Jelena Vuckovic: Photonics — a light on the computing horizon
Experts estimate that computers gobble up as much as 10% of global electricity.
They predict that that share will only grow as data centers expand and the internet of things brings scads of new computer-controlled devices to the world.
Jelena Vuckovic is an electrical engineer who sees a light on the horizon — quite literally. She is building computers that calculate and communicate more with photons than electrons. These “photonic” devices could cut energy consumption in half and empower exciting new technologies, like quantum computing, in the process.
Before that day can come, however, Vuckovic and other proponents of photonics — the science of using light for practical purposes — will have to design smaller devices and improve manufacturing techniques to compete with today’s ultra-small electronics. To hasten that arrival, Vuckovic is turning to artificial intelligence to develop new device designs and new materials that could usher in the age of photonics.
Join Russ Altman and engineer Jelena Vuckovic for a discussion of the power and promise of photonics on The Future of Everything.
Transcript
Russ Altman: Today on “The Future of Everything,” the future of photonics. What is photonics? The Internet, the source of all knowledge, says that photonics is the study and use of photons, or light particles, for practical purposes. More specifically, how can we use the physics of light to emit, transmit, modulate, switch, amplify, and sense things? Okay, we’ll see how that does.
Currently, most of our high-tech devices are based on electronics, the manipulation of electrons on chips, circuits for practical purposes. The last 50 or more years has seen rapid improvement of our ability to design electronic components that have smaller, more energy-efficient, and they’re very useful, and the full variety of devices that we know and use every day, and many devices that we don’t know about, but underlie an entire electronic world of communications computing, but light and photons might be the next big thing. Instead of moving electrons around, we can move photons around. They are more energy-efficient, they require less energy, and they provide a wider spectrum of frequencies that we can use for various applications, but they have been harder to work with, and the devices that manipulate them haven’t yet gone through the phase of extreme miniaturization that we saw for electronics.
Some examples of current photonics victories are fiber-optic cables that deliver communications. You may have heard of LIDAR: light detection and ranging, which is used on many of the new cars that have self-driving or assisted-driving capabilities and is the answer to RADAR which stands for radio detection and ranging, so maybe we’re moving from radio detection to light detection. There’s new types of imaging of the insides of solids, similar to ultrasound, but using light. Photonics is critical for quantum computing, a new type of computing, still fledgling, that promises however to provide a huge leap in computational capability when it becomes routine.
Dr. Jelena Vuckovic is a professor of electrical engineering and applied physics at Stanford University. Jelena, what got you interested in photonics, and how certain are we that we can usher in this period of photonics renaissance?
Jelena Vuckovic: Well I got interested in photonics in grad school. I came to grad school thinking I would be doing information theory, and then met my former PhD advisor who showed me some things that they were doing building crystals for light to manipulate the flow of light and localize light into very, very small volumes below few hundreds of nanometers.
Russ Altman: So that’s very small.
Jelena Vuckovic: Very small. Micrometer is a millionth part of a meter, so nanometer is a billionth part of a meter, and I was fascinated by the idea that you can really engineer and make something, these really sculptures on a nano-scale that could be really manipulating light in completely unexpected ways, and at that time, the applications of that were not completely clear. Of course, there were fiber-optic communications, but the applications of those crystals for light, photonic crystals, and some of these newer structures that people are designing were not clear, but I thought I would just be interested in learning more about it for the next five years of grad school. By the time I graduated, it became more interesting to also use it in the applications.
Russ Altman: Yes, and so you mentioned fiber-optics, and this is a little bit of a mixed bag because it is a huge win for photonics, but I think my sense of the reading is that in some ways, fiber-optics are really only a fraction of what they could be. Could you explain why? Why should I be slightly disappointed about my fiber-optics?
Jelena Vuckovic: Well I mean, optical fibers have been around for a really long time, and so did a lot of other optical devices including a variety of lasers and light-emitting diodes, and in principle, that is photonics and optical electronics, but what we’re going through right now and what we’ve been trying to do for the past 20 years or so is the idea of integrating a lot of optical devices on the same chip, so it’s not really new to build a new laser, single laser in a shape or an LED or improve its efficiency.
What is new is this effort of integrating a lot of things, millions of components on a chip that would lead to new functionalities, and that will be really analogous to what we’re doing with electronics where in 20th century, we went through this great revolution where from computers that occupied the whole buildings. Now we have much more powerful machines inside of our cell phones and mobile devices, and to build something like a compact LIDAR that you mentioned previously or augmented-reality glasses that would not really look ugly but would fit inside of your regular glasses that people would prefer to wear, you have to really build these optical chips that would integrate a lot of components and miniaturize photonic components inside of a very small footprint.
Russ Altman: So what I’m hearing is that we all hear about electronic chips and very large scale integrated chips, and we’ve gotten things very small. In fact I’ve heard that like sometimes, we’re pushing up against the limits of physics in terms of electrons move. So what I understand from you is that photonics has not gotten there yet.
Jelena Vuckovic: No.
Russ Altman: I’ve never seen a LIDAR device. Could you describe how big is it and how big could it be with success?
Jelena Vuckovic: Yeah, so many of you have seen self-driving cars from several companies around here being tested on the streets right now. LIDAR is this large thing sitting on top of the car which is.
Russ Altman: So it’s like the size of a toaster.
Jelena Vuckovic: The size of a toaster, exactly.
Russ Altman: So that is not small.
Jelena Vuckovic: Maybe not small. It’s also very expensive in tens of thousands of dollars. It’s not really visually appealing. People who care about the design of their cars would not like to have a toaster on top of their car. So in principle, there are no fundamental reasons why it has to be so big. It’s big because people just integrated off-the-shelf components, optics, mechanics, and that’s where they are now, but there, we could design things to be much smaller and in principle it could sit on a single chip that is smaller, that is half an inch in dimensions. That’s one of the things that we’re actually working on right now and others are working on. In that case, it would be pretty much like other sensors that you have in your car.
Russ Altman: Yeah, so tell me, what are the challenges? We’re so used to seeing things get smaller routinely, and not all of us have an appreciation for the underlying engineering advances that are required. What can’t we do with photons that we need to be able to do?
Jelena Vuckovic: Yeah, so there are multiple challenges. First, you cannot make things arbitrarily small in photonics because eventually, you hit something that’s called the fractional limit, so if you want to kind of localize things efficiently and not lose light, then you cannot really squeeze it into something that’s smaller than half of the wavelength. That would be something on the order of again, a few hundred nanometers in dimensions, so you can’t really go —
Russ Altman: Fundamental physics.
Jelena Vuckovic: There is a fundamental limit. I mean you can squeeze it into something smaller than that with metals for example instead of semiconductors, but then there would be losses, so there are issues there. So you can’t make things arbitrarily small, but having that said, traditional, state-of-the-art photonics is way bigger than that. For example in optical communications, people use devices that have dimensions of many tens of hundreds of micrometers, so it’s a little bit smaller than a millimeter.
Russ Altman: But something you could see with the naked eye.
Jelena Vuckovic: Yeah, almost right, or optical microscope, but on the other hand, for those same devices, we’ve shown and others have shown that they could be equally efficient, but way more compact. They could be something on the order of few micrometers, so you can kind of reduce their dimensions in all directions by hundred or thousand times, which means that you can integrate more efficiently. Then the other issue that we have with state-of-the-art photonics is that it’s very sensitive to the environment and to any errors in manufacturing, so if you for example make your optical chips and chip and you have electronics there and you put it in your car and the temperature of the environment changes.
Russ Altman: Vibrations?
Jelena Vuckovic: Humidity, vibrations, temperature, electronic heats up like all these things, your photonics gets completely messed up out of the operating points, so then you have to design a complicated system to tune it back to where it should be, and now what we’re doing is trying to make it really robust to errors and immune to those changes in the environment which is crucial because if people started with electronics building blocks that were so sensitive to any errors in the environment, we wouldn’t have billions of transistors right now in a processor. It would be impossible to integrate them.
Russ Altman: This is “The Future of Everything.” I’m speaking with Jelena Vuckovic about photonics, our current capabilities, and where we’re going.
Jelena Vuckovic: So we were just focusing a little bit on these really very practical challenges of environment, temperature, humidity, and also some of the physical limitations, but really, I wanna go back. I only mentioned a couple of applications. Tell me what the big promise is here. You had a choice of what to study, and you’re studying photonics. What are the visions of things that are not currently possible with electronics that photonics may allow to occur?
Jelena Vuckovic: Absolutely, so there are some really pressing issues that photonics will help address. For example, my colleague David Miller from electrical engineering a couple of years ago looked into the analysis of how much of the total electricity in the world we consume on information processing and computing, and the number is somewhere, conservatively somewhere around 10%.
Russ Altman: So of all the energy —
Jelena Vuckovic: Of all of the energy goes into computing which includes data centers. I mean a big fraction of that are data centers.
Russ Altman: So we hear about people doing Bitcoin mining. We hear about the data centers that Google and Facebook and Apple have. Is this the kind of thing?
Jelena Vuckovic: Yes, yes, exactly, and considering all of the expansion in the amount of information that we communicate and number of movies we upload on YouTube and so on, this is only growing, and if this current trend continues, pretty soon we will be consuming all of our energy on computing, on uploading cat videos.
Russ Altman: And then it will be hard to recharge our electric car.
Jelena Vuckovic: Yeah, exactly. Going back to that, the big fraction of that is just energy consumption in all of the connections and wires inside of this.
Russ Altman: So it’s not even productive. It’s not even the use of energy for productive, it’s just the dissipation.
Jelena Vuckovic: Sending, exactly, sending your electrons down the wires with very high speed, and that’s a pretty significant fraction of that number, and unfortunately, there is no really way to reduce that if you’re using just wires, copper wires, but you can reduce it dramatically or completely, almost completely eliminate it if instead of communicating by wires, sending electrons by wires, you are using photons to move information between different components of your computer or your data center.
Russ Altman: Are there estimates of how much energy we could save with even the partial deployment of these capabilities?
Jelena Vuckovic: Yes, so probably something on the order of at least 50% of this because it’s very hard to dig up precise numbers, but some estimates are that at least half of this is consumed in only wires. By replacing wires by fibers or optical connections, and of course by at the beginning of that, you have to transfer information from electrons to photons, and at the end from photons to electrons again. You’re not completely replacing everything.
Russ Altman: The computers might still —
Jelena Vuckovic: Computers are still —
Russ Altman: have electronics.
Jelena Vuckovic: Exactly, electrons, So we’re electronics is not going anywhere. You’re just kind of replacing connections there. You will save a significant amount of energy.
Russ Altman: And current fiber-optics can’t do this?
Jelena Vuckovic: No because there are, well people are already using fibers in data centers, but this is just to connect different servers inside of the data centers, and if you look at the pictures of Google data centers and other data centers, there are a lot of fibers running between different computers, but there are also wires inside of the computers connecting your processor to memory, or connecting different cores inside of the processor.
Russ Altman: And those are also wasting energy.
Jelena Vuckovic: Wasting energy. We all know that when you touch your laptop, it heats up. All of your electronic devices heat up, and that’s just energy dissipation, and that could be significantly reduced by photonics.
Russ Altman: This is “The Future of Everything.” I’m Russ Altman, and I’m speaking with Dr. Jelena Vuckovic about hot computers and how to cool them down using the miracle of photonics. So I know that you have recently written a bunch of very technical papers, but what the goal is actually prototyping some of these new miniaturized devices, and it sounds to me from looking at these papers that you’re bringing together all of the capabilities of modern science in terms of computation and modeling, deep learning even to some extent, and of course, your physical understanding of these sub-systems. So can you describe for us, and one particular paper which was just fascinating was putting light through diamonds. Why is that interesting, and what was learned from that experiment?
Jelena Vuckovic: So going back to the way we design and build photonics, which was the first part of your comment, I realized even in grad school that the way we designed these photonic devices is basically there is a finite library of things that you learn in grad school, and then you try to put them together and tweak them, and then you make your trip, and I was always wondering why we design things certain way, and it looked to me that in many of these cases, we just adopted some designs that people could make by hand many decades ago when they were building like microwave systems, not even photonics, and we just adopted them in photonics, and we just kept going with them.
Russ Altman: So it’s like the legacy of random stuff that people invented.
Jelena Vuckovic: Exactly, that you learn in grad school and keep doing. I mean it works okay, but then when you want to build very large systems, it doesn’t really work that well. It makes things quite complicated, and also components are very sensitive as we were talking about it. Overcoming environmental errors, and then they’re very big and so on, so I’ve been always obsessed with this idea of kind of finding the right solution to the problem. Why something has to be so big, or can it be more efficient, or can you design it for a particular function, and there are some new functions that we need to have, and of course, this was quite challenging, like 15, 20 years ago, but right now, we also have much better computing hardware with all the boom in AI techniques.
Russ Altman: All those hot computers are actually useful.
Jelena Vuckovic: Exactly, and so we started combining techniques of artificial intelligence, machine learning, and optimization. We designed photonic devices where you can really kind of explore all the possible solutions to find the right one, and what we realized is that there are so many solutions that are better than what we normally using in today’s systems that it’s quite remarkable. I mean, you can find dozens of solutions that are smaller by hundreds or thousands of times and equally efficient and robust to errors, which made us very optimistic about integration —
Russ Altman: And this is a computational design of a small component that you can then build and test to see if it does what you think.
Jelena Vuckovic: Absolutely, and then we spend a long time building prototypes at Stanford and measuring them and showing that they work. We spent almost the last three, four years working with major foundries showing that this kind of crazy-looking designs that look very unusual from the perspective of manufacturers of electronics and photonics are actually something that could be made in a foundry. We’ve shown that that’s possible which means that there are absolutely no barriers to building these systems right now with high-throughput manufacturing.
Russ Altman: So this is very helpful ‘cause now I see that you’re painting a future where first of all, we have more components that have been designed with these new technologies. The components maybe provide a greater set of choices for the engineer who’s trying to compose them together to build chips that do things, and the designs that you’re creating, and I saw this in the publications that you wrote, they’re kind of working in the first try, that the computer modeling really does work, and it doesn’t create something that when you actually plug it in, melts or just doesn’t work.
Jelena Vuckovic: Absolutely, yes.
Russ Altman: Okay, well this is “The Future of Everything.” I’m Russ Altman. More with Dr. Jelena Vuckovic about the future of photonics next on SiriusXM Insight 121.
Welcome back to “The Future of Everything.” I’m Russ Altman. I’m speaking with Dr. Jelena Vuckovic, and we had just mentioned the possibility of talking about diamonds. I love diamonds. You shot some light through some diamonds. Why did you do that? What did we learn?
Jelena Vuckovic: Yeah I mean diamonds are actually very interesting materials for photonics and also for quantum technologies, quantum computing, quantum communication, and we all know that you can buy diamond in different colors in jewelry. When you buy really pure diamond, that doesn’t really have any colors, so when you buy pink or blue diamond, it basically comes from defects in that diamond from presence of impurities inside of diamond. Those impurities are actually very interesting for building all these technologies that I mentioned, in particular for building quantum sensors or quantum computers or quantum communication devices. Of course, you don’t want to take a piece of diamond that’s already pink because that means that it has a lot of impurities that you can see with your naked eye. You actually wanna get it so that it has only some, like few of these impurities or maybe hundreds of them.
Russ Altman: Almost perfect
Jelena Vuckovic: Almost perfect,
Russ Altman: but not quite.
Jelena Vuckovic: But not quite. It would still have these impurities inside, and you can use these impurities as what we call quantum bits to kind of build all these technologies. You can use them to generate individual photons, and photons are individual particles of light. Light surprisingly consists of particles, and you cannot really divide that photon any further. That’s what Einstein’s Nobel Prize was for, photoelectric effect. That whole field of quantum mechanics pretty much started with those discoveries. You can actually use that single impurity to generate one photon, one particle of light, use it to communicate between two parties, and if you use that one photon to communicate, and someone tries to eavesdrop on you by measuring, by somehow observing that photon, you would know that someone was eavesdropping on you, so then you can make absolutely secure communication systems.
Russ Altman: That sounds extremely practical. So first of all, let me ask. The diamond, is it naturally-occurring diamond, or do they have to be grown specially?
Jelena Vuckovic: You can use either naturally-occurring diamonds, or most recently, almost all of us have been using synthetic diamond, ultra-high-purity diamond, and then we insert on purpose with a specific location these impurities.
Russ Altman: And then you know where to shine your light so to speak. Then you generate these signals from the impurity, but now you said something very practical which is under normal situations, we can have eavesdropping. Not with this technology, but with fiber-optics or electronic systems, you can put some sort of device around the wire or around the fiber-optic, and it can intercept the signal, and it’s not detectable, but what you just said if I heard correctly is that in this case, if they try to do the equivalent of eavesdropping into the system, they won’t be able to, so tell me a little bit about that.
Jelena Vuckovic: So in today’s fiber-optic communications, you send optical pulses, laser pulses, and that’s how you communicate between two parties, but someone can always pick up a little bit of your signal and essentially —
Russ Altman: Siphon it off and use it for nefarious purposes.
Jelena Vuckovic: And eavesdrop on you, right, but if you send information in let’s say a single photon or some other quantum state of light, and that could be what we call entangle photons or something else, but let’s say, let’s talk about single photon, and someone tries to eavesdrop on you, they can’t really pick up a little bit of a single photon. That’s impossible because that’s an indivisible particle of light. Nobody can cut half of it which means that if they try to tamper with your information to measure the state of that photon and pick up information, they’ll perturb it. Quantum mechanics —
Russ Altman: And you can detect this.
Jelena Vuckovic: And you can detect some interference on the channel.
Russ Altman: There’s a photon missing.
Jelena Vuckovic: Exactly. Exactly because you can kind of check some number of the photons you sent and see if they’re messed up. Oh yeah, then someone was eavesdropping on me, and then you stop and repeat communication.
Russ Altman: So this could be the basis, if scaled up, and as we discussed in the prior segment, if you can create all the components to manipulate this on a chip together, this could be the next generation of ultra-high-security optical transmissions.
Jelena Vuckovic: And there are systems. I mean, you can buy commercial systems for what is called quantum communication. This is quantum communication, but they’re limited to short distances, maybe something on the order of a few tens of kilometers, but if you would like to send your signal from east to west cost or from U.S. to Europe, then you cannot use these secure systems, but instead you need to build something called quantum repeater and you have to put this special —
Russ Altman: It’s like an amplifier.
Jelena Vuckovic: Exactly, but you can’t really amplify quantum signals.
Russ Altman: That was your whole point.
Jelena Vuckovic: Exactly, so you have to kind of build this, distribute what we call quantum mechanical entanglement which is basically the main resource for building quantum computing, quantum communication, everything, quantum sensors. You have to distribute this entanglement from east to west coast, and then, you can securely communicate over long distances.
Russ Altman: This is “The Future of Everything.” I’m Russ Altman. I’m speaking with Dr. Jelena Vuckovic, and now we’re talking about the quantum stuff, and a lot of people, they get fuzzy-eyed when they hear about quantum. So we’ve just talked a little bit about quantum communications and that some of these technologies are going to lead to extremely long-distance and secure communication channels. The other thing we hear a lot about is quantum computing, so before we get into how it might work, can you tell me what is the place and the role it might play? Some people wonder is it gonna replace our current supercomputers? Is there gonna be a quantum computer in my cell phone, or is that the wrong vision of the future?
Jelena Vuckovic: That’s completely wrong vision of the future because classical or traditional computing that we have is not going anywhere. Quantum computing really is much better and much faster than traditional computing only for a very small number of problems, and the problems were where it outperforms by many, many orders of magnitude to classical computing are factoring of large numbers, which is the basis of today’s cryptography.
Russ Altman: So once again, it’s going toward security again.
Jelena Vuckovic: Security yes, that one is related to security, then the other problem. There are few other problems including searches of unsorted databases, for example you know a phone number, but you don’t know who it belongs to, and you have a phone book, so how do you find who that phone number belongs to without reading all of the entries in a book, and some problems which are more physics-oriented like simulating other materials, discovering new materials, discovering new chemicals. So those are the problems where it would be outperforming traditional computing, but not like speeding up your Microsoft Word.
Russ Altman: Okay, so it won’t be an ultra-fast, although it would be interesting to see if it would help with PhotoShop, but let’s not go there, but let me ask you. Before we got on the air, you were saying that one interesting analogy is that if we think about current computers getting an input, doing a computation, and giving an output, that the fundamental approach for quantum computing is very different, and how did you characterize that?
Jelena Vuckovic: The basis of that is everybody knows in today’s computers we have bits which are 0 and 1. Your switch is in one of the two possible positions. That’s how all of the modern computing works. With quantum computing, you have quantum bits where your switch can be in an arbitrary position of zero or one. It’s not just zero or one, but it can be anywhere in between. That actually allows you to run some specific algorithms by kind of putting everything in a superposition of different possible inputs and different possible outputs.
Russ Altman: All at once.
Jelena Vuckovic: All at once. Right.
Russ Altman: And it can do the computation kind of in parallel on all potential inputs.
Jelena Vuckovic: Exactly. And then when you perform your measurements on the system it would collapse into a certain answer with certain probability.
Russ Altman: Actually that helps because you were talking about looking for a phone number in the phone book. I know how to look at one entry and compare it to my phone number and say yes or no. The problem is, in New York City, there’s four million entries, but if you can use these q-bits to look at four million numbers, then your answer — I’m snapping my finger — comes in very quickly.
Jelena Vuckovic: Exactly.
Russ Altman: So, where are we with quantum computing? Is this a pipe dream, or are we starting to see them deployed and built?
Jelena Vuckovic: There are machines that major companies such as IBM, Google and some startups, Rigetti for example, are building, and they are based on so-called superconducting quantum bits. That’s currently —
Russ Altman: They’re not using your diamond bits.
Jelena Vuckovic: Not yet, but eventually they will have to use them to connect into quantum internet or to scale them further. But right now, they have about 50 quantum bits. That’s roughly at the border where people are expecting, maybe we can start analyzing some problems that would maybe not be classically simulated. People are talking about quantum advantage. Maybe we’ll start seeing these computers really overcoming the power of classical computers, but it hasn’t happened yet.
Russ Altman: So we’re right on the cusp.
Jelena Vuckovic: Right on the cusp, but not really at the stage where these quantum computers are really convincingly more powerful than classical computers. That’s the current platform. You can view it as an analog of vacuum tubes in the 40s and 50s. Probably will be scaled to something like hundreds or thousands of quantum bits, but for a truly functional quantum computer that would be really useful for factoring large numbers and breaking cryptographic codes, we will need about one million physical q-bits. The question is how we get there. It’s quite conceivable that we will need to change the architecture and move from superconducting q-bits to semiconductors — impurities in diamonds, or some other material. A lot of people are working on these alternative platforms which would be equivalent to silicon transistors as opposed to vacuum tubes.
Russ Altman: So very early days with a lot of promise. Well thank you for listening to “The Future of Everything.” I’m Russ Altman. If you missed any of this episode, listen any time on-demand with the SiriusXM app.
