 So please share your screen and introduce yourself. Thank you very much. Thanks a lot for the invitation. I'm extremely happy to be here to talk about physics in the context of African research and also the needs that are connected to energy conversion and storage. So can you see my screen at this point? Can you see it or not? Yeah, we see it. So please make it a full and then we can go ahead. Excellent. So my name is Ismail Adabo and I'm an associate professor at Penn State University. And what I'm going to talk about today is how we can use data-driven, that is computational material discovery, in order to address some of the questions associated with the production of energy. And I'm going to talk specifically today about the production of hydrogen as a fuel from solar energy. And before I jump into the bulk of this research, I would very much like to thank the organizers of this African Physical Society conference. And I very much hope that this conference will continue to grow and attract young researcher and experienced researcher, and that this will ultimately become an important area where ideas can be shared in the area of physics and other connected areas such as material science. So before I go and describe more about this computational work, I'd like to step back a moment to just tell you my background. So my family is from Guinea. And I did my undergraduate studies in France at the College of Polytechnic. Then I moved to the US in Massachusetts to do my PhD, where I saw an opportunity to apply computational material research in order to solve important challenges related to energy conversion and storage. And I think this has a lot of potential, especially for countries around the world and in particular in Africa. What I would like to do today is to tell you about my work at Penn State University. That is a state you may have heard recently because of the US election. It was a swing state, and there has been a lot of discussion in the news. And so I work at Penn State, where I have a group that dedicate its activities to the use of computer simulations in order to accelerate the discovery of new materials. So the context of this is from a broad picture point of view is the issue of providing consumption, transportation fuel for a growing world population. And I've here a graph that shows the global consumption of fuel as a function of time. And you can see that there are essentially two main contributors. There's a great contribution that comes from light duty vehicles that, as you can see, is progressively being stabilized. And this is thanks to progress in efficiency and also the appearance recently, the emergence of hybrid vehicle and plug-in electric vehicles. And thanks to that, the consumption of fuel for the light duty fleet is actually quite stable. But you can see also that there is a blue contribution here that keeps increasing over time. And this is the contribution from heavy-duty transportation vehicles, such as those heavy-duty trucks, for which we don't have currently hybrid and electric alternative. And because of that, the consumption keeps increasing. And this means that for the foreseeable future, we will need to continue to produce a fuel for this heavy-duty fleet. So one option to alleviate this challenge is to use fuel cells. You might have heard recently about the Toyota Mirai. So Mirai in Japanese means future. And so this is what Toyota thinks could be a significant contributor to the future of transportation. It's a fuel cell vehicle that uses hydrogen as a fuel instead of conventional fossil fuels. And the benefit of hydrogen are multi-fold. First of all, it's extremely friendly environmentally. So the only byproduct of the use of hydrogen is water vapor. So it means that when you drive this car, you don't have any other emission than water. And also, it is extremely efficient in terms of energy efficiency. It is about twice as efficient as a conventional internal combustion system. But the bad, maybe the less advantageous aspect of using hydrogen is that hydrogen is very expensive. So I've here the typical price of one gallon of gasoline in the US, about $2.5 per gallon. And if you have to translate that in terms of hydrogen, the hydrogen is actually six to seven times more expensive than gasoline. And so this makes it a very economically problematic option for transportation. Another important aspect is that conventional hydrogen is often derived from methane reforming, which produces carbon dioxide. And carbon dioxide has all of the negative environmental impact, climatic impact that we know. And so as a result, it's important to develop alternative ways to produce hydrogen that will reduce this cost and avoid the production of CO2. One option to do so is photocatality, whether splitting or artificial photosynthesis. You have here two concepts to be able to produce hydrogen based on artificial photosynthesis. The first one is a single-bed particle suspension. So I save you the details, but you have essentially two plastic bags. And you see the dimension of this plastic bag here that contains a suspension of particles that can catalyze on one side the catalytic reduction of water into hydrogen and the catalytic oxidation of water into oxygen. And by this process, you will be able to produce H2. There is another concept in which you separate the two reactions. But the concept is essentially the same, that you will absorb sunlight, put water in those bags with a suspension of particles, and at the end of the day, you will produce hydrogen. So in that way, you can use water and sunlight to power your fuel cell car. So you don't need to extract resources from the ground. You can transform water, split it, and obtain hydrogen. So there has been a very comprehensive and very detailed technical economic analysis published in Energy and Environmental Science, which is one of the leading journals in the field of energy these days. And in this journal, they made an estimation that if we were to use the first step of technology, and if we were to achieve particles and materials that can achieve STH efficiency of 10%, so STS means solar to hydrogen conversion efficiency of 10%, we'll be able to reduce the price of production of hydrogen to $2 to $4 per gallon gasoline equivalent. This means that at that point, we will be competitive with petroleum-based fuel, and we will be able to power the transportation fleet systemically. And we can alleviate the need to extract more resources from the ground to produce fuel. So what we wanted to do is to address this question, can we develop materials that can reach an efficiency of 10%? And for that, we will rely on high-stru put computational material discovery. So this is an important evolution in the field of material science, whereby we have now computational methods that are extremely efficient and also predictive that enable you to predict the properties of materials before they are synthesized experimentally. And those are based on solving the equations of quantum mechanics, the Schrodinger equation. But the bottom line is that there has been a lot of activities in different fields, such as battery research, development of solar cells, development of thermoelectric devices, photocatalytic materials like what I'm going to talk about today, piezoelectric, dielectric. And so using computers, we can rapidly screen tens of thousands of materials very quickly and be able to determine whether or not there will be good candidates for a specific application. But as you can see from this graph, there has been a lot of activity, but few of those computational high-stru put searches have been associated with experimental validation. So as you can see, only about 20% in all range of those computational studies have been associated with experimental validations. So we wanted to use those type of techniques but to address this problem of how you can validate experimentally your prediction and how you can achieve this cross-communication between theory and experiments. So this is the protocol that we have developed in collaboration with people at Cornell University, the group of Tito Abruña, Craig Finney, and at Penn State, the group of Venkat Gopalan and Reshak, as well as the National Renewable Energy Lab, where you can see that essentially we use computers to first predict the chemical stability of the compounds. Then we look at the solar absorption. So we can also predict that computationally, at which point we are interested in knowing if the band alignment is correct, meaning that the exceeding energy of an electron from the semiconductor will be higher than the entering energy of an electron in water and the same for the hole. You want the exceeding energy of a hole from your semiconductor to be lower than the entering energy of a hole in the water solvent in order to produce oxygen. So then once we have done that personally, we can look at the toxicity of the material. We want to make sure that this is environmentally and from a health point of view safe. Then we want to make sure that this is earth abundant, that we're not going to look at very expensive materials, such as gold or platinum, because that will be not economically viable, at which point we are going to check the synthesizability of the material, whether it can be made. And for that, we go into literature. And for each candidate, we look if there exists a reference that explains how to synthesize the material. If it is the case, we go to the next step, where we apply refined computational methods that are called DFT plus U. That is completely ab initio and does not require experimental data that enables you to predict better bandgap. So these are the beginning where course calculations, very quick calculation. And here we do the very refined calculation, at which point we can check the phase purity of the material. And then the last step is to check for the photocardatic activity of the material. And based on what we have learned from this crossword validation between theory and experiment, we go back to the beginning and we can again screen more material. So it's a secret approach that enables you, ultimately, to narrow down the choice of interesting materials. So I'm going to just summarize, in one picture, how this goes from a quantitative point of view. So we started from 70,000 compounds that are listed in the materials project and the inorganic crystal structure database. And we first calculate the enthalpy of formation in the computer. And you check that this enthalpy is exothermic, meaning that it is energetically favorable to make those compounds and that will be stable. Then we use our quick estimate of the bandgap epsilon g to make sure that we are within a range that will be reasonable. Then what we do, we check that the valence band and the conduction band of the semiconductor are suitable for water splitting. Again, using a quick estimate, the calculation that do not require the most extensive iterations. Then we want to make sure that the lethal dose of the material, the LD50 coefficient, is greater than 250 milligrams per kilogram. So essentially, it is less toxic than lead. And then we want to make sure that it's earth abundant. It can be synthesized. And here we apply the refined DFT plus your calculations that enable us to have an accurate prediction of the bed edges. And at this point, we check the purity of the sample and we can test the 14 materials that we have. So essentially, 70,000. And we narrow down the choice to 14 interesting materials as the first cycle of this iteration. And you can repeat this iteration based on what you have learned at each step of the screening. So now, quickly the results. So these are the materials that were made by our colleagues. And these are bandgap measurements obtained by TORC analysis. And you see that for most of the compound, we have a pretty clear signature of the bandgap, except for some of them where we can have mixed oxidation states and mid-gap states that may affect the determination of the bandgap. But essentially, here we are fairly confident about the bandgap that we can measure, except maybe for those few materials. We then compare those to experimental bandgap computational prediction. So you see there are essentially two families of materials, the blue ones that are occlusion materials that are non-magnetic for which we tend to do quite well in predicting the bandgap within 0.5 or 1 electron volt. But you can see for magnetic material, we don't do as well. And this is something that we have learned as part of this cross validation that probably non-magnetic material are under control, but we need to do more work in order to describe the magnetic structure of other materials. Here are the most shocking measurement that we did at the National Renewable Energy Lab. They have unique facilities to be able to determine experimentally the bandgap. So you see in orange what they obtained. And in blue are simulations. So you can see that we are fairly good agreement between the two sets, except maybe for calcium in-date and this magnesium entimonate. And this might be due to the appearance of surface state. So I will not go too much into that. But essentially, we are quite happy about the result that we obtained. And we feel we understand some of the limitation of the method again. So again, very important cross validation between experiment theory. So then our colleagues at Penn State developed a very sensitive gas chromatography setup to detect hydrogen. Essentially, you want to be able to purge your chamber between each test. And you want to be very sealed. So they develop a very advanced gas chromatography analysis capability for that purpose. And this project would not have been possible without their contribution and this important setup. So these are the results. So among the nine materials that are described, the non-magnetic materials, we indeed found hydrogen production at the end for many of those. So you can see that for seven of the nine materials, we have hydrogen production. And for the other material, we don't see hydrogen production. And this is expected because some of them, you see there is a shift in the band edges that I've shown previously that is potentially due to surface corrosion. And this might be one of the explanations why we don't see hydrogen. But we were very excited actually by those results that were very much the result of a collaborative research between different universities and national renewable energy lab. And we feel that this could be an important contribution, a meaningful contribution to the search on new photo catalysts. And so we're doing more work in this area. And we have submitted this paper in Energy Environmental Science. It's under review. And we hope that it will appear soon there. You have one minute. Yeah. So with that, I would like to thank you. So thank you very much for attending. So I hope I give you a sense of how we can use computer simulations that are accessible, in particular to the ICTP Institute, that really enable you to make prediction and to accelerate the discovery of new materials. And I feel that this is an important frontier for the future of material science. And I feel that it can be an important opportunity for institutions across the continent. So with that, I would like to thank the people without whom this work would not have been possible, the students. So you see, it's a very collaborative work involving Penn State, EPFL, Cornell University, QC University in Japan. And we acknowledge funding from the DOE from the NSF. And these are the brilliant students who did all the work. So thank you very much for your attention. I'd be happy to answer your questions. Thank you for the interesting presentation. Ali, you can go ahead. Thank you very much, Ismaili. It's good to see you. So I had a couple of questions. The first one is, have you thought, I'm sure you have, about this connection between this data-driven approaches? Can this be reversed engineered to create like a universal DFT functional that will fix and be the holy grail of everything? So that's the first one. The second one was related to some of the band gaps that you showed comparing theory and experiment. One thing that's consistently different between the theory and experiment is that you have this red edge, this tail to lower energies in the experiment. That's missing it. So is this because of impurities? Is it because of phonons? Where does this come from? Yeah, very good question. So to answer your question, because of the time constraint, I didn't go too much into the details of the theory. But I do have a slide here that essentially compares some of the calculation that you'll find in the literature and the calculation that we do here. So we apply computational techniques that try to better describe the interaction between the electrons in order to understand their excitation where they are under illumination. So you can see here the typical DFT calculation will be in blue. And our calculation are in gray. So you see that we tend to do better, not always, but we tend to do better. Sometimes the matter is predicted to be metallic. We do predict them to be semiconductor. And this is an important validation of the technique. And it motivates us to further look into that those methods to develop better functionals and gaps. Regarding the experiment versus theory, so indeed, those states are due to mid-gap states impurities in the materials, as you can see here. And this is something that we discussed more specifically in this paper. So we go into the question of why those materials develop mid-gap states. And definitely, this is the frontier of what we can do here. It would be extremely valuable to be able to anticipate the production, the formation of mid-gap states. And that's something that I would be, yes. It's not because of distortions of the lattice or anything, not because of phonons. Like how does thermal vibrations can also, of course, create some broadening in spectra, so. Yeah, absolutely. This can be predicted as well. All the vibronic effects on the spectra is something that can be predicted. We are interested in electron phonon coupling, especially in the context of thermal electric. So this is another topic I didn't talk about today. But yet nowadays, there could be fairly good and reliable prediction of electron phonon coupling and their effect on spectroscopic properties. OK, OK, OK. It's a little frontier. OK, got it. Thanks. Thank you very much. Excellent. OK, great. So I'm encouraging the students to ask. I mean, it's a platform where you need to ask, even if you think it's a stupid question. It's not a stupid thing at the end of the day. So go ahead and ask. And I would like to mention, if there are students, we are very much interested in having some of the top students applying to Penn State. So I'll be happy to discuss offline as well. OK, great. OK, great. So if we don't have any more questions, then we can move on to the. Hold on. The time is already up. I think it's OK, OK, OK. Fine, fine, fine. Thank you. So thank you so much. You can discuss offline. So we are just going to move to the last talk, which we.