Battery DataBattery Intelligence
Together with Dr. Matthieu Dubarry find out about the latest discussions on #LFP batteries, degradation in li-ion batteries, and methods to estimate lifetime and performance. Matthieu has over 15 years of experience in lithium-ion batteries. He pioneered scientific methods that allow an understanding of the inner states of batteries based on voltage and current readings and recently published one of the most comprehensive synthetic data sets on battery aging curves.
Markham Hislop (00:10):
I'm gonna be talking to Dr. Matthieu Dubarry, who is the professor of the Hawaii National Energy Institute at the University of Hawaii about lithium iron phosphate batteries, which are becoming not as well-known as nickel, manganese, cobalt batteries, but are becoming much more common. So welcome to the interview, Matthieu.
Dr. Matthieu Dubarry (00:29):
Thanks for having me. Happy to be here.
Markham Hislop (00:32):
Look, can we start with an overview of what LFP is please?
Dr. Matthieu Dubarry (00:36):
Sure. So lithium iron phosphate is a battery matter you've known for probably around 20 years now. I think it's actually predates cobalt based materials. We have now, not lithium cobalt oxide, but the materials that are in batteries now appeared after LFP. But it's a really stable material. It has a really nice voltage plateau around 3.4 volts, and one of the main benefits is that it doesn't have any cobalt, nickel or manganese. So in term of environment, it's much more friendlier and bandwidths and also in terms of geopolitics where cobalt is not available for many countries. So that's why it's gaining interest now.
Markham Hislop (01:18):
Now I understand that LFP has cost advantage over the other batteries that use critical minerals. What kind of cost advantages are we talking about?
Dr. Matthieu Dubarry (01:29):
Well, that's not my specialty at all, so I'm not quite sure about how to answer that. I'm, I mean, iron is pretty cheap. Phosphates are fairly cheap as well. So compare that to cobalt, manganese or, and mostly cobalt and nickel are really expensive metals. So in that aspect, the raw materials, are considerably cheaper.
Markham Hislop (01:52):
Now, what are some of the other advantages of LFP? I understand that higher cycle life, for instance.
Dr. Matthieu Dubarry (01:59):
Yes. Mainly two advantages recycle life is one of them. Basically in that matter you only have one phase transformation. So lithium can go back and forth easily without too much strains. So that's why you have a long cycle life. And also, because lithium can go in and out easily, it's usually considered a high-power material, meaning that you can put a lot of current on it without too much problems, which is absolutely not the case for other cobalt-based materials.
Markham Hislop (02:30):
Does that mean that with LFP batteries and electric vehicles, for instance, that we could have faster charging times?
Dr. Matthieu Dubarry (02:40):
Well, yes and no. You need to consider the negative electrode. So yes, you can go really fast on the positive, but if you cannot go really fast on the negative, you cannot really also do fast charge. And the limitation for fast charge is mostly on the negative and to avoid issues like lithium plating and all those aspects. So, it could help definitely, but it's not the magical recipe to have battery that can charge in two minutes next year.
Markham Hislop (03:08):
Now I understand that LFP has a higher power density. Now we often hear about energy density of batteries, but maybe you could explain power density please.
Dr. Matthieu Dubarry (03:17):
Sure. So first you have to realize that all those materials, they have a specific voltage as well as their function. And your typical cobalt battery is around, average voltage is probably around 3.94 volts. LFP is much lower than that. The average voltage is probably around 3.5 volt, 3.4 volts. And so if you think of energy, energy is gonna be the capacity times the voltage. So LFP also has a less capacity per mole, more man, both over materials. So in terms of energy, it's not as good because you have a lower voltage and you have a lower capacity. So when you multiply both of them, but energy, it's obviously lower than for cobalt based materials, but you can go really really fast. So even though you have those limitation power-wise, when you just apply your current types of voltage, yes, you have a lower voltage, but you can apply a lot of current. So in terms of power is better, but in terms of energy it's not as good.
Markham Hislop (04:20):
And safety, of course is a big concern with NMC batteries. But I understand that with LFP batteries higher safety and lower toxicity, sorry, is that correct?
Dr. Matthieu Dubarry (04:33):
Toxicity? Yes. Safety, it's, yeah, it's a bit safer. It's really gonna depend on the additives you have in them. But in term of the energy release during thermal runaway LFP batteries release less energy than cobalt-based battery. But that doesn't mean that they cannot catch on fire. But that depend on the additives. Some manufacturers have showed additives for LFP batteries that prevent thermal runaway. So, it's really gonna be, dependent on the manufacturer. You cannot really generalize it. The truth is if less voltage, so you store less energy. So obviously if it fails, it's gonna be less energy.
Markham Hislop (05:18):
Now with the lower toxicity, does that mean that LFP batteries will be easier to recycle?
Dr. Matthieu Dubarry (05:24):
Yes, but I'm not sure the toxicity is the reason. It's a completely different structure than the cobalt oxide batteries. So it's completely different type of material. So I think the recycling for LFP is easier. I'm not sure if it's because of the toxicity or just because, if it's a phosphate and it's completely different structure and completely different chemistry than the cobalt-based batteries. But it is easier to recycle and to some degree it's more recycled than, than cobalt-based batteries already.
Markham Hislop (05:56):
Now I understand that LFPs have roughly 50% of the energy density of NMC battery, but is it the case that with changes to the anode that can be brought up to about 70% 75% of an NMC battery?
Dr. Matthieu Dubarry (06:16):
Well, no, because they have the same negative electrode. So if you change the negative electrode for the LFP, if you change it the same way for the NMC, you're gonna get the same results. It's really intrinsic to the materials. It's correspond to the amount of lithium you can put in compared to the mass of your material. So that's the set theoretical number. I think LFP is around one 30 milliampere per gram and the high nickel NMC materials are around 180 or 200. So, you can, that's nature, you cannot go more than that.
Markham Hislop (06:56):
Now I understand that LFP batteries are already commonly used in China for scooters and small EVs. How extensive are they used there?
Dr. Matthieu Dubarry (07:07):
As far I understand, yes, we have some, quite a bit of EVs that are LFP based over therein scooter too. And it is been using power tools for a long time as well. So commercially, those batteries have been available for, for quite a long time. Even in the US there have been some US manufacturers that produce both batteries,0 to 15 years ago.
Markham Hislop (07:28):
Now, I understand that some of the automakers like Tesla and VW are moving to LFP batteries. What's the reason for that?
Dr. Matthieu Dubarry (07:41):
Well I'm not sure. I'm not in their shoes. I think partially the reason is depending on what EV you making, you might not need a lot of range and maybe for small EVs, but are meant to stay in the city. LFP batteries might be good because you don't need as much range as the vehicle that are meant to go in between cities. So To some degree, I think, yeah, you get advantage of safety, it's easier to recycle, it's less toxic. And I think for some running downtown it's a perfect battery. If you don't do 200 miles per day, it's perfectly fine. So, I think they start to realize that you cannot have one single battery for every application, and you might not want to use the same battery, in Montreal, Bali and Honolulu. Obviously, the weather conditions are completely different. The roads are completely different. So it's probably an effort to adapt the battery to the road condition to maximize life and safety too.
Markham Hislop (08:43):
To a, a layperson like myself who isn't an electrochemist it looks like there's beginning to be a movement to, or a settling out of different battery chemistries for different applications. Is that a fair comment to make?
Dr. Matthieu Dubarry (09:00):
Oh, yes. And, that's absolutely expected. Battery degradation, lithium ion battery degradation is extremely complex and it's gonna depend-it's called path dependent. And that mean by depending on how you use, the battery is going to degrade differently. And it's not just using capacity slower and faster, but it's really the internal mechanism that are completely different. And that could lead to sudden death of a battery or not. And sometimes the difference is really, really subtle. And that's also to add to the complexity that's also chemistry dependent. So every battery chemistry gonna be really sensitive to some stress and not sensitive to others. And some of the chemistry, maybe the complete opposite. So, if you think of for an electric vehicle the temperature is gonna be a big role and the road condition, somebody they're gonna be driving the car mostly on open highways, it's not gonna be the same degradation as somebody that spends years in traffic or that's gonna drive on a mountain road or that gonna drive only flat roads. So all of that is gonna affect the degradation. And so some batteries might be really good for some condition and really bad for others.
Markham Hislop (10:13):
Now you've done a fair amount of research on degradation. Can you tell us a little bit about your work?
Dr. Matthieu Dubarry (10:21):
Of course. So we specialize in diagnosis and prognosis for lithium ion and we develop techniques to do that in operando, meaning that we don't require to open the cells, we don't require to do anything destructive to the cell. We really focus on trying to develop method that we can apply on any battery out there. So what we do is we analyze the voltage response of the cell. You don't realize it, but when you look at the voltage of a lithium ion battery, you see a lot of change of slope and some plateaus or slopes and it's not a monotony decrease of a voltage. And that's information, that's what you see is the change of structure in the electrodes. All the lithium is accommodated inside the crystallographic structure of a positive and the negative electro. And so we develop techniques to kind of use that information to go back to how the cell degraded and we cannot tell you exactly what happened, but we can tell if some lithium is lost, the lithium amount in your battery is finite. So if something eats some lithium, we need to know it because that's gonna use capacity loss. So we can quantify that. We cannot tell you why with those non-destructive methods, but we can quantify how much lithium is lost or how much material is lost, or if the kinetic of your battery is degrading. So we developed a set of tools to diagnose cells really efficiently just based on the voltage response of the cell.
Markham Hislop (11:49):
What are the real-world applications for that research? Or will it be primarily for electric vehicles maybe or is stationary batteries or all kinds of lithium-ion batteries?
Dr. Matthieu Dubarry (12:02):
Oh, it's for, for every lithium ion and you can extend that to sodium ion or any of new fancy batteries they're gonna come up with. I mean there's many applications, obviously one of the main application is for us in the lab when we test the impact of different conditions. Or I mean us or any other battery manufacturer, when we test batteries to see are they gonna be, what's happening when the battery is at high temperature, what's happening on the battery is discharged really, really fast and what the degradation and can we prognosis that? Does it gonna lead to thermal runaway later? So we can use the technique for that. Since its all only voltage based, we can also embed the technique into the battery management system so that your battery can constantly try to self-diagnose itself and to able to flag if we start to see degradation path that might need to drastic failure. It could automatically flag and detect what is gonna fail soon.
Markham Hislop (13:01):
Is it fair to say, and this is certainly my impression, you know, looking at your industry, but is it fair to say that the longer life better performance, lower cost of batteries is to some extent, maybe a longer extent dependent upon the kind of analytics work that you are doing so that you take a battery that, you know, maybe has a lithium ion, an NMC or an LFP chemistry, but it performs so much better and longer because it's managed better using your software and your kind of analytics approach?
Dr. Matthieu Dubarry (13:40):
Yeah, I mean a lot goes to understanding what condition that battery like or don't like. And I could show example where we took actually study we did on LFP battery, which one was really interesting, where we took one cell where we did fast charge and fast discharge. We fully charged and fully discharge in 15 minutes. It's really fast for a battery and for the same cell, I mean another service and batch we still charge in 15 minutes, but discharge we simulated driving for three hours. So on paper as far less aggressive, but the results the server did, the driving died after 600 cycles. The cell that did the fast charge and discharge lasted 5,000 cycles. So just to show you that sometimes even the less aggressive cycle could degrade bad cell. So it's really important to understand how to test the cells because obviously in that scenario we could have say, oh, bad cell last 5,000 cycles and sell it to some EV manufacturer where in fact if you use it for driving it's not the case, it's gonna die right away. So it's really important to do a lot of testing to understand what are condition that cell is good for and then if you do that, then you, you can deploy it and it's gonna last forever. But even sometimes a small change in the condition could mean that the cell gonna die after a year instead of 10. So analytics plays a role to monitor, but a lot of work has to come beforehand in the lab to understand what conditions gather, what degradation for the cells.
Markham Hislop (15:11):
What are the top two issues or trends in your work that you'll be dealing with over the next decade?
Dr. Matthieu Dubarry (15:22):
Well, I mean the battery field is really, really broad. So it's, it's, I mean are you talking about materials, are you talking about controls? It's, it's really broad. For my field is more on the control and the battery intelligence side of thing. And I think the key there is gonna be how to add some material science understanding to all those algorithms. But it's really easy to develop algorithm to monitor data. But if you cannot link that to what happened inside the battery in terms of where lithium goes or what's happening in the battery, it's never gonna be successful. So, to me the big job is to link having conversations between material scientists and understand what's happening inside the battery and people that are really good at building algorithms and doing all that machine learning. I don't think the battery is the one man word anymore. We need a team of people that are specialists on their own little thing and they learn to communicate together to get results.
Markham Hislop (16:25):
Are there any, maybe one or two examples if you have them of really exciting developments in batteries, particularly LFP, that we might see in the next five or 10 years?
Dr. Matthieu Dubarry (16:40):
LFP, I mean honestly on the research side of things, I haven't seen much work on LFP in the past decade. LFP was, every publication was on LFP 10 years ago. Since then LFP is not that many publications on it. I think, I mean our publication that I told you before where we simulate the driving or the fast charge, I think that one is really I think, important paper for the field because it shows the dependence. I think in terms of diagnosis was some synthetic data for LFP. So we use our method to simulate a lot of voltage curve and we simulated every possible degradation. We don't know if it's possible in real life. We don't know what could cause for degradation, but we simulated every possible degradation and we upload data for free. Anyone can download it and the goal is to help people validate algorithm so they can take that LFP data with 700,000 individual voltage curve and they can try our algorithm to see if they can diagnose it correctly.
Dr. Matthieu Dubarry (17:51):
To me that's a big development for the field when we have more and more data driven methods and it's great, those methods are fantastic, but they need good training data. You cannot expect to test one or two batteries and develop algorithm that's gonna be universal. That's why we invested a lot of time and effort to develop those synthetic data sets so that we can give people that doesn't necessarily know how to test cells or have resource of time to test lot of batteries under different conditions. So they can take our synthetic data and validate whatever algorithm they have, It can be machine learning, it can be a normal algorithm or anything they want, but they have a big data set they can use to validate. And to me that's really exciting for the future to keep providing those synthetic data under different conditions. We can simulate different temperature, we can simulate different rates, we can simulate a lot of different aspects so people can try and make sure that algorithm is valid under any given conditions, which is extremely important. Again, going back to my example of changing the condition a little bit might completely degrade the cell different. So you must be able to flag that with your algorithm. So I think to me what really interests me these days is synthetic cycles and how to replicate the battery. Having better models that we can that are not calculation intensive so we can do millions of calculation to validate any possible algorithm.
Markham Hislop (19:16):
Matthieu, thank you very much for sharing your insights. Really appreciate it.
Dr. Matthieu Dubarry (19:19):
Oh, you're more than welcome. Always happy to talk about the battery industry.
Get battery analytics trends and industry knowledge delivered directly to your inbox. Because we batteries.
Don't miss the new episodes of the podcast. Sign up for our newsletter.