Transcript of Dr. Supiryo Bandyopadhyay Interview. Inigo Howlett, 88.9 WCVE, reporting for Science Matters, April 2016

Transcription

1 Transcript of Dr. Supiryo Bandyopadhyay Interview Inigo Howlett, 88.9 WCVE, reporting for Science Matters, April 2016 My name is Supiryo Bandyopadhyay, and I work in various areas of spintronics, especially an area called straintronics. I m a professor of electrical engineering at Virginia Commonwealth University in Richmond. And my areas of research are spintronics, and especially one area of spintronics we developed here at VCU called straintronics. We have couple of patents in this area, one has to do with making computers more energy efficient. So some of the things we do here with straintronics involve switching magnets magnetization Direction tiny little mechanical strains the magnets and the strain is generated electrically with an electrical voltage. The electrical voltage is very small a lot smaller than what it typically is used in switching transistors and microchips and others types of circuits like in your cell phones Etc and because the voltage is so low you have to expend very little energy to switch these magnets so the fact that you might be able to switch them without a battery and just harvest energy from the surroundings like TV networks or radio or even passing wind passing traffic vibrations caused by the wind, et cetera. So if you can generate devices that can work without a battery then that becomes very attractive it's especially attractive for medically implanted devices for example a pacemaker or on a device implanted an epileptic patients brain that will monitor this brain signals and warn of an impending seizure so obviously you cannot perform surgery on the on the patient every 5 years or other reason than to change the battery in the device you would like a device you don't have to change the battery of the device is working so energy efficient they can get all of their energy from harvesting the patient's head movements or body movements Etc and would not require batteries 3:07 3:50 Finding list of the transistor the transistor is a power-hungry device so much so it's becoming very difficult to continue down scaling transistors sizes and making electronics progress beyond past where it is today. That's a major problem that is facing the electronics Industry and mostly comes from the fact the transistors consume a lot of power when they switch in a circuit and so replace the transistor with these tiny magnets and such would do the same thing as switch between two states in the case of transistor the two conductance states, two charge states if you will, and in our case it's switching between two magnetization states or two spin states. They are equivalent, they do exactly the same function only in the case of the magnets the energy required is two orders of magnitude less than what is required for the transistor so that's what the advantage of magnets come in. 5:00 It used to be we couldn't make very good magnets we couldn't make it to the Precision you today or make these devices a reality but today with advances in fabrication technology materials technologies et cetera we can make them here in our labs, so these things are now a reality and obviously the transistor industry has been around for 60 years and several billions of dollars have been poured into it, so in terms of fabrication, packaging cetera they are far ahead of us but that's because we are the new kid on the Block. We haven't been around that long as transistors and we certainly haven't had the kind of investment the transistor industry has had but I guess, with time we will be able to do things attract the attention of industry and other

2 powers that be and other people are interested new technology it is nourished, it progresses and I think they were hoping for that. That it will come true, and we mightn be able to take this to fruition and it does become a technology in certain types of devices. The sizes are very comparable. The magnets might be a little bigger actually, and consume more area, but the energy saving is so much, that is a small price to pay. 6:50 Your laptop for example, the battery is a fair fraction of the weight of the laptop. So if you can eliminate the battery theyre, you ll not only reduce the weight of the laptop, but there are other advantages. If you put your laptop down on your lap, and work on it for a couple hours, it will feel hot. The laptop becomes hot and you ll actually feel it. It s happening because you re consuming and dispiating too much energy. It has several benifits, not just in terms of reducing weight and making the devices more robust, but also in terms of dissipating less, and there are other issues that come itno play. One of the major issues that comes into play is that if you have too much heating, you cannot get rid of it. You cannot pack too many devices on a chip. One of the major issues in the electronics industry today, between when the transistor was invented in 1948 and today, 1987 a long time ago, one of my colleagues bought a one gigabyte memory. The disk was as big as my computer screen. Today you can get a 10 gig hard drive, a small little thing, and it s a flash drive. If the car industry had progressed as fast as the car industry, today you d be able to buy a car for 20$ or so. So that s something that the electronics industry is very proud of, but that cannot be sustained forever. One of the problems that is faced by the electronics industry is that the transistors are dissipating so much energy that you cannot pack more and more of them on the same chip anymore, and if you do that the heat dissipation will be so much that the chip will melt, essentially. 10:30 THE HOW 11:00 The way you do this is you need a pizoelectrtic substrate like lead zirconate titatanate, ( bismouth ferrite ( there are a bunch of pizoelectric substrates. Our favorite is one called PMN PT (Lead magnesium niobate-lead titanate, no wikipeadia article exists) What happens in a piezoelectric, is that if you apply an electrical voltage to the piezoelectric, the piezoelectric either expands or shrinks. So if you have a magnet sitting on top of a piezoelectric layer, when the piezeoelctric layer expands or contracts, it makes the magnet either expand or contract with it. So that generates strain within the magnet. there are certain types of magnets, known as magnetostrictive magnets, examples of that would be simple magnets such as cobalt nickel, or gallium iron galfinol, we call it. Those magnets are magnetostrictive, meaning if you apply strain on them, their magnitization will rotate. This is known as the Villari effect, or inverse magnetostrictive effect ( That has been known for a very long time. What was not known, because nobody had ever looked into it carefully, is that if you make the magnets magnitization rotate in this fashion, the energy you dissipate in the

3 magnet itself, plus the energy that you will need to make the piezoelectric expand and contract, the sum total of that energy is miniscule. That realization was very critical for us. We realized that in 2010, when we published a paper pointing this out, and we did a lot of work on it, a lot of simulations and now also experiments to demonstrate that. And it turns out that the amount of energy spent is an order of magnitude less than in the best transistors, and may even be two orders if we are able to optimize our substrate and so on. So there is an energy saving. 13:25 That energy saving has been many benefits one of the primary benefits is that now you can pack more devices on a chip if you pack try to pack too many of them on that ship in the same area H1 it dissipates a lot of energy and so that adds up the energy density per unit area becomes unbearable if these things can be made more energy efficient so they disappear less energy than the energy density will go up tremendously so that you can pack more devices onto a chip and make your computer's more powerful your signal power and so on and so forth because they have more devices working in unison to make your processor work well the chips are the size of the ship depends on how many processing units you have on it so better metric would be how small is the most primitive processing unit and the most primitive Processing Unit in traditional Electronics the transistor what you today can be on the order of 10 nanometers. 10 nanometers is roughly one ten-thousandth of a human hair. Our magnets are not that small are magnets are 5 times larger than that so we have a 5 times larger area but what happens when a chip it's not just the unit's area but how closely you can space them so I can have very small devices but the spacing between successive devices are neighboring devices is 10 times larger than the device size itself then I can see actually goes down and that's the problem with transistors the transistor itself is smaller than the magnets but the spacing between them to adjust them laugh be much much larger because the transistors generate more heat you cannot put them close together like magnets very much less heat so you can put them closer together. So one of the major challenges were trying to work on it since this is a new transitional technology the materials aspect of it is not as advanced as it is with transistors what happens is we have a problem sometimes with reproducibility so if we make 10 devices for example 5 of them work very well 5 of them dont so the question is why isn't the 5 it don't work work why do they not work and the answer is that we didn't make them very well we have to have immense control over our fabrication process and we don't have that that doesn't mean nobody else can we are a university I have an academic lab it's not an industry lab like Intel or IBM who have much more resources than I have and certainly much more expertise in fabrication and processing than we can have ina university. So when the constraints of an academic lab we can do the best we can and we have a yield issue which means that not every device we make actually works and even those that work 5 don't work 5 work and among the five the work they don't all work exactly the same there are slight differences between them so there's a variability issue and there was a real issue so these are the issues through these are sort of like the teething problems of a new technology. Not even Intel Obviously if intel were to pick this thing up an investor and how our Manpower resources and expertise this would progress very rapidly but until that happens it's still in the university and we're not the only people working on this we have collaborators from a bunch of other places on Platt Library University of Minnesota UCLA Bakersfield on this is well so it's very much in the academic lab it's still research in some ways it's not something that you can immediately get off the shelf and make money it's not going to be it's not going to be selling in the next 5 years it's a little futuristic it that so that's how most technology start transistor started you know transistors were discovered in 1948 if they

4 didn't really make them into computers until the 1960s even when the transistors Rivers cover people are still working with 3 vacuum tubes because the technology of vacuum tubes was so well-established the transistors even though they were infinitely better we're not in a position to displace vacuum tubes immediately so that's the problem with technology takes a problem for a new technology to catch up and catch on and ultimately displace the existing technology. We called multiferroic that is not a word that we going we going to the word straintronics that is it's an analog of electronics and that analogy is what We have kept in mind when we came up with this thing so strange products you can go and Google the word you will find it today refers to changing the magnetic state of a tiny Nano magnet with Mac with mechanical strain that word was going here at VCU and our group so we came up with that name and so that's what a lot of people refer to it as strange contacts and a lot of reasons people aren't you interested in this is because of the low energy aspect of it. this may not be very clear when you have a very energy efficient device you also Drive the cost down not the battery cost that's different matter altogether but because you're now putting more more devices and a ship simply because you can afford to you don't have to worry about heat heading heating it up too much and making it known that reduces the cost and so the cost per unit function comes down. 23:00 You turn a Let me explain what that means you turn a transistor on sleep with the charges in the transistor it turns on now if you turn the power off and go away and come back the charges as are all gone so the device is turned off so you cannot keep the device turned on without keeping the power supply on that kind of device called a volatile device the information is available it goes away magnets are different you switch a magnet and keep it in a specific magnetization State go away come back 5 centuries later it's still there the magnet doesn't leak the information doesn't go away so that's what device called a non-volatile because the information stays there if you can have that property you can do some wonderful things this individual the University at the University of massachusetts-amherst is working with us on this thing wouldn't think he's come up with as a circuit architecture that what that will do it will eliminate the boot delay time and a computer if you switch on your computer right now you'll have to wait about a minute for the computer to boot it doesn't take that long is because I have inside my computer processing unit with the programs for the processing unit are not stored in the processing unit they're stored in remove memory and so the processing unit has to fax that information from memory back and forth back and forth and that takes a lot of time so that's why it takes 20 minutes for computer to boot up because you're running the program and programme asked affects the memory what this technology does is is it stores the data the program in the processing unit itself because the processing unit is non-volatile there is no difference between the memory and the processor you don't have to set your pro instruction sets from memory and you've eliminated the boot delay and your computer will turn on just like a TV you turn your TV on and it was to the program and I was move it will happen with the computer there will be no boot delay that may not be a big deal for Consumer use applications you can certainly wait one minute but in certain types of critical applications medical applications defense applications that 1 minute delay may be critical That is another area where we work with this and as Professor from um and Amherst I'm here now we're going to work with another professor at the University of Chicago Illinois very young fellow he was actually very young at undergraduate intern at VCU worked in my lab got his PhD at Georgia Tech and is nowa Professo atr Chicago. So what he does is he comes up with circuits that do specific functions such as image recognition and does a lot of image processing

5 a lot of neural networking compuatiations, but not adding numbers together that's a different type of computation. He's been looking into that, and using our devices as opposed to transistors and got 250 times Improvement in terms of the space requirement, in terms of the space or power required and in terms of the number of computational steps required the number of steps is reduced using the types of architectures he's been very clever he's taking the very specific device characteristics of these strings on it yet stringtronic devices and use them to replace the transistor. He replaces the circuits with these new devices the new circuts performed about 250 times better. These things do not have any special requirements you do not need to call them they operate at room temperature if they're fairly robust, no less robust than transistors so in terms of how robust they are how reliable they are reliability is a complex business I don't want to get into that there are certain areas where this is not very reliable just why why we don't want to do things like Boolean logic just letting those computers do what we want to do do other types of computation there different types of of computational algorithms are used which are more forgiving of imperfection so we have an issue with imperfection that is mostly because of the new technology but there's certain applications that are very forgiving of those imperfections, that s where we focus because that's where we can defeat traditional Electronics. So they have to have basic characteristics are in basic chemistry to come into this program the math science Innovation Center is not a minority oriented it is a it takes high school students from all schools in the Richmond area and they come and spend time in our lives those kids we keep mostly these down a magnet based Computing techniques