's looking to have real impact and so we've recently participated in an XPRIZ competition where we scaled from a gram per day to the 10 kilogram per day regime a co2 to ethylene system we we basically said all that's our all that's are available to us in trying to accelerate the discovery of new materials for energy harvesting like from the Sun and also for catalyst discovery and so we've increasingly started to turn not only to computational methods which we've been using for a while like DFT but in fact accelerating the use of those with the aid of machine learning so that can be done at the computational screen stage but it can also be done at the synthesis stage the high throughput characterization phase and then there's a lot of fascinating work to be done with actually handling this huge amount of data and learning off of it and eventually guiding the next set of experiments using AI and I'll close with just saying that while I've spoken much about sort of applied goals and targets and wide searches like in my last slide I continue to feel that understanding the scientific underpinnings of all of these processes is the other just as crucial third leg of the stool in order to advance the this field with the alacrity that it deserves so with that I'm really looking forward to some discussion and some Q&A and some panel but I would like to acknowledge the supporters of the work and I'd like to thank the group that did all the work and to thank you and thank you very much for the wonderful talk really I open you really exciting so any questions from the students audience faculty members I'll start off excellent talk so is there any scenario where perovskites don't reach 20 year lifetimes and still display silicon yeah so I think there's a scenario where oh sorry and still do display yeah that your number you hear a lot yeah I was gonna go down a niche route which is they could have impact without necessarily achieving 20 year lifetimes but actually you know I said 20 year lifetimes which is why you said 20 year lifetimes but those are now becoming 25 and 30 and there's a 40 year lifetimes right and certainly when I do the analysis and I always do the analysis more importantly you know when if the competition can amortize its costs over a 40 year lifetime you basically have to be able to amortize your costs over a 40 year lifetime so no I think it's I think it's the right goal for the big climate impact for maybe niche your stuff and for making you know these beautiful windows into solar harvesters or maybe a flexible thing on your car or on your smartphone I think there may be ways to word shorter-term impact but I think the big climate impact comes from a very robust solution yeah very nice talk I have a question about the more related to CO2 reduction reaction okay you mentioned machine learning yeah I do DFT calculations but how do you think like a more from mental research instead of the machine learning direction such as for CO2 reduction Mark Cooper recently last year they show it must be cat times if no cat times there then no CO2 reduction this machine learning this machine cannot let I know that there's no way to know this result from manual part this must be scientists know that how about this very very from mental research to really really know the studies of like a reaction mechanism and does the active sides this part yeah I'm with you 100% and of course like you I'm a huge fan of Mark Cooper and and I'll just add that Mark as you said has some great works on cation effects in the last even just year and we've also been you know I didn't didn't mention it I guess because I wasn't getting into all the details but in the work where I was describing the direct use of the CO2 captured is to the amines you know when we tried that first with our regular electrolyte we got absolutely nothing and then when we started to change the electrolyte and we started to think about think fundamentally like you're suggesting about the electrochemical double layer and the fact that in MEA solution the cations the large organic cations are actually blocking the double layer and they're getting in the way of our ability to reduce the carbomate only when we started to think that way and only when we started to add potassium cations at high concentrations where we able to interact with the carbomate bound to the amines so I agree with you you couldn't sort of do a wide DFT screen of a catalyst and reach these kinds of answers because that actually not where the problem was and so you know in our work it was done more experimentally we used electrochemical impedance spectroscopy to see where we were getting blocked and where we were failing to interact with the anions but I tend to agree with you that there's always going to be a place for deep science deep mechanistic investigation including posing curiosity driven questions early like that we don't questions we don't know already that we actually need to know the answer to that to solve some problem but we just need to build up the basic knowledge because all of that comes in handy at some point you just don't know when hi so I recently read a paper on single-item catalyst you doped different metals on the copper surface so in that paper there's experimental and theoretical proof that the iron doped copper gives higher selectivity to methane so did you start with the experiments or like did you start with DFT calculations for that research paper yeah you know in general actually we tend to be a little bit iterative on that and so we're fortunate that our research group there's people like some fu hung who was the first author that paper who have a remarkable chemical intuition about what might be effective and you know who was thinking about how the doping with the iron could potentially influence the energetics of the key intermediates on the surface and just the right way to produce more methane which is what some food was after in that work and so there's people like some food but then they do wander over to the cube of their colleague and they say you know could you screen a bunch of candidate dopants and they sort of try these out and they see you know whether it looks like they're having the modulating effect and and then somebody goes into the lab and builds some catalysts and sometimes the ability to go make them and characterize them for performance that's actually the fastest thing to do I mean that's faster than running off to the synchrotron and doing the XAS and figuring out how you're modulating the oxidation state of the copper near the iron or it's faster than actually proving that you did make something with single atoms that stay stable with single atoms so often that's a fairly early part of the process is is you know with some guidance from theory and from intuition finding a really interesting catalyst and then often the most time-consuming part comes after that which is what do you actually have how is it working how can you learn about mechanism and there's another iterative loop with theory there so I've deliberately not answered your question which came first the chicken or the egg because we do parallel interactive chicken synthesis and egg synthesis with like a lot of conversations between the two chefs this one more question that so for ethylene pilot plan to use copper aluminum bi-metallic for that plant yeah so in that system I know what you're asking about we actually ended up using a pure copper catalyst and the reason is that in that work the best electrolyte for us to operate in was well it was in an MEA system so we didn't have a you know very high alkalinity electrolyte because we started to move in the direction of trying to manage and mitigate and reduce the carbonate formation problem and so in that system which had you know not quite our very very top fair day efficiency you know so we've we've achieved 70 78 percent occasionally 80 percent fair to 80 percent fair day efficiencies in these systems in the extreme alkaline conditions but in that work where the goal was scale long lasting of liability we had to operate for a month in the field continuously we worked with a pure copper system okay thank you for a very nice thought because it's very optimistic and maybe we can solve the global warming problem but the question I like to know is that you have all this chemical synthesis thing I'm not a chemical engineer nor a chemist how does this system compare in terms of efficiency to nature like we have chlorophyll photosynthesis that converts carbon dioxide into wood or plants how do this natural process or many plus process compared to to nature well I was fortunate to have lunch with one of your colleagues today and I asked him for a reminder of the kind of overall solar energy to stored chemical energy efficiency going from solar to biomass and he said roughly a tenth of a percent and so I can now is that did I get that right a roughly a tenth of a percent and I think it varies depending on the crop we're talking about etc but let's call it that rough range and so a solar cell with a 30% energy efficiency followed by an electric catalysis system with a say a target 50% energy efficiency will have an overall efficiency that's the product of those two numbers so it will be 30 divided by 2 or will be 15% and so it will have you know I guess that's two orders of magnitude it will have two orders of magnitude higher energy efficiency and in my head actually one of the crucial things that that provides is actually just better land use and because I recognize that with biology you can kind of spread out and you can consume a lot of land but if you're looking at from the point of view of a land use efficiency and land is a finite resource and they're competing uses for it such as agriculture is a good example of one if you can achieve a two orders of magnitude higher overall energy efficiency in one of these technological strategies or sort of artificial strategies then you'll be able to leave more land available for some of the other uses that we have for it. Hi I had a question about the perovskite solar cells so it seems like the it looks like the best way to get high efficiencies further is through the tandem cells like perovskite, perovskite or perovskite silicon so if we take that as as the way from like if you want to do computational screening and you know high throughput you know DFT based approaches should we be modifying our search criterion in a certain way to suit like tandem solar cells better and like you know a related question would be which features or properties would you be looking for like you know defect tolerance stability band gap you know lattice matching I don't know what else are we missing there. Yeah it's a great question so and the answer actually depends on which of the layers in the tandem we're thinking of so the tandem of course consists of a high band gap and a low band gap material a high band gap materials today have been mostly made by mixing halides so combining abromide and iodide but those have been introducing the challenges of phase kind of re-segregation into a bromine rich phase and an iodide rich phase and so if you're in the computational business if you can help us crack this problem of either finding a high band gap absorber that doesn't rely on the mixed halides or a mixed halide absorber that remains stable over extended operating periods at higher temperatures that would be very impactful for the field. The low band gap material used in tandem today is a lead tin mix and the tin introduces sort of a novel problem I feel like which is that it's notoriously hard to control its oxidation state and so kind of analogous answer if you could help us discover a solution here that either didn't rely on tin or if you could help us discover strategies that truly stabilized in a long-term way the mixed lead tin materials and the oxidation states of those systems while keeping all the other good qualities that we like because the efficiencies the diffusion lengths are quite nice that would be a very valuable contribution for the field as well. Any questions from students or from online audience? So you mentioned machine learning and AI as being one of the tools for kind of technological development because this is kind of I guess a sort of new tool in this area how do you see the introduction of this into academia-based research or industry-based research and then how do you see kind of like the long-term synergy I suppose between machine learning and then experimental research? Yeah I love your I love your question so actually for us we'd sort of formulated the desire to see if we could accelerate some of our discovery with machine learning we hadn't made too much progress and one of the issues this is I'm in sort of 2020 here early and one of the issues is that a lot of people in my group while very talented and brilliant didn't actually have the training you know in all of the background areas that would enable them to do even kind of basic available machine learning and then the pandemic hit and people started offering boot camps to one another so there were a few people who had a CS and a CE background in the group they started offering boot camps to one another where they take a week all online and kind of just do introduction to machine learning with a very sort of pragmatic bent that allowed us to sort of get going on our research in the area and actually the boot camps themselves ended up being kind of recyclable because we eventually ended up developing an industry consortium that was sponsoring research in using machine learning to accelerate this catalyst discovery but it turns out our industry partners had an appetite for the boot camps as well so they could start to engage with the data and roll up their sleeves so I will say this was kind of a reminder for me I guess one shouldn't either reminder but it's sort of reminder that research and graduate education are like intermittently intermingled with one another and then nobody actually does pure research and one of the reasons why why universities have such a competitive advantage in doing great research is the fact that we're bringing in these doctoral students who are at the stage where they're still learning and they can absorb new things and they can develop new techniques and they can combine them with other techniques the way folks who are a little further along into their careers might find hard to do so to me it was a lesson in the in the agility of the graduate student in picking up things like machine learning and running with them thank you one last question yeah thank you for the great talk I have a more broader question so you have two main thrusts in your research one is kind of more like preparing for the future creating less carbon emissions another is more about cleaning up the mess that we've made capturing carbons and like maintaining the current structure that we have but always as a researcher it's always a question whether these new emerging technologies can really penetrate into the industry and really industries are really capital efficient they're very they're very rigid in a way that they want to maintain their current systems so comparing these two technologies from your perspective as you've done research in both directions what do you think is in terms of time frame or capital efficiency which technology do you think among the two are more closer to industry adoption and what do you think are the main challenges that we would the one of main challenges that we need to overcome for either to penetrate into the industry and more successfully yeah thanks for the great question they're two pretty distinct ones and that the solar is already an established and rapidly growing industry it's a large industry and if we can have the kind of impact that we aspire to have I think we could potentially make that impact within the five to ten year timeframe in that area the CC us is less advanced the solutions are less advanced and of course while there are there's a hydrocarbon industry that's huge the industry for utilizing co2 and transforming it is less mature and there are actually some niche applications for it but certainly not turning it into today and turning into fuels and chemicals and feedstocks so I think that one's further out maybe that's one of the 10 or 15 or 20 year time scale and yeah I think I think the critical issues in in the areas my ranking always does start with further progress and performance until we have the performance is required to meet requirements it's always starts with performance but then I would say the durability science and then the development of the mechanistic picture that we've discussed a few times now are very high priorities for both fields as well yeah thank you very much for the questions and answers so let's end the session now and thanks to a speaker distinguished speaker again