Emerging Technologies Episode 1: Quantum Information Science

(Intro Music Starts) SH: Hello everyone, I am Sam Hansen SW: And I’m Sadie Witkowski. SH: And you are listening to Carry the Two, a podcast from the Institute for Mathematical and Statistical Innovation aka IMSI. SW: The podcast where Sam and I talk about the real world applications of mathematical and statistical research. SH: Sadie it is wonderful to see you and to be recording with you again! It has been too long! SW: It absolutely has, I am so glad to be back. Though while it is good to see you, I doubt you had me come down here just to exchange some pleasantries. SH: Too true, this is Carry the Two after all, not the Small Talk Hour SW: Exactly, and as it is Carry the Two what application of mathematical and statistical research do you want to share with me this time? SH: Well in this season of Carry the Two we are going to be examining how math and stats is helping scientists, engineers, and industry develop new and emerging technologies. SW: Oooo, what sort of technologies are we talking about? Are we about to go Iron Man on our listeners? SH: Not exactly, especially as these technologies aren’t fiction, they are real. I am talking about everything from Materials Development to Computational Medicine to Fusion SW: Amazing, and will we be focusing on any of those today? SH: Nope, today today we are going to be focusing on Quantum Computing and Information. SW: Ah, quantum computers feel like they have been emerging as a technology for quite a while now SH: That they do, and for good reason. It was back in the 80s, when the theoretical underpinnings of quantum computing was first laid out by luminaries including Paul Beniopff, Yuri Manin, and Richard Feynman and then through the 80s and 90s yet more names like David Deustch, Peter Shor, and Michele Mosca developed mathematical algorithms that could leverage quantum computer architectures to work faster than any traditional computers SW: The 80s, we have been talking about quantum computers for 40 years now? And it is *still* an emerging technology? SH: Ah, I am going to have to take a second here and put together my brain that the 80s were 40 years ago now, but, to be fair Ada Lovelace and Charles Babbage started laying out the theoretical underpinnings of programmable computation in the 1820s, then Turing put forth his formal mathematical model for computing in the 1930s, and finally in the 1940s machines that we would recognize as the progenitors or our modern computers started to show up. SW: Not to mention how Babbage and Lovelace were influenced by mechanical calculators for arithmetic that were first created in the 1600s and how those calculators were extensions of the abacus which has been around since, ummm, I think around the 3rd century BCE. Ok, so I guess 40 years really isn’t that bad SH: And it is not like there are no quantum computers out there at all. In fact the first proof of concept came in 1998 and there are a bunch of companies out there that are currently working in the quantum computing space. Some you have heard of like IBM and Google, and others that are not yet household names like D-Wave and Xanadu. SW: I guess my question then is, if we know how to build quantum computers and we already know of algorithms to use with them and mathematics that they can speed up, what exactly is the hold up? SH: There are many different answers to this question from how expensive and hard it is to construct the physical systems, called qubits, that are the heart of quantum computing to the need to continue to develop the theoretical and mathematical underpinnings of the field. But those are not what I want to start speaking with you about today (Silence for a couple beats) SW: Well… are you going to keep me waiting to find out what it is you want to talk about? SH: Sorry, I guess I am a little out of practice. How about I let Ben Brown, a researcher at IBM Quantum, tell us instead? BB: We have to build a quantum computer out of something, and lots of people have lots of different ideas about that. I work for IBM and there we're trying to make a quantum computer out of superconducting qubits, but other people have ideas on using, say, trapped ions, for example, or photons to encode qubits, which are processing the quantum information of a quantum computer. But well, to all of the approaches is that the qubits that we can make in the lab, they're never perfect. And sometimes you set them in some state and through no fault of your own, the environment or just some miscalibration or something will flip that qubit to the wrong state. Errors are a thing that we can't avoid, and we need a way to deal with that problem. SW: So what? Don’t classical computers also have issues with errors? SH: They do, but the way classical computers are built actually helps ameliorate them BB: By storing a bit in the direction a magnetic field is pointing, you've asked a whole Avogadro's number many of atoms to all agree to point in the same direction, or all up or all down. And if some of those bits try to flip, well, you know, you can think of these atoms as tiny little magnets that suddenly get pulled back to align with the other magnets. So they just sort of fix themselves. SH: Where as the way quantum computers are built may actually make errors harder to handle BB: You know, c lassically, we just have bits that flip up and down. But when we talk about qubits, qubits are, you know, this is a quantum way of storing information. And you can think of them as a bit where you can have a zero or one state. But you can also make superpositions of the states. This is something that classical computers don't have, which is partly why quantum computers are different. And we exploit the fact that they can make superpositions of states to do calculations. So, qubits, because they're bits that can also live in a superposition, they can flip like the classical bits of a regular computer, but there are other errors that can occur as well. SH: These errors can lead to losing information in the quantum computing system. Or instead of thinking of losing information you can think of it as introducing noise YQ: Okay, so when I talk about quantum noise, it usually refers to things like not being able to carry out a quantum gate perfectly. SW: Oh, a new voice! Whose is it? YQ: So my name is Yihui Quek. I am a postdoc at MIT and I am an incoming assistant professor at EPFL, Ecole Polytechnic Federal in Lausanne, in Switzerland. SH: Yihui shared with me the two different types of noise that appear in quantum computing systems. The first is called T1 or amplitude dampening noise YQ: For amplitude damping noise, let's say that you are computing with some atoms with two energy levels. So T1 noise is noise that kind of encourages an atom to go from an excited energy level to its ground state. Of course, there's also transitions from the ground state to the excited state, but they don't happen as often as in the other direction. So that's T1 type of noise. SW: Ok, so like going from a 1 to a 0 in a classical computer? SH: Exactly, and there is a T2 type of noise as well YQ: For T2 type of noise, that's kind of similar to what theorists understand to be depolarizing noise. So, yeah, that's just when your atom is kind of like becoming more classical, let's say. SW: And what does it mean for the qubits to start acting classically? Is the qubit playing Chopin? SH: Well in this case it means that instead of maintaining their superposition of different probabilities of 0 or 1 they are replaced by what is known as the maximally mixed state. The maximally mixed state is kind of like the quantum analog of uniform probability distributions. This means that the probability of every possible state would be equally likely. And so when we talk about this noise replacing a qubit with the maximally mixed state, that's bad because the maximally mixed state is non-quantum and doesn't contain enough information for you to do any sort of computation. So the fact that this noise has some probability of replacing its input qubit with this rather useless object means that the noise is kind of like decreasing the amount of information in whatever it sees. SW: Ok, yeah decreasing the information to a state where it is not possible to compute is clearly a big problem. Now I know that you wouldn’t have started us down this road if there wasn’t any way of dealing with these errors and noise SH: You know what, you should be a podcaster with that sort of insight. And of course, you are right. YQ: Error correction and error mitigation are two different ways of kind of correcting for the effects of the noise. SW: Great, but what do they mean? SH: Well let’s talk about error correction first. YQ: With quantum error correction, you are constantly monitoring your circuit to see if any errors have arisen. And if any errors have arisen, you measure it, you realize that there's an error, and then you do an operation that corrects the error. SH: Error correction has been an important part of computing and communications for a long time and mathematicians have created many tools and algorithms that can help identify and correct errors for traditional systems. The first error correction code was developed by Richard Hamming in 1950 and used 7 total bits where 4 bits contained the message and 3 were there to help detect and correct the errors. Similarly to this groundbreaking work by Hamming, Quantum Error correction also relies on multiple qubits working together BB: The error correction is all about how we can take several of these qubits and make them work together in such a way that collectively they can protect one logical qubit of our quantum computer. SH: More specifically BB: We make what we call an error correcting code, which is where we use several of these slightly noisy qubits, and then we make, well, we call them stabilizer measurements or checks and these are quantum measurements we make on all of the physical qubits and these checks are supposed to always give us a certain outcome if we make this measurement they're always supposed to tell us one for example right but if we measure them and they give a different answer yeah we can detect an error we know we have to do something about it SW: Well that sounds great, so problem solved? SH: (laughs) you know it can’t be that easy or this would be a very short episode. SW: (laughs) Can’t blame me for trying to set you up SH: No I, no I can’t, and you have done a good job because this error correction isn’t perfect BB: You know the principle behind error correction that works that if we spend more and more on physical qubits The failure rate should go down. The failure rate is the likelihood that you know the error correcting code doesn't protect the logical information SH: Which sounds great, except that qubits are very expensive to produce. Which means that Ben, and many many other quantum computer scientists, are looking for something very specific BB: We're trying to find the best possible code With the smallest possible failure rate with you know that uses the least number of qubits SW: I don’t know if many people would be happy with a failure rate above 0 for most of their computational needs SH: I know, which is why it is good that the quantum computer theorists have a theorem to fall back on BB: We have a celebrated theorem. We always write it in papers that it's a celebrated theorem that we call the threshold theorem. And that tells us that, well, it tells us that quantum computing is scalable at all with noisy qubits. Like, you know, we assume... We assume that the qubits are definitely noisy, and we assume some things about how the environment will act on those qubits. Maybe the noise acts locally. It doesn't correlate and act on two qubits that are far away from each other. Otherwise, we're in a real mess. But assuming certain sensible things like this and that our qubits are fairly good, so the noise on those isn't too bad. That it says is we can make the code bigger and bigger and bigger, and the failure rate will get smaller and smaller and smaller until essentially you have a perfect logical qubit if you're willing to spend. Or, yeah, if you're willing to buy enough qubits, you can make a logical qubit as good as you like. SW: And what is this threshold that the theorem is named after SH: I asked Ben that same question BB: The threshold is a number. So how noisy can the qubits be such that this scaling is still working well? So, yeah, if you're below this threshold rate of error, then your qubits are good enough that you can make your code bigger and bigger and bigger, and the logical error rate will go down really, really fast, and you'll have a nearly perfect qubit with a sensible number of physical qubits. SW: What about the qubits being made today, how does their noise compare to the thresholds being aimed for? SH: It turns out that they are getting very good BB: And so it's an exciting time to be in the field because now we're really seeing qubits are below this threshold seeing what we call break-even results, where the qubits that we produce on these quantum error correcting codes are better than the qubits that we use to make error correcting codes. SW: But so far we have only talked about how to detect errors, what does Ben do after he knows there is an error? SH: Hold your horses, Ben was just about to get there BB: Well, you know, maybe the standard option these days is to say we're going to correct the error. We're going to look at the checks that went wrong and we're going to make an educated guess on what's gone wrong and we'll try and fix it. But another thing we can do is we just say, well, we detected an error. Let's throw it out and start again SH: And in order to make this decision they use what is known as a decoding algorithm, which is a classical computing algorithm which sits on top of the quantum computer BB: So what's happening is down here we have the qubits and they're nice and cold and we're doing the check measurements here to try and spot the errors. We take that information and we send it to the classical computer, which is running the decoding algorithm. And the decoding algorithm looks at what checks have flashed and what checks have not flashed and given a no error result. And then it decides how to fix it. SH: Which gets us to work Ben did with Samuel Smith and Stephen Bartlett on how tolerant these decoders should be with these errors SW: Wait, what does tolerance mean in this case? SH: Well here is an extreme example Ben shared with me that I found helpful when thinking about it BB: You could try something like this: Here's the most intolerant version of decoding, where the decoder, quantum chip looks at all the information that comes out. It says, did any of my checks register detecting an error? And it says, if any of them registered, we throw it out, we're done. But you don't always want to do that, especially because as your quantum computer is getting bigger and bigger and bigger, you have to make more and more and more checks. And then just, you know, it's not difficult maths, right? Like eventually the probability of seeing a check is going to go to one, right? That's the way it is. But, you know, it's still maybe a good idea not to try and correct everything. SW: And they built this idea of tolerance into their decoder? SH: They did, which means they can vary the probability of error that they are willing to risk. Which is important, as that maximally intolerant decoder has a big limitation BB: It's going to work really well, because we're going to throw out anything any time we might have seen an error, but we throw out a lot. And it might just take the age of the universe before you get the perfect sample where definitely no errors happen SW: Yeah, I guess never finishing any computation at all, a la the spinning loading wheel of doom, would be much worse than an error SH: For sure, and the decoder they built tries to ride that line between performance and errors BB: So what our decoder does is we interpolate this number between the completely tolerant case and the completely intolerant case saying, you know, we're willing to let the decoder try a bit. But if it's looking a little bit too close, then we're not going to keep it. And we can quantify that gap between how close we want it to be. And what we see is we can improve the error correction performance a bit by throwing out some samples, but not too many samples. SH: And it works BB: So the logical error rate improves on a fixed number of qubits. And as I said, qubits are expensive. And we're going to take a little bit longer, but we're not going to take the age of the universe. So this is a very useful tool for improving a quantum computer to make a decoder that's a little bit intolerant to bad cases. SW: Wow, that’s great. SH: It gets even better as they showed that with a well designed scheme for choosing whether or not to throw out potential errors, this choice is referred to as post-selection, there is another important threshold BB: What we found was, if you choose your post-selection scheme a bit more modestly, the way I was describing before, there comes a point... Below the error correction threshold, there's another threshold where your physical qubits, if your physical qubits are at least this good, your post-selection actually gets better and better and better as well. So as you make the code bigger and bigger and bigger, with a sensible post-selection scheme, the likelihood of your computation succeeding and not being thrown out by the intolerant decoder improves. So the time it takes goes down the bigger you make your quantum computer if you're using these post-selection tricks to improve your logical error rates. SW: You did bring up that there were two different approaches to handling the noisy qubits though, didn’t you? SH: I did, the second is called error mitigation and we will hear about it from Yihui soon, but I want to tell you about a really interesting error correction method that she shared with me first. It is called the Quantum Refrigerator Construction SQ: Oooooo YQ: So the quantum refrigerator construction actually goes back to this issue that we've been talking about for a while, which is the issue of doing intermediate measurements. With error correction, intermediate measurements are necessary because you need to know when an error has occurred and to correct it. But the quantum refrigerator construction is a way of doing error correction without intermediate measurements. SW: But why do we care about these intermediate measurements? SH: Because we are dealing with quantum states, and one thing that I definitely know about quantum mechanics is that in order to measure a quantum system you have to collapse wave functions. The wave function basically describes all the possible states of the system and the probability of the system being in any of those states. When one makes a measurement, the wave function collapses into one of its many states. SW: Of course, and this construction some how works around this? SH Precisely, but it requires a lot of qubits and that those qubits are noisy in the T1 or amplitude dampening, not T2, way SW: So how does it work? SH: The details are very technical, but it relies on research from Dorit Aharonov, Michael Ben-Or, Daniel Gottesman, and Avinatan Hassidim about error correction without collapsing quantum states and leveraging how amplitude dampening noise primarily takes qubits and flips them from 0 to 1 states YQ: Just by allowing the noise to idle on the qubits, eventually you're going, you’re going to find that noisy qubits are going to be converted to qubits that are very close to all zero states. But the all zero state is exactly what was needed to facilitate this coherent sort of error correction that doesn't require intermediate measurements. SH: Which really means YQ: So the quantum refrigerator construction combines these two ideas and says that if you have a certain kind of noise in your socket, then your noise could actually be helping you instead of, instead of obstructing you. SW: Oooo , that is so cool! SH: Maybe even as cool as your fridge! SW: Booooooo! Sam, come on SH: Sorry, sorry. Will introducing the idea of error mitigation help make up for that terrible joke? SW: It will be a start SH: Good, let’s hear more from Yihui then YQ: With quantum error mitigation, there's no constant monitoring of your circuit. What you do is you run the entire circuit or variant of the circuit that you hope to run. And then after the entire circuit run has completed, you look at the output of the circuit run and then you try and do some statistical inference process to figure out what exactly was the noise that occurred during the run of the circuit and then you try and correct that in classical post-processing. SW: Ah, another method that doesn’t require intermediate measurements SH: Exactly. In fact the error mitigation approach was developed due to the cost of both the constant monitoring and of the qubits required for quantum error correction. That is not to say that mitigation doesn’t have its own downsides. YQ: You pay for this convenience. And one of the ways you pay for it is that you have to run your circuit many, many times. And if in order to achieve the same level of correction, you need to run your circuit too many times. Then at some point, you don't really save too much by doing quantum error mitigation. because you're kind of like, yeah, paying for it in terms of how many times you need to employ a quantum device, and it could potentially just be a huge time sink. SH: In fact Yihui has shown in her research that there are some cases where the number of runs is incredibly unwieldy YQ: We constructed circuits that require a superpolynomial number of circuit runs in order to be error mitigated. SH: The specifics of what superpolynomial means is not important, just know that it means way more runs than would be in anyway practical. But, this specific case was constructed to be really hard and there are many cases where error mitigation is practical SW: But with all of these different possibilities how can quantum computer scientists know which way to go? SH: I'm sorry, but the answer really is, that it depends YQ: With quantum error mitigation and quantum error correction you can think of them as being on two ends of the spectrum where you can think of the scale as being like what amount of quantum control you need or quantum resources so quantum error correction is something that requires a large amount of quantum resources it requires intermediate measurements for one and with quantum error mitigation it's kind of a quantum budget friendly way of doing error correction SH: So if you have access to a lot of qubits, or a lot of money, then maybe error correction is the way to go but if you are more resource strapped maybe you will lean toward mitigation. Though some research is now showing that the two methods can be used in tandem as well. And while the specifics are not quite clear yet, Yihui does think that good progress is being made with errors in quantum computing YQ: I think right now we are still at the stage where we are kind of getting very close to good error correction and we should probably turn our attention to finding more and more algorithms for quantum computers. And once we have done that, then we can think about whether any of these algorithms can actually bring some commercial value. SH: And there is definitely one thing that quantum computer scientists have had, and continue to have on their side toward achieving these goals YQ: The amount of attention that quantum computing has received in the press is something that I'm very grateful for because I think a lot of this has also led to greater investment in quantum computing and that has led to the research that is making error correction progress so far today. SW: Press attention that we are contributing to right now! I guess we can now start calling ourselves quantum computer scientists too with all this help we have provided to progressing the field! I’ll add it to my linkedin right now! SH: You go right on ahead, while I definitely know more now than I did before starting reporting this episode I in no way am conversant enough to call myself a quantum anything. But maybe I will feel different after sharing a novel application of qubits SW: And when will you share that? (ad music starts) SH: After the break of course SW: Ha, of course SH: Have you ever wondered what goes on inside a black hole or why time only moves in one direction? Or what is really so weird about quantum mechanics? Well, you should listen to why this universe. On this podcast, you'll hear about the strangest and most interesting ideas in physics broken down by physicists Dan Hooper and Shama Waksman. If you want to learn about our universe from the quantum to the cosmic, you won't want to miss. Why this universe? Part of the University of Chicago Podcast Network. SH: Excited to hear about a very different application of qubits? SW: You know it SH: Well, before we do I need to ask you a question. Do you know what dark matter is? SW: Does anyone? SH: Right answer DT: Well, the short answer is that no one knows what dark matter is. We have overwhelming astrophysical and cosmological evidence for this mysterious fluid, this mysterious matter that permeates as far as we can tell, throughout the universe. SW: I do want to know whose voice that is, but first I want to ask about this overwhelming evidence, what is it? DT: W e can look at how galaxies spin, measure the speed of the stars orbiting the center of that galaxy, and say there's not enough, from all the stars that we can see, all the gas and the dust that we can measure with our optical instruments, we would not see enough mass to hold those stars on orbits in that galaxy. SW: Since our star and our galaxy don’t seem to be spinning apart that seems like rather great evidence. So who shared it? SH: That was Dylan Temples DT: I am a Lederman Postdoctoral Fellow at Fermi National Accelerator Lab, and I work on dark matter direct detection and kind of the intersection of dark matter direct detection and quantum information science. SW: Oh, I know Fermilab. They are one of the Department of Energy national labs and do a lot of particle physics right? SH: Right. They smash very small together with very large accelerators, they run the Deep Underground Neutrino Experiment, and they also are trying to detect dark matter with the help of people like Dylan SW: I know scientists love detecting things, but why dark matter? It seems to be doing its job and doesn’t want anything to do with us and the rest of the visible universe so why try to detect it? SH: I asked Dylan a similar question DT: One compelling in the sense of it is very important to the story of the cosmological evolution of the universe. How did we come to be here in the first place? SW: Wow - understanding our cosmic origins. That's pretty deep. But are there other questions motivating Dylan? DT: There are a number of, a number of properties of the dark matter that make it a challenge to detect in the lab. And so to get there, you need to develop new technologies and more powerful, more sensitive technologies to help you get to the sensitivities that you need to be able to potentially detect whatever the constituent particles of the dark matter may be. And so one of these areas is quantum sensing that is recently taken off to look for dark matter in the lab. SW: Ah, so while trying to figure out this big question from basic science they are also creating a lot of cool technology that could be useful in other areas? SH: Just like how NASA has helped contribute to things from vascular bypass surgeries to firefighting to cochlear implants to better mattresses. Basic science research has led to so many of the things we use in our daily life that we would be here for days listing them all SW: And I just don’t have that sort of time, so how about instead we talk about the specifics what it means to detect dark matter SH: Ok, that we can definitely do DT: What we want to do is, from a particle physics point of view, understand what are the particle constituents of that dark matter, what are the little pieces that make up these large-scale dynamics that we're seeing. SH: And the way they are trying to sense this dark matter is through the use of qubits, which of course begs the question DT: So why might you even consider using a qubit for detection of radiation or dark matter in our case? SW: That is a very good question SH: Yes it is, and to answer it we need to talk a little bit more about qubits. DT: Anything where you have two energy states separated by some amount of energy can behave like a qubit. SH: Which does offer quite a few different options DT: There are many things that can be used as qubits. So you can use atoms, you can use trapped ions all of these things that have different energy levels as a qubit. What our group focuses on are superconducting qubits. SH: These superconducting qubits are circuits that have been chilled to within 1% of a degree Celcius above absolute zero, so ridiculously cold, and then if you send in a small amount of microwave photons you can get the qubit to change from one state to the other. In other words it will exhibit quantum behavior. And once you have your qubits DT: If you can completely isolate your qubits, your two-level system, it will last indefinitely. It has no interaction with the outside world. And so you can put it in some state, and it'll just stay there. SH: And this is the start of the answer to why they are using qubits for dark matter detection. The rest of the answer starts with how it is impossible to fully isolate anything, and continues with the fact that DT: These qubits are very, very sensitive to interactions with their environment. SH: Which when combined with what we already know about how easily qubits are impacted by noise leads us to DT: If you have something that is sensitive to very, very small changes in the environment, you can potentially use that to detect very, very weak or very, very rare signals that might be interacting in your chip. SW: Wait, wait, wait. Are you saying that qubits being so easy to cause errors in, the exact thing that we spent the first half of this episode talking about, is the reason they are able to be used as hyper-sensitive detectors? SH: How about we ask Dylan? SH (From interview): So you're essentially using the fact that quantum computing is really, really hard to leverage the thing that quantum computing folks hate to do your work? DT: Yes, exactly. SW: I love this, I guess there really is a silver lining everywhere SH: Seems so. But Dylan doesn’t see these approaches as being in opposition DT: We want to understand the same things and then optimize in different directions. So we want to understand in our group, how does our qubit react to radiation? And taking that information for the quantum computing side, they want to engineer more robust, more robust qubits to ignore these effects, whereas we want to maximize the information exchange with the environment in certain ways that we can control. SW: So they have these qubits, how do they do about actually try to use them to find dark matter SH: Well first they have to try and isolate them as much as possible. In this case that means burying them deep underground to try and shield them from as many interactions as possible. Thankfully, while the earth will absorb the majority of the particles and radiation DT: The things that we care about, dark matter, doesn't see the rock as a big burden. It's equally as likely that a dark matter particle will interact in the rock as it is in our detector. SH: Then once the qubit detectors have been triggered scientists have to determine if what caused the interaction was dark matter or not. So they then compare the information from the qubits to models they have developed far what they expect non-dark matter interactions to look like and what they would expect dark matter interactions to look like DT: Okay, this is what my background, my standard model background looks like. This is what I expect my dark matter to look like. Then I do my experiment and I say, is my result statistically compatible with the background or is it statistically compatible with the background plus a little bit of dark matter? And so it's all about understanding on a large scale what signals you expect to see and what a dark matter signal should look like. SH: Which is all really fascinating, except there is a big problem SW: Uh oh. But I know you Sam, so I totally saw that one coming. SH: Ha, I guess I can be predictable. Which is not true for dark matter (SW Laughs), so, it makes sense that DT: We don't know the mass of an individual dark matter particle. SW: And let me guess, there are some possible masses for dark matter that qubits can’t detect? SH: Before we get to that we should probably dig a little deeper into how little we know about the mass of dark matter particles SW: I am scared now, but ok DT: So there are some very loose constraints on what that mass could be and that spans about 50 orders of magnitude SW: Oh my gosh, what! I was right to be scared! 50 orders of magnitude? I have no idea what that sort of mass range could possibly mean SH: Neither did I, so I had Dylan do so quick calculations DT: So, 50 orders of magnitude is quite hard to wrap your head around. But the ratio of the heaviest dark matter to the lightest dark matter is one with 50 zeros behind it. That is roughly the same ratio as the mass of the Milky Way galaxy compared to the mass of an E. coli bacterium. And so these are enormous scales. And so depending on where you're looking in that 50 orders of magnitude, you have very different phenomenology, different ways of that dark matter manifesting itself in the lab. And the technology that you need to look for dark matter that's very heavy is far and away different from the stuff that you need to look for the ultralight stuff. SH: If the dark matter is on the heavier side of things the qubit detectors have a chance of working, but for the ultralight side a different approach needs to be taken. One that relies on some pretty mind-blowing physics DT: It could induce time-dependent fluctuations in the fundamental constants of nature. So the constants of nature become no longer constant. you get a very, very small correction to the mass of the electron and to the fine structure constant. SW: Huh, so if dark matter is ultralight then it could cause things that we have always relied on to be unchanging everywhere, like the speed of light, to now be not constant?!?! SH: That’s right, but only for a short period of time and only locally. Thankfully what changes can be measured DT: If you have something that is very good at measuring the mass of the electron or the fine structure constant, you can start to be sensitive to these very, very small oscillating corrections. and that is where MAGIS comes into play for the search for dark matter. SH: MAGIS-100 otherwise known as the Matter-wave Atomic Gradiometer Interferometric Sensor is just such a measurement tool. And one that again leverages quantum mechanical behavior to do its sensing. They start by cooling down atoms a factor of a thousand more than the super-conducting qubits DT: This is a measure of kinetic energy and it's not really a macroscopic temperature. You can't put your finger on it and go, "Boy, that's cold." It just means I've sucked all the energy out of these atoms. SH: And at temperatures, and energy levels, like this, thanks to the quantum wave-particle duality, objects will start to display wave-like behavior. In particular they cool down Strontium 87, an atom that is popular for building highly precise atomic clocks. DT: And so it has these two clock states, the ground state and one of the excited states that we care about. And so we drive transitions between those things. And the interference pattern that you get out at the end is dependent on the energy difference between those two energy levels. That energy difference depends on the mass of the electron and the fine structure constants. SW: Ah, so if MAGIS sees a change in those constants… SH: You got it DT: The output of our interference measurement will change in time and it will oscillate at the same frequency as the dark matter. So that is one of the ways that you can look for a very specific type of dark matter or specific class of dark matter models with this matter wave interferometer. SW: And… SH: And what? SW: And does Dylan think one of these will actually work? SH: Uhhhh DT: Let's differentiate between work and find dark matter. So these experiments, they will work. SW: But working doesn’t mean detecting does it? SH: It does not, but that doesn’t mean nothing will be learned DT: So we are in this business of, at least in the dark matter sense, measuring zero better and better. SH: It is also worth remembering that these detectors will have an impact well beyond dark matter research DT: Qubits have been around for a while, but we're still far away from building a real quantum computer. And the more that we can understand about the noise sources and the backgrounds that influence the information loss in quantum bits helps us get towards building a realistic, large scale quantum computer. SW: Of course, and I bet that MAGIS will also lead to breakthroughs that industry will be salivating over, but what if they do really detect something SH: Dylan has some thoughts about that DT: What would it mean if we actually detected dark matter with one of these things? Well, someone gets a Nobel Prize. Probably not me because I just built the thing. It's probably whoever thought about the model that we discovered. SH: And no matter if they can conclusively say that something they detect is dark matter, any signal could be helpful when searching such a wide set of possibilities DT: Right now we're searching 50 orders of magnitude. Once we see something that might be a signal, we can start to focus effort on the technologies, new technologies, different technologies that would be sensitive to dark matter in that mass range, and really start to suss out all of the different expectations that we have for what dark matter should do, and see that we're observing that in different systems, different technologies. And I think that's really exciting. SW: Hahaha, yeah the analogy of a needle in a haystack doesn’t even come close to this challenge. Just being able to narrow it down to looking for something between huge ranges like Bacteria and Human or Solar System and Galaxy would be a huge leap forward (Outro Music Starts) SH: I mean I bet Dylan would like to get closer to say between cat and human but you aren’t wrong. Do you still feel like we should get to call ourselves quantum computer scientists? SW: You know what, maybe I did jump the gun earlier. Though I am so going to call myself a quantum podcaster now! SH: Oh, hell yeah. Me too! SH: Don’t forget to check out our show notes in the podcast description for more Ben, Yihui, and Dylan, including links to their work we discussed on this episode SW: And if you like the show, give us a review on apple podcast or spotify or wherever you listen. By rating and reviewing the show, you really help us spread the word about Carry the Two so that other listeners can discover us. SH: And for more on the math research being shared at IMSI, be sure to check us out online at our homepage: IMSI dot institute. We’re also on Bluesky at IMSI dot institute, as well as instagram at IMSI dot institute! That’s IMSI, spelled I M S I. SW: And do you have a burning math question? Maybe you have an idea for a story on how mathematics and statistics connect with the world around us. Send us an email with your idea! SH: You can send your feedback, ideas, and more to sam AT IMSI dot institute. That’s S A M at I M S I dot institute. SW: We’d also like to thank Blue Dot Sessions for the music we use in Carry the Two. SH: Lastly, Carry the Two is made possible by the Institute for Mathematical and Statistical Innovation, located on the gorgeous campus of the University of Chicago. We are supported by the National Science Foundation and the University of Chicago. SW: Where’s you chalkboard? Ok, I’m ready SH: Literally didn’t know that is possible SW: Laughs SW: Absurd, lies SH: Laughs SH: I can all the way down here and go real smooth SW: Laughs SW: Heh, heh, heh you’re failure is my joy SW: What is happening SH: Oh, I immediately forgot the pronunciation SW: Laughs SW: So sorry, that was gross SH: Laughs, couches SW: Laughs, That can not go in the outtakes laughs SH: Laughs SW: Are you going to talk, are you going to lose it again? Laughs SW: Mathe-mathical SW: Vocalizing a song SH: Vocalizing a song (Outro music ends)