Panel Discussion on Thermoelectrics
Thermoelectrics Session, MRS Memorial

Taken From "the Battle Against Phonons" Video
Taken From "the Battle Against Phonons" Video

Gang Chen: Let me get us started with some questions. My first question is: What are the major challenges for the field? Anyone want to take that on?

Jos Heremans: Applications — or, rather, the lack of applications. Like Jihui, I come from the automotive industry, and in 2008, there was a very nice prototype of a waste heat recovery system developed by a company that is now called GenTherm, and it made one kilowatt of power — which would have saved a considerable amount of fuel, and it was beautiful. It worked very well! But it was totally unaffordable. And so, although the whole automotive waste heat recovery also gave the thermoelectric field a boom — because that problem is a really important application in fuel economy — but it just didn’t work because it was too expensive. And that is actually a major problem for the whole field. If we’d had, back then, a material that not only had a good ZT, but was also totally affordable (and when I say ‘affordable’, I mean by two orders of magnitude; thirty dollars instead of three thousand dollars, for instance), then the field would have taken off. But that didn’t happen. Therefore, I think that major effort should really be an affordability — not just the materials, because materials costs are only about 30% of the thing — but the entire system cost. And affordability, there, means simplification.

A card sent to Millie in 2011.
A card sent to Millie in 2011.

Jihui Yang: I completely agree. We have actually very few people attending the MRS session on thermoelectrics from people who are working on engineering of modules and the systems. This is something that we may want to change, as a community. I was talking to Professor Chen from the Chinese Academy of Science, yesterday. Even on the material level, we have very few people working on lower cost thermoelectric materials — which is extremely important.

Arun Majumdar: Maybe I can give a slightly different perspective in the application area. We had a Paris agreement in 2015, and that is very well known. What is often not well known is the Kigali amendment to the Montreal Protocol, which pertains to HFCs.  Now, HFCs (hydrofluorocarbons, which are used in our cooling systems) have 2000-3000 times lower Wong potential compared to CO2. So the idea in the amendment is to phase out HFCs. This means that we need new technology for cooling. Simultaneously, the demand for cooling systems in emerging economies — which are mostly in the tropical regions — is about to go up. So you’ve got this double whammy coming.

I think the use of other technologies for cooling systems is extremely important, and thermoelectrics is going to have to compete. Now, large system vapor compression is very efficient. The COPs are on the order of 5-7, or so. But if you try to take a vapor compression system and make it smaller — it doesn’t work. Your COP goes down. And the rest of the world — the emerging economies, especially — do not need refrigerators as big as the ones we use in our homes. It’s okay to make smaller ones. So I think this idea of smaller cooling systems is very important. But it will have to compete with other systems.

On the scientific side, the challenge, frankly, is overcoming that Bermuda Triangle of electrics, which I mentioned in my talk.  It’s a strong one, and a very important one. And we really have to look at new science coming out of the science side. Whether you have different entropy — like the ones that you were talking about in terms of spin — or if you have correlated systems that have different kinds of degeneracy on the electronic side, that’s one. On the phonon side, we’re looking at various things we have been discussing for decades now. It’s an open area. I know that Gang is trying to do that, as well. Can we look at wave interactions of phonons and can we do Anderson localization? We have not been able to do so in a very definitive way.

So I think the introduction of new science into thermoelectrics is going to be the next big phase. And connecting it to some of the applications we’ve talked about — those are the big challenges.

Millie and Gene celebrate their anniversary at a Thermoelectrics Conference. Photo courtesy of the Dresselhaus Family
Millie and Gene celebrate their anniversary at a Thermoelectrics Conference. Photo courtesy of the Dresselhaus Family

Zhifeng Ren: I agree. Applications are very important. Anything we do, we’re using taxpayer money to do research. And if that’s only for our own curiosity — in the end, that will not work. We have to have something that will come back to serve the society that paid us to do the research. But right now, I think that finding more applications would definitely be urgent. If you look at the thermoelectric community, probably many more people work on materials — trying to improve materials and trying to discover new materials — but it’s still a long way to go, to get from materials to applications. And even then, even if you can find engineers who can make devices work, you need to really ask yourself... where’s the market? Right now, we do not have any end-user to really say, “I want to buy 1 million modules!” If we had that, it would trigger the whole system and materials development — better materials, better contacts, better devices. Therefore, I really feel that, at the moment, there’s no end-user. While it's true that thermoelectrics is very promising and has a lot of potential — potential for both power generation and for cooling devices — we still always need to ask ourselves, “Where is the market? Where is the buyer?” That’s the problem.

Mona Zebarjadi: I just want to add to that I completely agree. I’d also like to add that one of the applications that has, perhaps, not been the focus of our electronics thus far would be a system for the electronic cooling of chips. For things like that, there are not many options available. So maybe thermoelectrics could be an option, there.

Gang Chen: I’d also like to jump in on this discussion. I completely agree with Zhifeng.  Jihui probably remembers when DOE/EERE had this automobile project — I thought to myself… ooh, automobiles!  That’s the hardest application!

Millie delivers a thermoelectrics lecture at Union College.  Photo courtesy of the Dresselhaus Family
Millie delivers a thermoelectrics lecture at Union College. Photo courtesy of the Dresselhaus Family

But at the time, I was also really optimistic, and I kept saying, “I really hope this will be successful, because if this is successful, then everything is easy!” Because, of course, automobiles have a mass market, and making a breakthrough in the automotive industry would give thermoelectrics a lot of visibility. But automobiles have a weight-cost sensitivity, and the expenses were too high.

When Zhifeng and I were trying to commercialize thermoelectric applications in our company, GMZ, we looked at the combined utilization of thermoelectrics for hot water and electrical power. We generated entropy in a home furnace, where we have such a large ΔT, but where we can basically use hot water to heat up the house. Essentially, what we’re really generating is entropy. Therefore, I thought it might be good for the thermoelectrics community to look at getting into the market for these types of things.  It’s not necessary to start with the ambition of, “I’m going to conquer everything!” We need to have a product and really get into the market. That’s one.

Secondly, we know that any time we go from material to market, it takes a long time. Now, when GMZ first started, we were all very ambitious, and I remember Zhifeng saying, “Okay, we’re going to sell powders! We’re going to sell the material!”  But nobody came.

It turns out that if you think about taking a material to market, you first have to realize that you need contacts.  The black magic of companies are their contacts. You see, everybody in business spends a lot of time making contacts, but those who already have them don't want to tell new people how to make their own.

So while it’s true that there’s a lot of good basic science being done, and that the science we’re doing can translate to a lot of potentially interesting applications — it’s very different actually trying to create and sell that application.  You have to put a lot of time and effort and money into it, and although you develop some really interesting science, trying to make it all work, it’s the kind of science that’s hard to publish in Nature or Science or any good scientific journals.  The success of the applications really is defined by whether or not they sell on the open market.

So really, every step from contacts to collect the heats, reject the heat… there are a lot of challenges that need to be solved to take a material to the market. Those are really the lessons we learned.

Jos Heremans: I wanted to add a little bit to Arun’s comment. It is true that it is easier to get a high ZT in a waste heat recovery system, because of just the T. If you're going to be working at 1000 Kelvin, the T really helps you. But it is also true that the vast majority of the market is actually in cooling. It’s bismuth telluride, it's at room temperature, and the T is only 300 Kelvin. So if we look at cooling, now we have problems. Right? Because we have a lower T and, in fact, there is a demonstration possible that will show you that the ZT actually scales as T to the power of seven halves. That then gets us to these other new ideas.

Dresselhaus and Hicks. Photo courtesy of Paul McGrath
Dresselhaus and Hicks. Photo courtesy of Paul McGrath

The electron system conventions. Okay, so there are two types of correlated electronic systems that have been looked at for a long, long time. Jerry Mayhem’s review in the solid state physics series. And essentially, you can have either Kondo effects, or you have these mixed valence and valence fluctuation compounds. So all the good thermoelectrics and the mixed valence compounds. The Kondo effect doesn't really help us much. I mean, I don't think that is a very good direction, because what you have there is, essentially, the formation of a collective electron population around spin polarized atoms or impurities of the whole lattice. And so that gives you an enormous boost in thermopower at very low temperature. But then, it very quickly stops. And then the thermopower plateaus at a temperature that's about equal to the Kondo temperature. That's very low. And so, that's been around for 50 years, and nobody's ever really found a way to make that work into a high ZT. Mixed valence systems and valence fluctuation systems, that works quite well but the ZT is on the order of 0.2. And yes, it’s metals and suitable for cooling. But 0.2 is all we’ve ever had in 60 years. So I would suggest that, instead, we should look at totally new ideas.

Jihui Yang: I would like to add to Arun’s comment. Maybe a slightly less challenging problem could be looking at buildings. If you look at the total energy consumption of buildings, it's a gigantic number. Maybe we can borrow the idea of localized cooling from automobiles. And if you can set your thermostat at a somewhat higher temperature, we're providing a localized as-needed cooling system. That would be something worthwhile. And that particular application could be less challenging than automotive waste heat recovery.

Arun Majumdar: Let me just add to this discussion. Thermoelectrics has to compete. Whether it’s GMZ or Alphabet or others, we must try to compete. We are now entering a world of very cheap electricity from carbon-free sources. Wind and solar is getting extremely cheap. You put a solar panel on the rooftop, and you're getting electricity at less than 10 cents a kilowatt hour. And thermoelectrics has to compete with that for power generation. I think that’s a very hard proposition. But I think, on the other hand, if you look at the cooling side — something is being phased out because of regulation, and that means we have an opportunity. There’s an opportunity. One has to compete with other things. And that's where I think the opportunity is. I think someone mentioned — it's not just the ZT; it’s also the system level. Like you were talking about. The contacts and the heat exchanges and all of that has to be brought into the picture. But it has to compete.

Zhifeng Ren: So I would like to add to what you just said. Gang and I learned a hard lesson when we started GMZ (which you just mentioned). We are scientists. Right? We, as scientists, do not have money to develop commercial products. We need venture capitalists to give us money. So when we were talking to the venture capitalists, they asked us, “What’s the market?” We answered by talking about thermoelectric cooling systems. The venture capitalists went out and did their own study on this, and they established that cooling was about $200 million of potential market. Well, they looked at that and said, “Forget it! If you have something that will only generate millions of dollars, we’re not interested.  We need at least billions.”

Now, their study also found that with power generation, you can easily make multi-billions of dollars. So that’s where we ultimately wound up putting our money.

And then came the hard part, and we learned such a hard lesson!  Such a hard lesson. We spent so much time and effort. Eventually, our company developed its first product.  It was a low concentration that uses solar for the thermal concentration, and then, when combined with solar thermal power, they produced 1.8 m long glass tube and could generate 3 watts of electrical power plus the original hot water. It was beautiful! But then, all of a sudden, boom!  The PV price came way down. Completely killed our chances!

So, overall, as you can see, actually selling thermoelectric applications is very challenging.

Gang Chen: Let’s turn to fundamental science, because, as you go to a lower temperature, it becomes increasingly interesting to study the fundamental science side. We’ve heard about spin. We’ve heard about 2D materials. And Arun has told us all about localization.

What are the opportunities for all these different things? Could you elaborate more on that?

A project GMZ created using Half-Heusler thermoelectric materials.  Photo credit: Green Car Congress
A project GMZ created using Half-Heusler thermoelectric materials. Photo credit: Green Car Congress

For example, Zhifeng and I studied about this when we were trying to understand why some of the Half-Heuslers have a very high power factor, and we found that crystal symmetry plays a very important role in protecting the high mobility of the electrons. So thermoelectrics is very fascinating from the science point of view, because of the different fundamental challenges.

Maybe you can jump up the Bermuda Triangle, as Arun mentioned; that’s one way to bypass the challenge. But also, if we were to fundamentally understand electron-phonon transport, the impact of that would be felt far beyond the field of thermoelectrics. Over the past few years, we got into the computation of first principles.

So, what are the opportunities in the science of thermoelectrics and the potential impact?

Jos Heremans: I’d say topological properties of materials. That’s the ‘new frontier’. Of course in many other transport properties, once you go beyond the classic bath structure that has to do with the periodic potential of the atoms and electrostatic interactions between free electrons and atoms, you go into properties that are really due to the topology. This has exploded in the last decade or so, maybe less, but all the emphasis has been on electrical transport. And there is a little bit of emphasis — maybe two groups in the world — that are looking at thermal transport. And thermal transport is really completely different, because it adds this energy derivative of the electrical transport, and it's full of richness. That is for the electrons alone. Of course Jihui gave his talk on the phonon properties. I would dare say that there should be topological properties of phonon transport. Tom Lubensky has a fantastic Nature Physics paper on that. So there should actually be phonon edge states and things like that. We have never looked at this. This is totally unknown.

Mona Zebarjadi: What I wanted to point out is that even the simplest structures, as you mentioned silicon or gallium arsenide... I think one thing that we’re trying to do in groups like Dr. Gang Chen’s group and other groups, is that we’re trying to build a predictive model so that we can actually predict the thermoelectric transport properties before going ahead and making all those samples, so we can optimize it experimentally. And it’s a challenge. And a lot of the first principles calculations that we have available are very time-consuming to perform and not accurate enough to tell you exactly what the properties would be. And I think that’s an area that a lot of people are now recognizing and trying to work on, and hopefully, we will improve those theories.

Jihui Yang: In addition to the topology that Jos mentioned, I think there are maybe a lot of room for spin and magnons in semiconductors. This is a rich area which hasn't been delved into. I see a lot of opportunities, especially in low temperatures for cooling materials.

Arun Majumdar: I’ll just add that, from the electronic side, on the spin effect, certainly, I think that’s interesting. I’ve been encouraging Jos. Every time I see him, I say, “Okay, what’s new?” I’m curious to see what’s coming out of this, because there’s opportunity out here. There’s something out there that, perhaps, we haven’t explored.

On the phonon side, going back to what we were talking about, I don't find any physics that is missing in trying to show Anderson localization. But no one has shown it yet! And the question is why? And I think it is the anharmonicity of the phonons, which is, perhaps, not there as much in photons or in the electron wave function localization. That is really difficult. This is broadband. If you look at all the photon localization, it’s all monochromatic. But here, we are talking about broadband across the whole Brillouin zone, and trying to localize it is non-trivial. But that would be a big leap, if we can demonstrate that. I think that, on the phonon side, would be new physics coming out. I haven’t seen the paper in Nature on the edge states of topological insulators, but... yeah. That would be the kind of thing that would be very interesting and that has not been explored before.

Jos Heremans: It’s all theory.

Arun Majumdar: It’s all theory?

Jos Heremans: Yes.

Arun Majumdar: But that’s okay! That’s okay.

Gang Chen: I know we’re having a good discussion, but I also know that people want to ask questions. So I'd like to open it up to audience questions, please. Yes, John, could you please come up to the mic.

Question Asker: I’ve been energy conversion for over 40 years, and I especially followed thermoelectrics. And I'm very pleased to hear Arun bringing some new fresh air in. We must combine systems. The thermoelectric people, I think, are sometimes too narrow-minded. I’d like to know what your opinions are. What do you think is the near future? And how far away can we hope to have some medium-sized systems?

Arun Majumdar: It’s a good question. As I mentioned, in the cooling systems, thermoelectrics has to compete with other refrigerants. After all, we are using electrons and holes as refrigerants. It has to compete with propane. It has to compete with CO2. CO2 is a fantastic refrigerant which is used in automobile systems, etc. I think that’s the competition that we have to really look at. I can’t tell you when the thermoelectric systems are going to go mass market. What I can say is I have a wine cooler which is thermoelectric. But that's a very niche market! It’s not a very big market.

Question Asker: Well, there’s no universal product to begin with, in this market.

Arun Majumdar: Right. And so, we have looked at (as I presented) other ways of doing it. Entropy is not limited to only electrons and holes. We’re looking at electrochemical reactions and leveraging what is there and what’s been developed in other fields. I mean, there are tens of thousands of engineers working on batteries! And I know you guys have worked along that track as well. I think we should be looking at that carefully.

Question Asker: But how can we make the world enthusiastic to do it? How can we make the universities enthusiastic?

Arun Majumdar: Well, I mean that’s the applications... we want one company to be successful.

Question Asker: We need another Millie to have a bomb of an idea!

GMZ's generator for the army.  Photo courtesy of Business Wire
GMZ's generator for the army. Photo courtesy of Business Wire

Zhifeng Ren: I can add to this debate. You’re talking about systems. That was our effort at GMZ, as I recall. In fact, when GMZ did a DOE grant for the auto application, the Army did a piggyback for another $2 million for a one kilowatt power generator. In the end, GMZ did, indeed, deliver a one kilowatt generator to the Army. But where is it now? How does it perform?  We don’t know. So in the end, as I've said, we don’t have a customer. If there is a customer base, then I would have another one million such units, and as word of mouth spread and the market demand for the product went up, people would put more effort into making these things work better and more efficiently. But we don’t have a customer base. That’s the issue.  As I’ve said, the real problem is — where is the market?

Gang Chen: I remember that, at the end, GMZ was developing a self-powered boiler. Granted, at that point, we were also running out of money. But if we hadn't run out of money, I really think we could have gotten into the market with that kind of product — for two reasons. One is because you’ll save a little on your energy bills. But that’s not the major reason. The major reason you’d buy this would be for security. Imagine if you’re in a snowstorm or a hurricane or any other natural phenomena that cuts your power — I believe you can sell a self-powered boiler for that, and with a fair amount of success. But even keeping that in mind, what we found is that, in order to create this boiler, you’ve got to go to boiler makers (and that is a very stable industry). You’re going to have to ask the boiler-makers to redesign that boiler, and they won’t do that for you unless you have a few million dollars in cash on-hand.

Millie at her 80th birthday. Photo credit: Patsy
Millie at her 80th birthday. Photo credit: Patsy

Arun Majumdar: Can I add one thing? Since I’m from California, I can tell you that in California, most of the water heating in homes is done by natural gas. That’s going to be changing. It’s going to be mostly heat pumps in the future. So there’s the opportunity, where people are going to try to eliminate natural gas for heating because of CO2 emissions, and going to electricity. And I think there’s an opportunity. Maybe it’s not exactly the boilers, but using thermoelectrics for heating or cooling, in terms of heat pumps — I think there’s an opportunity there.

Gang Chen: Any other questions or comments? Yes. Could you come to the mic?

Question Asker: I enjoyed this session. I’ve got two questions. One is scientific. It’s clear that a transition to two dimensional structures has many advantages. Increased density of states and also phonon scattering, which is excellent. But what about electron scattering? For example, we tried the layered brutal and porous structures and electron scattering is much worse than the other advantages. Can you comment on this?

Gang Chen: I think this is a question that Mona wants to answer.

Mona Zebarjadi: Yes. We always have this struggle with electrons and phonons. And each time we try to suppress phonons, electrons are also suppressed a little bit. But that’s the whole idea beyond nanostructuring, right there, also. Because of the difference between mean-free path of electrons and phonons, you get the advantage to go small enough and to suppress only one and not the other that much.

Question Asker: But this is the limiting factor. It’s a few nanometers.

Mona Zebarjadi: Well, it depends on the material and it depends on the temperature range that you’re working in. And in terms of 2D materials, it’s not necessarily true, because... if you look at graphene, for example, it has a record high mobility and it’s just one layer. So if you look at the in-plane transport, that’s not a big deal. But if you look at the cross-plane transport then... you know. Also, if you just have a few layers, it’s ballistic transport, so you’re not suppressing anything regarding the electrons. If you try to make nanostructures, then you have to be careful, and we have to look at the mean free path and engineer it that way.

Question Asker: Is there any way to engineer the interface properties?

Mona Zebarjadi: Definitely. That is also another thing that you can do. So if you have an interface that allows electrons to pass with whatever mechanism, and blocks the phonons, that would, of course, be a good way to go, too. So it’s the same struggle that you have for bulk materials — you’re going to have the same thing for 2Ds as well.

Millie delivers a thermoelectrics lecture.  Photo credit: Geof Cooper
Millie delivers a thermoelectrics lecture. Photo credit: Geof Cooper

Jos Heremans: I would encourage you to dream a little bit, along these lines. So if you are going to play with dimensionality... Imagine that you have a potassium metal, so the Fermi surface is nice and spherical. Imagine, now, that you have a purely 2D system that would be potassium — the Fermi surface would be discs. So if you now had layers, with variable coupling between the layers, you would have to progressively go from a sphere to a disk, which inevitably brings you through a Fermi surface that would be a corrugated tube of some type, and you can imagine situations where actually it would become P-type in one direction and an n-type in the other. Because the corrugations would give you a P-type thermopower. So let your imagination bring you to this situation. This is this is a typical Millie Dresselhaus approach to things, right? But you can imagine intermediate situations in 2D materials where you can create amazing things!

Gang Chen: To add a little more on that, it turns that now, with the calculations, we realize that in these materials, because the electron mean free path is heavily doped, it’s actually very short. When we started this composite, we were not sure it would work, and one big reason why it did work is because the electron mean free path was very short.

Now it’s time to wrap up this session. Let’s thank all of the speakers and the audience.

Thank you all.

Photo credit: Xiaoting Jia
Photo credit: Xiaoting Jia