【Expert Interview】Academician Hao Yue Exclusive Interview: The application prospect and future development of ultra-wide bandgap semiconductors

Hao Yue

Academician, Chinese Academy of Sciences

Hao Yue, Academician of Chinese Academy of Sciences, internationally renowned microelectronics scientist, professor and director of Academic Committee of Xidian University, deputy to National People's Congress, Vice chairman of Chinese Institute of Electronics.

He has been engaged in scientific research and personnel training in microelectronics and solid state electronics for a long time, and has made systematic and innovative achievements in wide band gap semiconductor materials and devices, microwave and millimeter wave semiconductor devices, new integrated circuit devices and new materials. He was elected an academician of the Chinese Academy of Sciences in 2013.

At present, he is also the director of the Information Science Department of the National Natural Science Foundation of China, the convenor of the Electronic Science and Technology Discipline Evaluation Group of the Academic Degrees Committee of The State Council, and the director of the Electronic Information Teaching Steering Committee of the Ministry of Education.

Q Dr Lu Min  Editor-in-chief, Compound Semiconductors

A Academician Hao Yue  Academician, Chinese Academy of Sciences

Q  Then we're done with Silicon Carbide. In fact, one direction that we are more likely to pursue in the academic world is that some people currently call ultra-wide bandgap semiconductors the fourth generation semiconductors. Do you agree with this statement? Why?

A  I think the third and fourth generation, as I said at that time, we should still call it wide bandgap or ultra-wide bandgap from academia and industry, which is in accordance with international practice. Then the term "third generation", I have always said, is a popular science statement, that is, how to tell the people more easily understood, it is called the third generation.

In fact, the whole semiconductor is now, the first generation is still mainstream, silicon is still mainstream. More than 90% of us are still silicon, silicon materials, so the third generation, the fourth generation of the name, especially the fourth generation, some experts and entrepreneurs have asked me whether it is appropriate, I said you want to become a generation of materials, you must have industrial background support. If these have been industrialized, large-scale, then we say to become a new generation is OK.

Now most of them are still in the research stage, and in what aspects the industry will have updated breakthroughs in the future. Now objectively speaking, there is still a period of time, this can be worth waiting for, there are uncertainties, you call it a generation of semiconductors, then it means that it has been industrialized, I think the society or more or less will have a certain misleading, misleading because when it comes to the fourth generation, it still thinks that it is very advanced, and then our investment, the government began to pay attention to these aspects.

But I think it will take quite a period of time for it to really be applied, so I think it is better to be cautious. What do you call it is more convenient for communication, easy for people to understand, and perhaps also to facilitate the need for financing.

Q  Still has its value. What do you call it is more convenient for communication and common people to understand. But be sure to clarify the background of why you say this to him

A  But when you talk about it, people think it's a good time.

Q Thought it was time for the fourth generation to replace the third generation.

A  It's a kind of information that's not accurate, so I think we need to be more rigorous in this area.

Q  So Academician Hao Yue needs to do more definitions of this kind of science popularization.

A  Yes, many experts have told me that we should boldly talk about the fourth generation of semiconductors. I say boldness is one aspect, but you have to at least see that some of our materials, ultra-wide band gap some of our materials already have very strong, not local, but very strong prospects for industrialization, right? We say that in this case, we need more attention from the society, we need more capital to enter, at this time I think to put forward a new term, this may be conducive to this aspect, but if you do not have the opportunity, you put forward may give us a whole misleading.

Q  Propaganda is necessary, but propaganda must be measured and based, right?

A  It is not foundation. The timing is important.

Q  Ok, so we know that Gallium Oxide should actually be the most active in ultra-wide bandgap semiconductors at the moment, and should actually be relatively mature. What are the main scientific and technical problems in the field of Gallium Oxide?

A  Gallium Oxide because our team should talk about in the domestic and even international is still relatively early, up to now it has been 4~5 years or even longer.

I said Gallium Oxide its advantages are outstanding, ultra-wide band gap, band gap width 4.6~4.8ev. But it has its weaknesses, one of which is that it is a monoclinic crystal system in terms of crystals. Unlike our Gallium Nitride hexagonal, Silicon Carbide cubic/hexagonal. In addition, Gallium Oxide is the same as Silicon Carbide, it is multi-type, and now it is generally believed that β type is relatively stable, like α type, like Silicon Carbide is more, more than 200 kinds of crystal type.

In this case, I think if we want to talk about what aspects we still lack, first, I think this substrate material needs high-quality substrate materials, need to continue to break through, there are several companies in the world, several companies in Japan can do, the domestic time is relatively early, but the quality is general, this is the first aspect. The second aspect also needs to solve the problem of Gallium Oxide thermal conductivity, because the thermal conductivity of Gallium Oxide is too low.

Q  Worse than silicon, isn't it?

A  Worse than silicon, and even worse than many dielectric. So its thermal conductivity must be solved, that is, the heat dissipation problem to be solved, if the heat dissipation problem is not solved well, its role will not play out, we have done a lot of work in these aspects in the early stage.

Then Gallium Oxide P-type doping, oxide semiconductors have this problem, Zinc Oxide, IGZO including Gallium Oxide P-type doping are difficult, of course, Gallium Nitride P-type doping is not easy. Because of the wide band gap, any impurities into most of the energy levels are too deep, so it is difficult to ionize at room temperature, it is difficult to activate, so in this case, P-type doping, is also a worthy of attention. If P-type Nickel Oxide is used to make a heterojunction, now there is also this way, even PN Junction can be done, P-type Nickel Oxide, N-type Gallium Oxide.

Q  Is such a workaround, what about performance?

A  The performance now looks good, there are Nickel Oxide, there are other materials can also be developed more, this is in the research and development stage, the potential is still there, but this is still very difficult.

In general, in these aspects are the need to focus on. In terms of the device itself, the progress is fast, the device is now the fastest progress, power density, integrated voltage, etc., it can be said that every day is doing, are making progress. Devices are mainly in the academic community, paper I am more concerned about, this progress is very fast, the world is concerned, when we do not have too much attention, even the academic community feel unlikely, heat dissipation is in this case how can it be? But with some of our work later, we feel that the heat dissipation problem, we transferred it to Silicon substrate or Silicon Carbide substrate, feel good, also transferred to diamond, diamond because the cost is too high, which means that even academic can be transferred to diamond, future applications are also a threshold, because of the cost, unless in some special super power of this application, The general amount of this large, wide range, but also need low-cost, so the transfer to Silicon and Silicon Carbide is still a very important aspect.

Q  Well. That's actually all developing, whether it's materials or devices, of course materials are definitely the first thing that needs to be solved, then you push back, at the application level, people actually care about the market is the application. Because now these application scenarios are actually occupied by other materials, that is, do you think this Gallium Oxide is most likely to first appear in which application scenario, or corresponding to Gallium Oxide is which power level, which voltage, which current scene?

A  I think Gallium Oxide is most likely to be the first step must be diode, directly compete with Silicon Carbide, that is, with Silicon Carbide SBD, to replace its part, because its on-resistance is lower, so this will directly in a very short time, I estimate there will be a market in this area, this is I believe that whether it is high voltage or low loss, Will be in the diode side. Later, it must be in the power electronics triode, and now we also try to do microwave, for example, its frequency can also do tens of G frequency, which is very good. For example, last year we had an article in IDM about Gallium Oxide microwaves.

Q  What (advantages) does it have?

A  It is low on resistance, low loss, because the frequency characteristics are good, but its current weakness is that its current is still relatively small, the power can not be done up, but its loss is really low.

Q  Low loss, just in the small power area

A  Yes, because microwave, the loss is low, it means that it is high efficiency, so in our microwave field, we pay special attention to efficiency, in the end how much this conversion efficiency can achieve, this is very important.

Q  Ok, then we know that the current Gallium Oxide substrate is actually considered by the industry to be a melt method, which may be done very cheaply, of course, in fact, it is not so easy.

A  Right, it's not that easy

Q  So in this context, making devices, there are actually different routes of homogeneity and heterogeneity epitaxy, how do you evaluate that?

A  I think now to do the device to really make it industrialization, large-scale industrialization, or homogeneity, this means to do on the homogeneity, in any case, that sapphire can now do a little, this quality is definitely not like the quality. That is to say, use a homogeneous substrate, and then peel it off, and then transfer to other substrates.

Q  Because homogeneity is, after all, a good epitaxial quality.

A  Not only epitaxy, that is, transfer, but also much better homogeneity, to really make devices, the quality of homogeneity is still much better.

Q  Do you think the heteroepitaxy Gallium Oxide has a future?

A  I think I can do it academically anyway, the industry has a certain difficulty, now many are ALD, MOCVD heterogeneous above a lot of long, mainly sapphire above a lot of long. Are talking about the quality is good, but I will say that the quality is good, and ultimately rely on the device to explain.

Now mainly used to do more detectors, is to do the deep ultraviolet detector more, this because the detector on the material requirements are not so high, but also belongs to photoelectric, it is mainly absorption, the main absorption of light can produce a pressure difference is finished, then really want to do electronic devices, is to take heteroepitaxy to do devices, do less good.

Q  Ok, so actually in the field of ultra-wide band gap, in fact we know that there are Aluminum Nitride, Aluminum Gallium Nitrogen, Diamond, and of course these three are a little bit behind. Then you can also evaluate these three directions in general.

A  Because when Aluminum Nitride and Aluminum Gallium Nitrogen, in general we are called III-V group. In fact, the Gallium Nitride devices we do now also use this barrier layer, and most of them use Aluminum Gallium Nitrogen. So using Aluminum Gallium Nitrogen as a substrate, I believe the question you asked is not to make this barrier layer, but to make a channel, called a channel layer.

Aluminum Gallium Nitrogen as a channel layer, I think it can be studied academically, because the biggest advantage of Aluminum Gallium Nitrogen is its wide band gap, which is wider than Gallium Nitride. So this is its advantage. But its weakness is that it is ternary, ternary is often used as channel material, it will encounter a lot of problems, such as its components are not uniform, which will form additional scattering.

So in this case, you say that Aluminum and Nitrogen, Aluminum and Gallium components in the end is zero, in fact, it has different inhomogeneity in different places, because you use it as a barrier layer does not matter, it is just the regulation of charge, the problem is not big, but you really use this channel, then it will cause some scattering caused by this component is not uniform, So I think the academic community can do this, I think that the ternary is used directly as the substrate is still a challenge, of course, so that IGZO can also be, four can be, but I think the more elements, the more complex the non-uniformity of its components.

Aluminum Nitrogen I think the future is promising, but the biggest advantage of Aluminum Nitrogen is its band-gap width of 6.2eV, but because it is so large, it has encountered a lot of problems, doping is a problem, ohmic contact is a problem, and Aluminum Oxidation is also a problem. If you want to really use this thing, after the content of aluminum is high, it is easy to oxidize, so why is the substrate of Aluminum Nitride difficult to epitaxial? That is to say, there is a natural layer of aluminum oxide, it must be removed before it can be further processed. Those are the challenges, especially the ohmic contact, you have to regrow, regrow the ohmic contact layer, like Gallium Nitride, or regrow Aluminum Gallium Nitride, and then this middle part of the device is made with this Aluminum Nitride, this is the first part.

Second, Aluminum Nitride, in general, its mobility is not high enough, much worse than the migration of Gallium Nitride. But its advantage, that is, the wider its band gap, the lower the loss. Diamond, these aspects we are doing. The biggest advantage of diamond, good heat dissipation, all this material is diamond heat dissipation is the best, the thermal conductivity is very high. The problem is that its atomic number is too low, so its bond energy is too strong, so it is difficult to doping this doping, he is not talking about doping what, the key to form a substitution, more difficult.

Another is to say that whether it is N-type or P-type, its activation rate is quite high, so it is very difficult to engage the body. At present, for example, with hydrogen terminal and silicon terminal, a two-dimensional hole gas is formed, and now this is the mainstream. And now we're doing a lot of work on that. And what we're hoping for now is that we can find an n-terminal, a two-dimensional electron gas, and that's good.

Q  In fact, diamond is this N-type doping difficult.

A  N-type doping is difficult, P-type doping is simple, but it is generally difficult, Boron also has several hundred meV in it, generally less than 100 is good, several hundred activation rate will come down, such as three or five hundred. So this activation rate is too low.

Q  Actually, people say that diamond is the ultimate semiconductor. How do you evaluate this?

A  I think that's right. Because diamond it is indeed overcome after, can solve a lot of problems, I can only say that the solution will not be used in general, only in some extreme cases, super power. This requires that in the case of super power, its temperature can not be high, I think the ultimate semiconductor is right, in this case I am still more looking forward to, and finally diamond if you can break through, then our real high power low loss can be achieved.

Q  That's the ultimate semiconductor, do you think it's possible for carbon to replace silicon, to extend to microelectronics, to logic?

A  That's a more sensitive question, and I usually don't like to talk about something too sensitive. Since you asked, I will say so. I think it is very difficult for an industry to change from one material to another, especially for a large-scale industry. You're the one who asked me, can carbon replace silicon? So it should be said that silicon integrated circuits have been around for more than 60 years today, 66 years this year.

Internationally in this human, material and financial resources, I do not think any industry investment is so large, You want to replace it with other materials, you must be really can not go to the extent of it. If there is any possibility, I think it is unlikely that there will be other materials to replace it, as a supplement is OK.

That is, additional supplements, heteroepitaxy on the above, especially now there are more heteroepitaxy, then you can make some supplements on the above is OK, such as Memory, such as some End Connection, are possible. But to say completely do not silicon, do not silicon devices, do not silicon IC, completely use other materials I think so far, this technical route is not clear, the future is not clear.

Q  Right, because silicon now Moore's Law is still pushing forward.

A  All kinds of things are being done.

Q  This one I see they're planning it down to 1 nanometer.

A  But you have to count this thing, I often say why integrated circuits can be infinitely reduced in size, and why optical devices are not small, why can't optoelectronic devices be reduced? Everyone may think that I optical integrated circuit, of course, when it comes to optical integrated circuit, maybe some experts think that teacher Hao should not comment more on this integrated circuit of light, but now it is not photonic integration, is not also a very important silicon-based photon.

I think no matter which aspect of you is related to the wavelength of its elementary particles, its scale is related to its wavelength, you can not violate this law.

Now you make devices with light, so the wavelength of a photon, now the infrared wavelength for example, is a few microns a bit. A few microns of light communication among us, even if it's ultraviolet, the current ultraviolet laser, like 193 nanometers, even if it's extreme ultraviolet, you make 13.5 nanometers, that's still at least 10 nanometers. So why don't the electrons take this into account, so it's integrated again, you can't be shorter than the wavelength. So the size of the device depends on the wavelength. The de Broglie wavelength of the electron is a few tenths of a nanometer, just like the wavelength of the matter wave is a few tenths of a nanometer.

Well, humans have started to think about a few tenths of a nanometer in silicon, and if I were to be objective and use other materials to calculate the de Broglie wavelength of electrons in other materials, then you can know what our human limit is. This is a law that we can't break,

Q  This is the first principles of physics.

A  So in this case, the optical integration of more than 1,000 devices, 10,000 devices is very large scale, but your integrated circuit is now 100 billion. On a single IC, 100 billion of that transistor. So so the gap is from 10,000  to 100 billion, so how do you compare the two?

Q  So that creates a problem, in fact we are talking about this field of microelectronics research or application of electronics, electronics of course, it actually brings another problem, which is power consumption. Therefore, in the future, the problem of power consumption is not always proposed. To solve the problem of power consumption is to use photon computing, quantum computing, photon computing that is not behind the optical computer, quantum computer, its power consumption can be done down, but listen to your talk about its size is difficult to make small, right?

A  It's hard to make the size small and the features big.

Q  So how would you compare the quantum computing route in the future to the traditional microelectronics route we have now?

A  Now a couple of paths in quantum computing, because I don't do quantum computing, so because in the long term in the foundation has heard about these aspects many times. Because I don't do quantum computing, I think quantum computing now has several ways, one way is superconductivity, and another is light.

Because I think this method of light, it is very difficult to be industrialized, such a technology of pure light to be industrialized, or think that it is mainly a problem of scale, the system is huge, and the industrialization is still very difficult, but it can solve some special problems and special applications.

Q  Special applications have no volume limit on size.

A  Then do superconducting quantum computing, which I think has a very important prospect in the future, so the international including IBM, Google are doing superconducting quantum computing anyway. Domestic also do superconducting quantum computing, superconducting quantum computing is actually mainly at low temperature, how many mK to this, in fact, it is also very troublesome. Later I thought, if we electron you give it down to low temperature mK level, it is not also very good. Of course, it doesn't need to go down to that level.

Q  At low temperatures, the wavelength of the electron decreases.

A Right, so I'm pushing a low temperature electron on the electron side, which is how integrated circuits, at low temperatures, the temperature is a little bit lower, you can also push its limit further, you can also increase its ability further forward, which is also a way.

So I think we are all working hard on this. In the future, I think quantum and electron will gradually replace some of them. I think the best way is the current super-large quantum computer and large electronic computer, supercomputing.

Quantum can replace some of the current electronic computers in some special aspects such as matrix operations and vector operations. For example, the proportion of the current calculation can be said that 5% is calculated by quantum, 95% is calculated by electrons, and slowly move, just see how fast you can develop on this side. Slowly move, see how common it is.

Now because they are very special, mainly vectors, slowly can the last 10%, 20%, so that step by step, called electron quantum fusion.

Q  This is like competition between materials

A  Electron quantum fusion, so as to crack the development of our high computing power, AI, big data, I think it is possible. Otherwise, if you continue to engage in quantum for a few years, you will not apply it, and the industry will slowly lose patience, so it has now reached such a stage.

Q  So there is no one size fits all, it's all the right scene for you.

A  Because an industry, in addition to science and technology, there are also markets, costs, and people's psychological bearing capacity. When the LED came out, you said you save energy, people said I don't buy, you are too expensive. You no longer energy saving, I do not think, and now automotive electronics are the same, said Silicon Carbide is very good, Gallium Nitride is very good, why?

But expensive so I would rather use silicon, silicon is cheap, although its energy consumption is higher, but in the minds of ordinary people, driving in the mind, this is not directly reflected, the body feeling is not too strong, but you want him to use so much money to buy, he can immediately appreciate, because I used Silicon Carbide, I used wide band gap semiconductor, so I want to increase the price, Then the people think that I would rather lower the price is enough for me, and this LED is the same, you must put its price down, the most important market is to rely on cost.