Resources

• Argonne National Laboratory
• Department of Energy (DOE) Battery Materials Research
• Joint Center for Energy Storage Research (JCESR)
• International Energy Agency: Grid-Scale Battery Storage

Host and Guest

Climate Rising Host: Professor Mike Toffel, Faculty Chair, Business & Environment Initiative (LinkedIn)

Guest: Logan Ward, Deputy Scientist, Argonne National Laboratory (LinkedIn)

Transcript

Editor's Note: The following was prepared by a machine algorithm, and may not perfectly reflect the audio file of the interview.

Mike Toffel:

Logan, thank you so much for joining us here on Climate Rising.

Logan Ward:

I'm really glad too.

Mike Toffel:

So Logan, why don't we start with a brief introduction? Tell us a little bit about how you ended up at Argonne National Lab.

Logan Ward:

Yeah. I've worked at Argonne for about eight years now. Before that, I was a PhD student at Northwestern studying a mix of material science and machine learning. My PhD is in material science and when turning the road on the end of my graduation, I had the options of working at many places, industry, national laboratories, and I was really motivated to work on big societal problems with large teams. So the opportunity to go do that with the Department of Energy was really too good to pass up.

Mike Toffel:

Great. Now I know you've worked on battery technology, and we'll talk mostly about that today, but you've worked on a range of topics in the eight years at Argonne. Can you just give a brief thumbnail sketch as to what those projects have been?

Logan Ward:

Sure. So all of them, to start out generally, because my background is in materials, I studied some computer science along with that. And Argonne's, the division I work in, is focused on running the world's largest computers. have some mix of those three. Sometimes the materials are for nuclear reactors. Some are for batteries. Materials for carbon capture and conversion into useful products, airline fuels, lots of things all with that general through thread of energy plus AI plus big computers.

Mike Toffel:

And okay, so let's dive into the battery work. So now there's lots of battery technology development going on these days, anywhere from battery electric vehicle technologies to grid scale batteries. And you've been working on the grid scale battery solutions, is that correct? And so let's talk a little bit about the need for grid scale battery and what the conventional technologies are. And what the problems are with those technologies that motivate the research that you're doing and then we can get into what the research actually is.

Logan Ward:

Yeah, that's a really broad range of technologies that go into storing the grid. And we'll start out by at least detailing a few of the reasons why one might do that. Reasons to store energy over this course of let's say seconds to minutes, maybe an hour to ensure that the grid runs at the right frequency all the time as different power plants turn on and off or loads change. The reason to save it for a few hours because people use more energy at the daytime. The sun is shining, and renewables tend to work more in the day time and you need to hold on to that energy till the night. There are reasons to store it for months due to seasonal variation or disaster recovery. And each of those, because they have such different requirements, have different technologies that suit them best. I was at a presentation yesterday about there's a technology of making ice and heating oil extremely hot and storing energy that way. Or as what I've worked on more personally, storing it in big tanks with electrically charged molecules, which are flow batteries.

Mike Toffel:

Yeah, so a lot of people think about batteries only as the electricity storage, like the batteries we're most familiar with. But in grid scale, as you're noting, there's thermal. That's the example you gave. You could store it in heat, or you could store it as cold. There's mechanical, where you can store it with pumping water up a hill and then having it turn turbines to come down. So there's actually quite a range at grid scales, I understand. Is that right?

Logan Ward:

That's right.

Mike Toffel:

So you're working on the batteries that we're more familiar with, as you mentioned,storing electrons in liquids in this context. And what are the properties, what is the chemistry there? What are the materials that are the conventional approach for doing that? And then what are the challenges of that that motivates your research?

Logan Ward:

So the battery that I work on most is called a redox flow battery. The way they work is at least with two types of molecules. One, which is really the workhorse for the battery, is the one that actually stores the electrons. What you do is take that battery molecule from a normal state and then you electrically charge it and then move it elsewhere for long-term storage.

Now that battery or that molecule in order to move it has a solvent that goes along with it. That can often in many cases be water. It can be some other types of liquids that are a bit more resistant to large electrical charges. But the same basic principle works. You've got a molecule that holds on to that energy in the form of having one more electron than it should. And then it has to be shuttled off somewhere.

Mike Toffel:

Got it. And I imagine, I'm just speculating here, that some of the properties you look for in these chemistry are cost, availability, environmental impact. You hear about labor impact in extracting some materials, especially on the EV side. Maybe politics, you know, which countries you can source the information from. We hear a lot about battery electric vehicle batteries. Are those also the issues at play at grid storage at grid scale storage?

Logan Ward:

Yeah, so especially across, yeah, those economic factors become huge when we're talking about installations that store days’ worth of electricity for a city.

And when it comes to sort of the specific technology, so cost is of course the major thing running over any kind of design that has to work within the economy. But also there are a lot of things that need to work just for the technology to function. So what you really need to do to make the flow battery work is that molecule has to be able to hold a lot of electrical energy and do so without changing to something else while it's in that active state because that molecule wants to go back to the way it was and can do that by breaking into pieces by reacting with the solvent that it's carrying reacting with the walls of the vessel. So we try to find these really special molecules that not can just hold energy and a lot of it but hold on to that for the days and hours we need them to and then be able to do that again and again for years so we can keep using that battery without having to continually pay the cost of making a new one.

Mike Toffel:

Got it. So, and as I understand, why is this issue more important now than ever? And I think it's because as we try and scale renewables, in particular, for example, wind and solar, which are intermittent, sometimes they're there, sometimes they're not, in order for us to continue investing in them, we need to make sure that we can leverage their energy generation, even when it exceeds the demand in a given moment in time that currently sometimes you see folks turning off wind turbines. Though there's wind, they've pivoted the blade so they're not generating power. Or you hear sometimes of a negative price, where in order to dissuade the production from some sources, you actually make them pay to offload it into the grid. These are problems of mismatch in supply and demand, where in those moments the supply exceeds demand.

So even for today's deployment, sometimes there's gotta be a better way than turning off blades that can generate power for virtually nothing. And batteries are one solution there, but it seems like in order to increase the share of renewables in grid, we have to figure out ways to allow them to generate a profit even when their production alongside everyone else's production exceeds demand. Is that the right way to think about it?

Logan Ward:

Absolutely, and that's really the core for needing to store that energy. It doesn't arrive when we need it to, and so we have to hold on to it for in some cases a month. And that's really the long duration type of batteries.

Mike Toffel:

Got it. And so you mentioned cost or cost per kilowatt hour is the key differentiator that you're looking to reduce through your research. Is that right?

Logan Ward:

Mm-hmm. That's right. And you can do that in many ways.

Mike Toffel:

So when we think about the cost of a battery, one holistic way to think about that is, well, how much are the materials one has to source? Then what's the cost of manufacturing it into a battery, deploying the battery, maintaining the battery? All of those go into...

Mike Toffel:

the ongoing, the capital cost and the operating cost. But as I understand it, you're really working on the ingredients portion. So how do we figure out a cheaper set of ingredients that have the durability and all the other properties you mentioned?

Logan Ward:

Exactly. We're seeking to do two things. One, find a material that is cheap to begin with. Or, if it isn't cheap to begin with, make that cost as effective as possible. To be able to store the most amount of energy for the longest without replacement for the amount you push in.

Mike Toffel:

Got it. And we're going to talk about AI in a minute, which is the approach you're using. But what's the conventional approach to finding new battery chemistry? What's it looked like over the past decades before we had available these AI tools that you're using?

Logan Ward:

Yeah. So the way that one will search out has two components that AI now serves are a helping role in the where to look and the how. So the where to look is often human intuition.

I know from having a background in chemistry that there are certain types of molecules that won't be as reactive as others. So I can think from reading all the science papers that maybe there's a unique molecule combination. And then I can go off and test that concept. It will be different kinds of computations, which thanks to more clever theory and faster computers have been getting more intricate over time. But that basic recipe of coming up with new ideas by really sitting and thinking about the problem and then designing a way of going and testing it as cheaply as possible is a way that's going about. And actually that basic recipe hasn't really changed with the introduction of AI.

Mike Toffel:

So let's talk about these two in turn, the selection and then the how. So the selection process used to rely on intuition, as you're noting, by experienced scientists who understand the properties of various chemistries. So they're taking informed bets on which chemicals or which substances to actually evaluate. And what's the AI enhanced approach?

Logan Ward:

Yeah, the AI enhanced approach is to provide an alternative way for a computer to mimic that process of human creativity.

There are tools like GPT or the variety of models that have seen a bunch of ways that words get strung together such that when you provide a new query, how would you react to this new bit of information? It can go back through those patterns and be able to generate something that matches the situation.

We do very similar things with AI where we provide examples of here are the ways that molecules look. Here are the ways that a good molecule looks that I know works for this problem. Can you come up with a few examples of other molecules that resemble things that you've seen before? And that kind of generative model approach really provides not a precisely better route for coming up with new materials, but one that really augments the way a human looks at the problem. So it's basically like getting another set of eyes. A colleague of mine likes the concept of having an alien intelligence that thinks very different than you, but it's all that general idea of just getting another mind to contribute to the thinking process.

Mike Toffel:

That's really interesting, the comparison to the large language models that some of us are playing with chat GPT and others. How do you, whereas those models that we're more familiar with in the lay language will use information from news reports from the web in order to figure out which words tend to get strung together. In your case about the chemistry, are you training the model using peer-reviewed science or lab experiments from the Department of Energy? How do you train these models in order to figure out those patterns? Because I imagine it's not by just looking at the general corpus of text out there.

Logan Ward:

Yeah, well, you actually could be right about it going on the general text. That's something that people are trying to do, have it read papers and use just them, not just the molecules in them, but all of the context of the ways they're described to help the generation process. But that's the really new way to do it. Thanks to better technologies on being able to read text.

The more old way of doing it, I humorously put the word old there because the first paper I remember reading around this was only five years ago, which is a long time in AI world, but I'm beginning to realize is maybe not quick for many other technologies. But that aside, the way it works is very similar to the text. Instead of giving examples of,

conversations on Reddit or newspaper articles. We give it examples of here are molecules that we've used in the past and molecules kind of have a Lego block. Maybe to keep the analogy closer to text, their letters and their words and molecules, and it figures out how they often get stitched together and with that information can generate things that look like new molecules. There's a lot of subtleties in the way that exactly happens, and scientists have been thinking hard about it. There are dozens of ways of doing that, but all of them rhyme with that general strategy of clinking Lego blocks together in ways that look like things it's seen before.

Mike Toffel:

I see. So beyond exposing your models to peer-reviewed science, are you typing in these equations and talking about, you know, here's the inputs yield these outputs or where else are you actually training them on?

Logan Ward:

So there are actually two fun ways of using the AI. There's one that's a generative model, which is working just, I give it examples of things. I give it a prompt and it produces something that looks like it. So that, in nowhere in that description, did I explicitly say a certain molecule behaves a certain way. It's all just sort of absorbed into the general way that model functions.

There's a second strategy where you give examples of molecules equals this, and you can train a model that will be very good at learning how a molecule, how its words and letters, stitched together implies how it will behave. And once you have that, you can do a lot of different things. You could say run through all examples of known molecules and pick which one looks like it's the best based on those patterns. Or you could take that model, understand the regions it doesn't work very well in, and use that to help you search through. So you're more likely to find something that is not just an incremental improvement, but is something that's more unexpected. And there are reasons for searching both, those ways are indeed the two different approaches that people use it in science.

Mike Toffel:

Great. So that's about the what molecules to search for based on your hope to figure out what their reaction would yield and whether those are likely to be successful in a battery given the characteristics that you're seeking. Let's talk about the how because that was the second piece you said. You said there was AIs being applied to revolutionize the what to search for.

Logan Ward:

Mm-hmm.

Mike Toffel:

And then what you do once you figure out what you want to run your experiments on. We'll talk about the portion. Once again, what's the conventional approach and what's the AI enhanced approach?

Logan Ward:

So the way we find a material is we enumerate what are the things it needs to do and how can we measure how well it does them. So let's say I know my battery needs to store a lot of energy per molecule. Now I can equate that to how much energy does a single molecule stores when I give it a charge.

That's something that you can measure. You can make the material in a lab. You can test its voltage and that will be able to feed in to your decision. Should I run other tests on it? And then you'll start running other tests. Does it react with everything after it's charged? And you can do that and you can build that up. The thing and just the nature of the process is each one of those steps. It takes a lot of time. Making something in a lab isn't always assured you can do it. Even if you can, those tests cost money, they cost time, and that limits how many materials you can actually test.

Mike Toffel:

And the AI enhanced approach.

Logan Ward:

Sure. The AI approach is to take those same processes, the things we can measure, and make them faster. When we earlier talking about how there are ways of, I can take a pattern of a molecule and I can tell how well it does, I can use that to emulate each of those experiments. If I run that experiment enough times, or I can go to the science literature and I can find enough examples of other people who have. I can train a model that will repeat that process, but much quicker and give me a guess about how well it will work. And that can tell me, should I test this material? Should I test something else? And that's really the kind of the number one, the first way that we started using these kind of AI models to be able to do the same process, but do it by guessing rather than actually running the experiment.

Mike Toffel:

Got it. So that basically another in a, if you think about it, innovation funnel, which is how we teach innovation here at HBS to our first year students, where you put lots of things as candidates and then you're trying to narrow down which candidates to continue investigating, because each level of investigation is costly. And then you have these gates where you decide which one should proceed to the next stage. And then we'll invest some more in testing out those ideas.

So in this case, it sounds like in the testing process, not only in the decision about what to investigate, AI is enhancing, well, instead of throwing these hundred things in, let's throw these other hundred things in, or maybe there's some overlap, changing the basis of that intuition. That was on the selection side. And now on the narrowing down the funnel, it's another stage that you've layered in here that says if we have 100 things that we're going to test in the past sort of manually, let's use AI to figure out how to narrow that down to the most likely candidates who we should then test. So instead of testing all 100 in the lab, we might test some fraction of that.

Logan Ward:

Exactly.

Mike Toffel:

Yeah, right. So what's the magnitude of improvement here? So if, I've just used my number just to fix ideas here. So if we had 100 candidates that we might have run through a physical process in the lab each time figuring out, does it hold the charge and then does it stabilize and is it reactive? But you're at least starting that process with 100 in a physical way.

If you inject this AI enhanced process in the middle there, how many would that narrow you down to? How many would you then cull and not waste your time, hopefully waste your time, using them in the lab?

Logan Ward:

Great question. For some of the projects that I've worked on, we've evaluated a million, 10 million candidates with AI and tested 100. So that's what thousands of times the difference in the number of resources it takes. Testing something with an AI model takes fractions of a second, so we can employ them as vigorously as we want.

Mike Toffel:

And then how do you measure the accuracy of this culling process? And here I would guess, you know, the typical way I think a lay person like me would imagine how to think about that question even is type one and type two errors. So type two errors being things we ruled out that we wish we hadn't because actually it would have been a successful candidate. And type one errors would be things we ruled in but turns out to not have the properties that we hoped it would have. All of this not in the abstract but relative to the conventional approach. So how do you think about, speed is obviously a great thing but not if it makes the wrong decisions.

Logan Ward:

Yeah, that's a question that we wrestle with continually during the process. As you're building the AI model, before you're even considering using it in practice, you employ a lot of tests where, say, I withhold an amount of the data. I train the model on everything that.

I didn't withhold and then I measure those kinds of type one and type two errors on if I use this model, how likely it would have showed me down the wrong path and being able to look and we'll do that very extensively can gradually make the problem of how much you don't tell the model harder to really see when it will and when it won't work.

And let's say it gets to a point where everything works fine. That's still not the end of that sort of testing. That's not the end of the day. As you start testing those new materials, you're monitoring how much it was off? Do I really trust this model? And that's something you'll design into how you use it.

You don't always test the predictions that it tells you are good. Sometimes you look at the ones that the machine learning model is says it's unreliable. The predictions are unreliable and test those just to really make sure if everything is working the way that you hoped it would.

Mike Toffel:

So let's step up to a level now that we understand how AI is enhancing the discovery or the decision about which substances to test and then dramatically expediting the testing process in a way that one hopes is managing type one and type two errors as you're noting sort of a continuous challenge. This is being done at the US Department of Energy lab. So does that mean that all the work that you're doing is, in a sense, will be freely available to companies and organizations around the world who want access to it compared to a private sector R &D lab?

Logan Ward:

Yeah, Argonne has a very open stance towards sharing their research as most of the Department of Energy labs do. Most of the software that I produce for doing these kind of AI studies is open source. So you can go download it. I would be thrilled if people sent me complaints about whether it was working or not, because part of our role as laboratory is to make sure that research stays out there.

Now the materials we find will vary depending on who's funding the research, and if the actual material we find will be made available. Some of the materials will be licensed for companies to be able to work with Argonne to be able to scale up the manufacturing process. Others are research funded by a company. And that material we find stays within that company for their decision on how to operate or how to use that knowledge. So it really runs that spectrum. But in general, our role is to take these kinds of technologies and materials and make them available to industry.

Mike Toffel:

I mean, that brings up a related question as to why this work is being done at a national lab, largely funded by taxpayers, although as I understand what you just said, there's some funded by the private sector as well. So why does this even exist as an area as opposed to leaving it to the private sector to do their own R &D or maybe universities, labs to on the cutting frontier when there's 20-year uncertain paybacks, we often see that's an area of sort of fundamental science that some of the academic institutions are leaning into. So what's the justification for the taxpayers when you go out and say, this is what we're doing, or you or your bosses, bosses, bosses, going out and saying, you know, we need millions or even billions of dollars to do this work? What's the response to folks who think this should be a private sector effort?

Logan Ward:

Sure. So the notion between being very high risk and for an academic organization to really take on the risks of studying and it being a surefire idea, almost everything is ready and it's time to bring it to the market, is really just two ends of a spectrum. And the national laboratory system fits in the middle of ideas that have been refined enough by academia to be able to know they've got some merit, but still need some further exploration in order for them to really be immediately useful. And that's the role that the Department of Energy Sciences plays, being that transition between something that's a cool idea that could be worked out to something that is ready for being available to society and to industry at large.

Mike Toffel:

Got it. So this is the area where academics have the incentive to be on the cutting edge. It's okay if things don't work as expected. But then there's this intermediate spot where you're saying there's promise, but there's still so much uncertainty that it's unlikely to attract private sector investment. And so in order for it to de-risk it further, that's where the DOE steps in.

Logan Ward:

Absolutely, there are parts of laboratory that specialize on technology transfers of taking the ideas that have been refined and vetted enough and help have them be used by an existing company. If I recall correctly, a lot of the lithium-ion battery technologies used in electric vehicles are licensed out of Argonne. It was a material that was further refined at the laboratory, and then at that point became something that spurred a lot of technological developments after there.

Mike Toffel:

Interesting. I didn't know that history of lithium-ion batteries. Now when you license it to organizations, do you tend to, and I know this is not your area, so if it's outside we can move on, but does DOE tend to license it to a single organization or do they refuse to do single and say, since we're publicly funded, we want to license it to multiple players to sort of foster competition?

Logan Ward:

Great question. I think there are examples where you'll see both, but I don't think there's an explicit mandate that things need to be marketed elsewhere. And sometimes the form of how the technology is rolled out to the public is actually a new company forming under, with the assistance of the laboratory. And there are plenty of small business grants at the Department of Energy grants to, well, small companies and to really help those new ideas make their way up the ladder to something that can be really transformative.

Mike Toffel:

So in the Argonne lab and in your own lab within Argonne, how do you decide what to work on? You must receive a bunch of pitches from companies, from NSF grants, from universities to seek your attention and time. How do you sort through those and decide which ones to pursue?

Logan Ward:

That's a great question. And it takes many forms. A lot of it comes directly from the Department of Energy. The Department of Energy will, in consultation with other areas of the government, find those technological areas that really need a lot of attention now. And they'll solicit requests for proposals from the laboratories, saying that we need some AI technologies that are really going to help the development of grid-scale batteries.

For example, and my role at the laboratory includes coming up with those new ideas that help the DOE find the research that needs to be done. So it's really somewhat of a collaborative process. The government finds these are areas that need advancement. They know the general shape of the kind of technologies like flow batteries must be cheaper could be the overall goal. And they'll work with us scientists, we'll send them proposals, and that will help shape the direction of where the work goes.

And that process is very integral to being a scientist. I play both the role of providing the ideas to potentially be funded for those proposals in areas I didn't submit. I might also be the reviewer to help the Department of Energy find what is the research that's really going to be the most impactful for the amount of budget they have available. So many different aspects of the pipeline.

Mike Toffel:

Got it. Let's look toward the future. And as you're seeing rapid developments in AI, you mentioned five years is a really long time in the AI development sphere, given the rapid pace of development.

And also, I imagine battery technology interest is only growing as people take climate change more and more seriously in the hunt for new energy technologies to try and increase the deployment of renewables, not just in the US, but around the world. Seems like both the demand signals are strong, and the supply capacity is strengthening. What do you see as the roadmap for the next five years in terms of both the development of these AI models on the one hand to help you do this work, but then also on the battery technologies themselves at the grid scale level where you're focusing.

Logan Ward:

Okay, yeah, so I'll start with the AI question first. The complexity of questions we can ask AI to help us answer has dramatically changed. A general idea of something I thought was unreasonable was why don't you just ask the AI model what the next material for a certain application is?

Three, four years ago, I would have thought that was a foolish proposition, but now there's actual good evidence that that's a viable way of doing things. There are enough ways if you just look at the literature alone that are reading text and newspapers and science articles that one could make a reasonable guess. Now it's not to say it's the way that's going to be the most competitive to solving AI problems now, but it illustrates the increased role AI can take instead of just being you are my equation now that I feed you inputs and you give me outputs, getting closer to being like a lab mate wild to me.

Now for the grid technologies, what I'm really excited to see is the diversification of the number of technologies that are available for storing energy. Early on in our conversation, we talked about pumped hydro, which has been around for a long time. There are flow batteries, there's lithium, there's sodium ion batteries. There are many variations of that.

Sometimes it's energy that needs to be stored for different amounts of time. Sometimes the geography. Sometimes the market environment means a technology is more expensive somewhere than it is else. And with these sort of really kind of explosion of numbers of technologies, I'm hoping we'll really start to find answers to a lot of the problems that we're seeing and how do we integrate renewable energy more broadly. So it's not just going to be one idea, it's going to be dozens of them in different ways of doing things, which is I like having more than one answer to a problem. It means the risk is lower of us having actually found the right answer.

Mike Toffel:

And do you see different sets of players entering the game in the next five years than have been around different government agencies, different industries or startups? What do you view as the major players? How do you view that change over the next few years?

Logan Ward:

Yeah, I see a lot of very young companies working on those technologies that have just made it out of the laboratory, Department of Energy or otherwise, and moving them forward to become noticeable impacts on how electricity is done. So there are technologies that we work with that come from the traditional already massive companies, but with that diversity of new technologies comes in a lot of new people and new companies. I certainly don't see any winners there. I see, or any one winner. I see many of them, just because the need is so large to be able to move clean energy forward as fast as possible. But also just a number of ways to look at that problem are so expensive that there are going to be a lot of opportunities.

Mike Toffel:

So you mentioned earlier about different ways that the national labs interact with the rest of DOE. But if I understand correctly, you're also interacting with other US government agencies, at least, for example, NASA or Space Administration, maybe others. Can you say a few words about how you interact with other agencies and maybe other agencies around the world?

Logan Ward:

Yeah, so they take both a variety of forms. I've got one project that involves NASA and some universities where we're all funded part of the same project. There's a specific outline of goals. This one's for electric aviation. Our team has an expertise in an area that's very helpful towards that problem.

There's another set of teams that have different parts of the issue figured out. So we're all working together there because finding the right tool across different agencies works very well. There are more informal interactions. A lot of the panels that are used to help the DOE find and understand better where it should place its next set of funding are things which involve other agencies investments in one part of the government affect others and internationally. I've been part of many discussions which involve representatives from other computing facilities in different nations all really focused on the ideas of how we exchange knowledge better across each other and ensure that we're exploring as many ideas as closely as possible.

Mike Toffel:

Yeah, I imagine there'd be some motivation by all of you to ensure that your efforts are stratified rather than duplicative.

Logan Ward:

Absolutely. That's a very key decision given the size of the Department of Energy, ensuring that we explore many different technologies thoroughly.

Mike Toffel:

So as machine learning evolves in the coming years, how do you see its use either in grid scale batteries or other types of batteries changing, or do you see it getting better at the two tasks that we discussed today of selecting candidate materials and simulating the properties of them before bringing them actually to physical testing?

Logan Ward:

Yeah, the way that we use AI will likely remain the same because those two processes of science, the hypothesis generation and experiment, are the way that has worked for centuries.

The tools that we're using are evolving and AI is one of those tools. We can use it to help us come up with ideas, pose evidence that might already exist to help us refine that hypothesis before we even design a test and then help us execute.

So one of the ways we're using AI and really exploring it heavily is how can we have the AI run the experiments better for us? Where we take a general concept, help me test this idea of a material, and have it design the experiment. What is the equipment that we have that could possibly do it? What is the procedure which we should specify? That's something that used to be a purely human endeavor and is still something that humans will do, but having that extra toolkit to help them really be the most effective is a way that AI is being used now in the early stages that I think will become even more important.

Mike Toffel:

Well, how did you get involved in this area yourself? What motivated, what led you to batteries?

Logan Ward:

so I got into batteries somewhat by accident. I had a specific area that I'm really passionate about. I love modeling materials at the atom level. That's my particular specialty. And in pursuing that, I got involved in teams that are looking at the larger scale problems. Like how do we understand how to design a material, a battery, who doesn't lose capacity as quickly?

There are many different aspects of that problem. One is understanding the atom level. So that's my role in the team. And I like working on the problems that have a clear impact on the future and clean energy obviously is one that is going to remain important probably for the next couple of decades. So being able to work on that little thing that I've got a particular passion in on a team of people who link together to attach it to that societal need is sort of the pathway that had worked together for me. And I presume that's how many of my colleagues brought their unique experience into being able to build that.

Mike Toffel:

Interesting. When you have a particular interest, there's sometimes lots of different domains one could lean into. And for you, this has had an impact on the future, which led you into energy to deploy your skills, which is very interesting. I think that that will resonate with a lot of our listeners who are trying to make a difference in some way in the climate space. So let me ask a question that I ask all of our guests, which is about advice. So I imagine as you give talks to different audiences, you must get the question, this is really interesting, the intersection between AI and chemistry and physics and thinking about the energy space, especially the important issue of grid scale batteries. And I know that you'd have different answers for PhDs in machine learning or PhDs in biochemistry. But what about for folks who are not in that set? So I'm going to try to bring it down to a level in their education where they may be more generalists or maybe they're engineers or finance people or general management people who would support the work but not do exactly the work that you're engaged in on the technical side. Where do you see opportunities for those folks to plug in? You already mentioned many labs. You've mentioned private sector startups. What are some pathways that you see for such folks?

Logan Ward:

Yeah. So they are really expansive. So I hope this wasn't too apparent in our conversation, but there's a lot of things about batteries that I don't know. And that, and I only have a clue about it because I work on a team with people with all kinds of expertise backgrounds. So really the being part of a role in say using AI for advancing energy technology is not about being good in any specific thing, but being good at working on problems that are too big to fit in your own head, too big to fit in anyone's mind. So really the skill sets there are continue doing what it is that you're passionate about and being good at working in a team such that they can all link together well. And that's something that transcends being able to work at a national laboratory, an industry or working on problems that are not just yours, but they're a collective problem. So the advice I'd actually lean towards giving your audience is the same one that I talked to young scientists about, which is become good at working on a team. That's the one thing that really makes solving these problems go quickly.

Mike Toffel:

Well, Logan, thank you so much for spending time with us here on Climate Rising. I've learned a lot about batteries and about machine learning and the discovery process at our Department of Energy Argonne Labs. So thanks so much.

Logan Ward:

Glad to. Thanks for the invitation. I enjoyed our chat.