诺贝尔物理学奖得主:改变一切的量子跃迁——John Martinis
Martinis 与诺贝尔奖相关的实验之所以有价值,不只是因为它完成了证明,更因为它奠定了超导量子计算这一领域的基础。 他在1985/86年完成的电路实验显示,一个包含数十亿电子的宏观电气对象也能呈现离散的量子行为;Google后来沿着这一路线扩展,在2019年完成了53量子比特的“量子霸权”演示。他判断真正突破的标准是,它是否能带来“其他实验、其他论文和其他发明”。
商业化鸿沟残酷地体现在数字上:今天可控的超导系统大约有50-100个量子比特,而通用机器可能需要约100万个。 量子比特在噪声抹去记忆前,或许只能完成100次、1,000次或几千次操作,因此必须持续进行量子纠错。Martinis 说:“一百万是个好用的整数”;实用机器既要大得多,也要干净得多。
Martinis 的公司目标是在大约8-10年内做出有实际意义的成果,同时也承认,行业已经“预测10年”有一阵子了。 现有机器能够运行真实算法并产出可发表的实验,但仅仅声称计算任务很难,并不能让它们具备经济价值。他直言“炒作多于现实”,因此这个预测仍取决于制造和噪声瓶颈能否被解决。
扩展路线的核心是工业化制造,而不是实验室里又一个量子比特里程碑。 Martinis 的公司正与 Applied Materials、Synopsys、Hewlett Packard Enterprise 及理论初创公司合作,转向现代300 mm半导体工艺,目标是实现类似GPU的成本和质量。如果这次制造跃迁成功,他相信“我们可以非常快地扩展规模”。
Martinis 认为中国团队在技术上已接近同一水平,并担心中国当局可能会限制论文发表,直到西方出现可比结果。 他说,复现Google量子霸权实验的中国团队“知道自己在做什么”;Google发布明显改进的结果后,中国很快就释放出性能相近的信号。他提出的制造业优势,是获得中国尚无的先进制造设备和工艺,包括某些300 mm设备。
AI 可能帮助量子计算,但 Martinis 不接受它可以替代高质量硬件和清晰控制这一说法。 他认为AI可以用于建模、量子纠错解码以及量子-AI混合算法,但仍然“有点老派”:制造出来的硬件噪声过大,软件再强也无法把它挽救成出色性能。近期重点是工艺工程、材料、控制和纠错,而不是假定AI存在捷径。
1. 从车库动手造东西,到把 Leggett 的问题变成实验
Martinis 将自己的实验直觉追溯到San Pedro:他的消防员父亲没有读完高中,却是“一个非常聪明的人”,总是在车库里做各种项目。物理学之所以让他着迷,是因为他看到动手实践的世界背后有数学规律;在UC Berkeley,导师John Clarke让他接触到电气器件中的量子行为。
Anthony Leggett 提出的关键问题是:“宏观物体是否遵循量子力学?”候选对象是一条包含数十亿电子和原子的电路,其集体电流和电压可以检验:量子规律是否也适用于那些理论最初针对的微观构成之外的宏观对象。
Friedberg 将微观物体描述为概率分布,而不是沿着预定轨迹运动的点。Martinis 进一步用原子把这个想法讲清楚:按照经典力学,吸引力会把电子和原子核拉到一起,但电子是“模糊的”,是一个延展的波;它允许存在的驻波频率,决定了原子的大小和特征光谱。
Leggett 关于“薛定谔的猫”悖论的核心是经验性的:把猫视为可能同时处于生和死的状态,前提是宏观物体能够占据量子叠加态,但当时没有实验证明这一点。Martinis 看到的是一个异常深刻的博士论文题目:与其在哲学层面争论悖论,不如直接在电路中检验这个假设。
2. Josephson 电路让宏观量子行为变得可测
量子隧穿是最初提出的测试方案。粒子的波大多会被势垒反射,但有一小部分可能出现在势垒另一侧——“就像穿墙一样”。这一效应已经应用于存储电路:电子会穿过只有10-20个原子厚的绝缘体;磁存储器也依赖隧道结。
人体实际上永远不可能穿过墙,因为必须同时有太多原子处在恰当的位置和动量上。电路改变了概率:在接近5 GHz的频率下,它可以“每秒尝试50亿次”跃迁,让一个原本极其罕见的宏观事件获得足够多次机会,最终变得可观测。
Martinis 解释说,在超导体中,“所有电子都凝聚到同一个状态”,集体运动而不会发生随机散射。Friedberg 儿时用液氮冷却超导圆盘、让磁铁悬浮的演示,让这种持续性变得直观;MRI磁体同样能让超导电流和磁场维持很长时间。
Josephson 结在两块超导体之间放置绝缘势垒,让Cooper pairs成对无损隧穿。与电容结合后,其非线性动能电感构成LC谐振器。决定性观察是离散的能量频率,类似钠灯特定的黄色谱线——这直接证明了宏观电路遵循量子力学。
3. 一个反常结果,变成了一套计算架构
这项工作大约发表于1985或1986年的Physical Review Letters,并引起广泛关注,包括Scientific American的一篇报道。但Martinisin回头看并不浪漫:证明宏观量子力学成立当然值得关注,但人们仍然会追问:“它有什么用?你接下来要做什么?”
在UC Santa Barbara的一次会议上,Richard Feynman在闭幕演讲中提出用量子力学进行计算。Martinis 承认自己没有听懂全部内容,但人群围住Feynman的场面让他确信,这是“最有意思的基础问题”。Peter Shor在1990年代初提出的因式分解算法,后来给出了一个具体问题。
Martinis 继续在法国Michel Devoret的团队中研究器件物理;在NIST工作时,David Wineland的团队就在走廊另一头。他还开展了电子计数和计量学实验。到了1990年代末,随着理论日渐成熟、美国政府资金开始出现,他决定“全押上去”。在UCSB,他的实验室从基础器件一路推进到五量子比特和九量子比特计算机。
Google提供了资金和团队连续性,这是学术界很难维持的复杂机器项目。2019年,Martinis 的团队发表了53量子比特的量子霸权实验,其数学输出更难由经典计算机模拟。他强调,这“并不实用”,但证明了量子计算机的能力,也证明它已经能在有意义的规模上运行。
4. 今天的机器已经能计算,但距离实用仍差着几个数量级
超导量子比特在概念上仍接近最初的实验:Josephson 结电感、电容,以及一个接近5 GHz的振荡器。微波脉冲改变其状态,读出电路进行测量,电容耦合则把多个这样的元件连接成阵列。
当前超导系统大约有50-100个完全受控的量子比特,已经能够运行真正复杂的算法。中性原子是“这个领域的新来者”,大规模阵列已经得到演示,但Martinis 说,它们的门操作仍需要更好的控制。
瓶颈在噪声。根据质量不同,量子比特在丢失记忆前可能完成大约100次、1,000次或几千次操作。Martinis 将其比作必须不断刷新的动态随机存取存储器,但这里的刷新依靠量子纠错,这会把通用系统推向约100万个量子比特。
被问及实用计算的时间表时,Martinis 说他的公司和许多其他团队都希望在未来8-10年内做出成果,同时立刻保留不确定性:人们已经预测10年“有一阵子了”。当前系统足以支持科学论文和实验,但“还没有大到真正有用的程度”。
5. Martinis 选择的扩展杠杆是制造,而不是AI
当被问到AI是否会像加速聚变或材料科学那样加速量子工程时,Martinis 的回答刻意克制:“可能会有”有用的建模应用、纠错解码和量子-AI混合算法。但硬件制造质量差、控制不清晰,无论软件多么复杂,都无法带来出色性能。
他的公司判断,关键技术瓶颈就在现有制造方法中。工艺一代代推进:从1985年的简单器件,到2000年前后的更复杂工作,再到2019年的量子霸权芯片;下一次跃迁,是采用现代300 mm半导体设备、标准工艺和新的工艺配方,目标是实现类似GPU的成本和质量。
中国让这件事更加紧迫。Martinis 说,复现Google量子霸权实验的研究人员理解背后的理论,也做出了“很好的结果”;Google发布另一个明显改进的结果后,中国团队很快就释放出接近同等性能的信号。Friedberg 说,他听到过同样的担忧:中国当局可能会阻止论文发表,直到西方媒体披露可比结果。
Martinis 认为,获得先进制造能力可能构成一种防御性优势,因为他们计划使用的300 mm设备,据他说在中国无法获得。他的联盟包括Applied Materials、Synopsys设计工具、Hewlett Packard Enterprise和理论初创公司。这个赌注是,工业工程能够带来“巨大跃升”,同时帮助守住他们的领先地位。
6. 影响力,而非新奇性,让80年代的结果成为诺奖故事
Martinis 通过后续影响衡量这项实验的重要性:如今有1,000名、或许几千名研究人员在开发超导量子计算机,也有公司在销售机器或机器使用时间。这个领域还推动了材料、制造、控制系统和测量等方向的进步——他称之为美丽的“工程与物理”。
Martinis 说,他和Michel Devoret等同事曾参加诺贝尔研讨会,组织者会评估这个领域的活力,以及潜在获奖者能否做一场有分量的代表性演讲。他明白自己正在被考虑,但强调,仅仅收到邀请本身就已经是极大的荣誉。
早些年的诺奖公布日,他曾短暂失望;后来他开始不喜欢这种心态,因为仅仅进入考察范围就已经非同寻常。今年他忘了公布日期,妻子接到了凌晨3:00的电话,但让他一直睡到5:30,免得6:00到场的记者来时他心情不好。
离开公司后,Martinis 仍然被仪器制造者吸引。他特别提到Ben Mazin为系外行星搜索开发的超导探测器,这与Martinis在1990年代帮助奠定的工作有关。从车库项目到量子处理器,这种吸引力始终如一:“我喜欢制造仪器”,尤其是在更好的设备能够打开新的观测领域时。
Welcome. Today, I’m very excited for this All-In interview with this week’s Nobel laureate, winner of the Nobel Prize in Physics in 2025, John Martinis. John, welcome to the All-In interview.
Thanks for inviting me. I’m quite excited about this talk and love explaining to people what this prize is all about.
The Nobel Prize is the most prestigious honor, particularly in physics, that can be awarded. You’re in the record books, and it’s going to be an incredible ceremony coming up for you. Maybe we could go back to the beginning of your history. I’d love to hear a little bit about where you grew up and how you got started with your interest in physics.
I grew up in San Pedro, California, and lived there my whole childhood. My father was a fireman, and my mom stayed at home and took care of us. Through the years, I was always interested in science and technology.
One of the things I’ll say is that my dad didn’t have a high school education, but he was a very smart person. He was always building things in the garage—various projects. So I grew up knowing how to build things, which also tells you how things work: an empirical, tactile view of how physics works.
When I took physics in high school, I loved it because there was math behind it, along with concepts that really made sense to me. I fell in love with the subject, then went to UC Berkeley, did pretty well there, and enjoyed it a lot.
In my senior year at UC Berkeley, I took a class from John Clarke, who became my adviser, and found out what he was doing. He was just starting to look at quantum mechanics and electrical devices, and it sounded really interesting to me. I guess I could see when something might take off, so I started doing graduate-school work with him.
You went to Berkeley for graduate school.
I went to Cal for graduate school, which you’re not supposed to do.
I was originally a physics and math undergraduate at Cal. I changed my major later and actually got my degree in astrophysics. There was an upper-division math class that really turned me off to math as a major. There were so many proofs; it drove me nuts.
Physics was always exciting, but I liked working in the astro lab, and I actually worked at Lawrence Berkeley Lab.
Oh, okay. Yeah.
But then you stayed at Berkeley and went to grad school, right?
Yeah, I stayed at Berkeley and went to grad school. We started this project a couple of years into grad school—I forget the exact date.
What was interesting is that this was a question posed by Professor Anthony Leggett, who won the Nobel Prize for helium-3 physics in 2003.
Was that superfluid?
Superfluid helium-3. Yeah, that’s right.
He showed that if you put helium-3 in a cold enough environment, it develops a new sort of characteristic in its physics—in how it moves and how it works. It has superfluid behavior, but it’s very complicated because of the more complicated nucleus of helium-3.
This had been discovered, and people worked for a while to figure it out. He helped develop the theory for it, so he was quite well known and a very smart person.
Although he won the Nobel Prize for that, there’s not much helium-3 physics going on. But for the question that led to our experiment, there’s a huge field. The question was: Do macroscopic objects behave quantum mechanically?
A macroscopic object might be a small ball. In our case, it’s an electrical circuit with billions of electrons in it—billions of atoms. Is the collective motion of, say, the ball, quantum mechanical?
If you think about throwing a ball against a wall, it’s going to bounce off. But if you make the wall thin enough and the ball light enough, every once in a while it will tunnel through because of the laws of quantum mechanics.
Hold on. Let’s pause on that for a second, because I think that’s really worth spending a moment on.
Yeah, great.
When we talk about quantum mechanics, when we talk about the relative position, energy, or movement of a particle at the atomic scale—as small as an atom or smaller than an atom—we have to use probabilities to describe where things are going to be.
That was the big understanding of quantum mechanics in the early 20th century, right? There’s a probability of things being where they are and moving as they’re moving. It’s not deterministic like what we see with the ball that we throw around. When you get very small, things get very fuzzy, and it’s very hard to know exactly where they are.
You hit upon the key idea here, maybe by accident, but it’s very important. Quantum mechanics was developed as a theory of small things: electrons, atoms, and the fundamental constituents of matter.
If you take an atom, it’s made from an electron and a nucleus. Classically, they attract each other and would just combine together, and then atoms would basically have no size. Why do atoms have size? That was one of the strange things.
It’s because an atom is not a point particle. I used to say to my kids that the electrons were fuzzy. Quantum mechanically, an electron has a wave function and is extended. You can think of the electron as being all around the nucleus at the same time.
It’s just very strange behavior of small things, and of course it’s very important to how atoms work and how we describe nature. Quantum mechanics ultimately became a field that people say is very nonintuitive in terms of understanding where small particles are, the energy they have, and where they’re moving.
We had to figure out that we needed to use these functions. It’s not just a single point, but a distribution—a whole bunch of places—and there’s a probability of where the atom or electron could be. There’s also a probability of how fast it might be moving. All of these things become probability functions.
You develop a mathematical theory for doing this that takes you until your third year in university to really know enough math to understand. But basically, that’s right.
These are forming waves—the waves of the electron. You have a kind of wave around the nucleus describing what the electrons are. These are like standing waves, like hitting a string. Different-length strings and strings with different tensions form different notes. These vibrations of the electrons around the atom can vibrate at different frequencies.
Rather than think about an electron moving around an atom in a predescribed path, where I can know where it is at any point in time, the right way to think about an electron around an atom is that it’s in a wave. There’s a wave that describes where it is.
You have the electron and the proton attracting each other. The whole wave theory combines those two and gives you a description of how the atom works—a quite accurate description, too.
The probability of something extreme or extraordinary happening is small, because everything at the microscale is described by wave functions. One example is that Stephen Hawking figured out you could have a particle and antiparticle come out of nowhere in the middle of space. The antiparticle goes into the black hole, and the particle shoots off.
The probability of that happening is very low, but it happens enough that the antiparticle actually starts to deplete the black hole. That’s how black holes evaporate.
Can you tell us what quantum tunneling is? This is another feature of quantum mechanics that arises from the fact that these things are waves and probability functions.
If you have an electron traveling through space and hitting a wall, let’s say, there’s a little wave packet—a wave function—to it. It’s not a single particle; it has some extent to it.
When that particle hits the wall, quantum mechanics says there is some small amount of this wave function, or, if you like, the particle, going through the wall and appearing on the other side.
Most of the time it bounces off, but every once in a while it goes through. This is seen in everyday devices. If you build a very small memory circuit, you have to worry about electrons tunneling and charge leaking off your capacitor. There are magnetic memories that depend on these tunnel junctions.
This is a very well-known phenomenon. If you make the barrier—the insulator—just 10–20 atoms thick, that’s thin enough for it to go through.
You can actually predict the number of electrons that might tunnel through one of these barriers—one of these insulating barriers, as they're called—to the other side, which really is crazy to think about. It's just like walking through walls, right? I mean—
Yeah, that's the idea.
Yeah. So, going back to the story you were sharing, you're in grad school—
Right?
And then Leggett proposes this idea. Maybe you can share a little bit more now that we've got, I think, a bit of the basics on what was discussed, which was: zooming out a bit, rather than just thinking about all of this happening at a microscopic scale, is it possible for it to happen at a bigger scale?
Yeah. Again, we've been talking about quantum mechanics as the physics of nature at the microscopic, atomic scale. But the question was: if you made a macroscopic object, would it obey quantum mechanics also?
That was the basic question, and it turns out that there's a very natural system to look at: an electrical system. You can look at the currents and voltages of an electrical oscillator and ask: does it behave according to classical physics, or does it have this quantum-mechanical nature to it? That was the question.
Now, when you think about quantum mechanics, there's the quantum behavior, but at some point you have to measure it, which then turns it into a probability. There's something called the Schrödinger's cat paradox. In the paradox, you have radioactive decay, and then you let it happen for, let's say, half of the radioactive decay time. Then you have a radiation detector and a bottle of cyanide, which will kill a cat, and you ask: after some amount of time, is the cat in the dead or the live state?
Physicists have discussed this question. Einstein brought it up, or Schrödinger brought it up, and a lot of people discussed it. But Leggett pointed out that the reason this is a paradox is that you can believe that a macroscopic object, like a cat, could be in a quantum-superposition state. In fact, there was no experimental evidence that this could happen. That was his point.
He said, “People should be testing this. Let's see if it's true.” As a young graduate student who had just learned about quantum mechanics, I thought, “That's a really great question. That's something that we should try to do.” We should try to do an experiment on the suggested system to look for quantum mechanics. The original proposal was looking for tunneling. It turned out to be more than that, but the original proposal was to look for tunneling.
Let me just describe it another way. The macroscopic system could be my entire body. Could I walk through a wall?
That's right. The probability of all of my atoms being in the perfect momentum and the perfect position to be able to cross through the wall is so low, it would never happen in this or many other universes.
That's the problem: for most macroscopic objects, when you try to think about quantum mechanics, that won't happen.
Right. There's a small probability that 1 electron can cross over a barrier—
But the probability that many cross over at once is lower and lower and lower, and that makes it very difficult to see at scale. What happens is, if you look at an electrical circuit, the parameters become favorable for seeing this kind of macroscopic behavior.
It's hard to go into the whole physics of all that, but it's basically because you can make a circuit that operates at microwave frequencies. Instead of trying to go through the wall once a second, it tries to go through the wall 5 billion times a second. You have more chances to go through.
The other thing is that the various parameters involved in quantum mechanics are favorable for seeing this kind of phenomenon. You have to do the experiment right, but the parameters are favorable for doing that.
So, one of the parts of your experiment was that you created what's called a Josephson junction. Is that correct? This is 2 superconductors with a barrier between them, right?
I got really fascinated by superconductors when I was maybe 12 years old. I went and bought a superconducting disc of yttrium barium copper oxide.
Yes, that's right.
From the back of Popular Science. Then I went to UCLA and got a jug of liquid nitrogen, and then I floated a magnet above the disc because of the Meissner effect. I had it at the science fair, and I did very well with the science fair that year.
What year was that? Was that when it was discovered?
Must have been 1989.
Okay. Yeah, that was close enough. That was good. The hard part is getting the liquid nitrogen.
Yeah, I had a friend whose dad was a doctor at UCLA or something like that, so he was able to get the liquid nitrogen for our demonstration.
Right. Yeah, that was the hard part.
I've always been fascinated by the physics of superconductors. Maybe you can just explain one of these important features of superconductors as it relates to resistance and current flow, and then we can talk about your experiment.
When a material goes superconducting, all the electrons condense into one state. To give you an analogy—not a perfect analogy, but a close analogy—if you have a normal metal, any metal at room temperature, it's like a gas of electrons. It's like gas in the air.
Sorry, I think we should just explain that. You have a metal, and all the electrons are moving around. They're perturbed, and they're all at different energies and in different states.
That's right: different energies, different states. There's some Fermi statistics—we won't go into that—but it more or less looks like a gas. You think of a gas, and then when you cool it below a certain temperature, it then coalesces into, let's say, a solid, like atoms will. The electrons coalesce into something called the Cooper-pair BCS condensate, where all the electrons are kind of locked together and doing the same thing.
The nice thing about that is that it's not like they're frozen in place. They have a free parameter that allows all the currents—all the electrons—to flow in some direction, which is the supercurrent.
In a superconductor, meaning a material that's cool enough that it reaches its superconducting critical temperature, all the electrons can still move. They can still create a current, but—
But they're moving together, like in my analogy, like they're in a solid instead of a gas.
And because they're moving together—
When you work through all the physics, they aren't randomly scattering off things. They're just moving together, and then you get a supercurrent. For example, if you made a ring of superconductor, that current would basically flow forever around the ring. This is what you saw with the floating magnet.
Right. That's so interesting. I've always thought—and there have obviously been companies started around the idea of creating an infinite battery—that you could store electricity technically forever because the electrons are just moving around. If it's superconducting, they can just spin forever around that circuit.
People actually do use big superconducting magnets to store energy. When you get an MRI, you're in a liquid-helium machine with a superconducting magnet. They charge it up, and that magnetic field is basically there forever, waiting for people to go inside it. It's kind of strange to be inside this super-cold magnet, but they've designed it very well. It works well.
So, this Josephson junction is 2 superconductors on either side of a barrier that you create—an insulating barrier. Maybe just explain the experiment and what you guys measured. This was all while you were in grad school, right?
Yeah. Yeah. This Josephson junction is where the Cooper pairs have to tunnel through, but they kind of tunnel through it together without any loss. This actually forms what's called an electrical inductor in circuits.
An inductor is normally a coil of wire that stores energy in its magnetic field. Here, this just stores the energy of the electrons tunneling through it. We call it a kinetic inductance, and it forms a nonlinear inductance.
With a capacitor in the circuit, that forms an inductor-capacitor resonance circuit, which is like the circuits in your old radios. You have filters of LC resonance circuits to filter your signal and do other things. This is a very common microwave and radio-frequency element that you use all the time to make electrical circuits.
I just want to simplify that. You have these 2 superconductors split by this barrier. Some electrons tunnel through the barrier to the other side, and then you can effectively measure all of these different changes as you change the temperature. You guys were putting different voltage states into this circuit that you built.
What you saw and measured, and what you demonstrated, was that there were very discrete or specific changes that happened, which basically demonstrated quantum mechanics at scale.
That's right. This inductor-capacitor resonator, which you just treat as a charge and a current going through, has a wave function because it's quantum mechanical. There's some uncertainty in these, and given the way the simple electrical circuit works, you can demonstrate quantum mechanics. One example is tunneling, which is a little bit hard to describe here, but you can see tunneling.
I think the easier thing is to look at the energy levels of this. When people discovered atomic physics and started doing this, they excited a gas, and the light coming out of that gas would be at certain colors or frequencies. If you go outside and see the sodium lamps, these yellow lamps have a single frequency coming out of them. Nowadays, if you look at LEDs, there are certain frequencies that come out of them.
This is a quantum mechanical effect involving how the electrons travel around the atom. There are only certain frequencies that they oscillate at. Classically, you would expect all different frequencies as the electron spirals around or spirals into the nucleus. That's what you expect, but we saw these discrete frequencies.
By measuring those discrete frequencies, you now had proof that there was quantum mechanics happening at a macro scale.
That's right.
You published this work. Was there a lot of attention when you published it? This was in 1985 or 1986.
1985—or I actually forget, but 1985 or 1986.
Was there much attention on this work at the time?
Yeah. This was a big question, and people wanted to understand it. We published it in Physical Review Letters, and it got a lot of attention. I think we had a little article in Scientific American that wrote about it, and we were very proud of that. It was kind of a big deal.
What did you go on to do at that point? Was it considered groundbreaking, Nobel Prize-winning work? What was the story at that time when this came out?
It was an important piece of work, and people noticed it, but we showed that quantum mechanics worked on the macro scale, which was nice. One could still argue, “What is it good for? What are you going to do with it?”
The secret to an important scientific breakthrough is whether it leads to other experiments, other papers, other inventions, and the like. That took many decades to happen because it was so new, and people had to develop it. I would say it was noteworthy at the time, but not necessarily something for a Nobel Prize because it was just weird. People wondered, “What are you going to do with it?”
What happened at the time was very interesting. At the end of my thesis, there was a conference at UC Santa Barbara, where I came for the first time. They were talking about this experiment, but the very last day, the last talk was by Richard Feynman, a very well-known physicist.
Of course—the greatest.
The great, yeah, right. I idolized him and read his books and whatever. He was talking about using quantum mechanics for computation, which is building a quantum computer.
He gave a talk that was really amazing. I'm going to be honest: as a student, I didn't quite catch everything. Michel Devoret, my dear friend, said, “Yeah, maybe some of the things weren't quite figured out at the time.” But afterward, Feynman was absolutely mobbed by people asking him questions because it was so interesting to think about taking this basic law and actually doing computation with it.
Right.
I was a graduate student, so I was at the outside ring. You had the professors in close, and I was just a lowly graduate student. I could hear a little bit, but what I learned from this was that it was a great question and something worth doing as your life's work because it's so deep, so interesting, and maybe practical.
The big idea was to use quantum mechanics and these properties of quantum mechanics to do computing.
That's right. Soon after that, other people in the field got a little bit more specific and showed how you would do it. Then, in the early 1990s, maybe 5 years later, Peter Shor came up with this factoring algorithm to solve a real-world problem with it.
It took people a while to figure out. It was very abstract, and people weren't quite sure what to do. But, like I said, I could see from the crowd around Feynman asking him questions that this was the most interesting fundamental question: how to combine quantum mechanics with computation. It's really amazing.
You started to do that with your life's work. You went on to have a very good career.
My career path was that quantum computing was getting developed, and it took me a while to really go all in on it. What happened was that Michel Devoret was from France, from CEA France. He went to Berkeley and went back, and I went there as a postdoc and worked with them.
They were young and unknown at the time. People said, “You're going to go to Europe, and you're not going to get connected to U.S. science.” But I knew Michel Devoret, Daniel Estève, and Christian Urbina, the people I was working with, were absolutely brilliant, and they've had very illustrious careers. I went over there because I knew it was great, and we continued to do experiments on this.
After that, I came back to the U.S. and worked for the National Institute of Standards and Technology. Just down the hall was David Wineland and his group, who won a Nobel Prize for atomic physics and quantum computation. I worked on experiments involving counting electrons and metrology, and then did other experiments.
In the late 1990s, I again went all in on building a quantum computer. There was funding available at that time. The theory had progressed enough that the U.S. government started funding it to see if people could do it.
Then, a couple of years after 2014, I think you ended up at Google's quantum lab in Santa Barbara. Is that right?
I was at UCSB for 10 years or so, which was wonderful, and built up the lab from very basic things to building a 5-qubit and then a 9-qubit quantum computer. During that time, Google got interested, and I decided that although academia was great, it would be hard to get the team together and keep them together for a long time to build this complicated machine. Google had the money.
We went there and started off fairly small, mostly with people coming from my UCSB group. Then, in 2019, we published this quantum supremacy experiment with 53 qubits. We made a lot of qubits and made them really good and fast so that we could run a mathematical algorithm that produced some output that took much, much longer for a classical computer to emulate.
It wasn't practical, but it was a demonstration of the power of a quantum computer—that it worked.
Maybe give your description of a qubit, and then we can relate how we build these quantum computers from qubits to the Josephson junction and some of the early work you had done that you ended up winning the prize for.
Very simply, we have a metal wire and another metal wire that are put together on this Josephson junction, which represents an inductor, with current flowing through it. From one wire to the other, we have a capacitor, and we set that up to oscillate at about 5 GHz, or cell phone frequencies, to form the qubit. It's this oscillating thing.
At low temperatures, the superconductors and all this magic allow us to get quantum mechanical behavior out of it.
Then you can measure that quantum mechanical behavior, create a representation, and use that to run your computing.
That's right. You put on microwave pulses to change the state of the quantum computer and change the way it oscillates. Then we connect it to complicated readout circuitry to figure out what state it's in.
Okay. Then you connect an array of these and use capacitive coupling from one wire to the next to couple them together. It's more complicated than that, but that gives you a good idea.
To understand your work that you won this Nobel Prize for, which demonstrated this quantum mechanical phenomenon at scale, is that part of the design of a qubit and the circuitry? Did that inform that design work?
Yeah. It was the very basic, simplest circuit. We were using analog simulators at the time. I took data with a computer, but this is far back enough that it was very rudimentary.
Over the years, the whole field got more sophisticated in its design, with many, many people contributing. We were able to put things together in a way to actually build a computer now.
Right.
The reason it's interesting from the Nobel Prize perspective is what it led to. What it led to right now is 1,000, maybe several thousand, people around the world doing research to build this superconducting quantum computer. It's turned into an enormous field: a large number of papers, a large number of people, and people selling quantum computers. IBM is selling quantum computers, and people are selling time on quantum computers.
The fact that it was a useful idea brought all these different experiments and ideas into being, and many, many people contributed to this.
It's very interesting. I think the broad question or observation is that sometimes an inquisitive mind leads to research, which leads to a set of discoveries that are completely not apparent until 40 years later.
The effect or impact it may have had on building an industrial field—there's now quantum computing. Everyone feels it's on the brink of actually achieving what people have talked about in theory for decades, and it seems to be getting very close to doing it.
Yeah, I can talk about that, but I would say this field has generated many other ideas about how to build a quantum computer. It is a very exciting, quite large field, and I would say the science is very deep, too. To get these things to work, you have to invent lots of different devices, think about materials, fabricate them, and build complex control systems. Engineering and physics are, to me, quite beautiful.
Just to tell you a little bit about me, I grew up building things and, as an experimentalist, I like to build instruments and experiments to show this. This was kind of the ideal project for me because, from very early on, it was, “Let’s do this great physics, but let’s also build something.” By asking, “What do we have to do to build a quantum computer?” that led me to understand what physics we had to test and what kinds of things we had to build.
That’s just the way my mind works. I’m much more practically oriented, so it was a perfect field for me to get into, and that’s what intuitively led me to want to do this in graduate school. I think it’s fascinating how much engineering and technology you have to do to make this work.
Where are we in the evolution of quantum computing today? What’s the state? At what point will we have—call it—generally accessible and generally useful quantum computers that can do all of the amazing things everyone has talked about for decades, things one would be able to do with quantum computers?
That’s right. Right now, we’re at about 50 or 100 qubits in the superconducting case, but they can be fully controlled, run real algorithms, and do very complicated things. There are a lot of other systems that can do that. I think the newcomer on the block that looks good is neutral atoms, where they’ve made big neutral-atom systems, but they’re still working to get the gates controlled really well and the like.
What’s happened right now is that we can run genuine algorithms on them, and people have ideas they want to run. But because these qubits are not perfect—it’s an analog control system, and fundamentally these quantum bits have a little bit of error and noise—you can only run a project of limited complexity. It’s good enough to write scientific papers and try things out.
Every once in a while, people say they’ve done something that’s hard to compute, and that’s fine, but they aren’t really big enough to be useful yet. They have to get bigger and better, with less noise.
Do you have a point of view on the timelines? This is everyone’s speculation, and there’s been more hype than reality.
Yeah, there’s more hype than reality, and it’s hard. I used to not want to speculate about that, but since I started a company, I can do that. What we want to do—and it’s the timeline of many other groups—is to do something in, let’s say, the next 8 to 10 years, something like that. But the problem is that people have been predicting 10 years for a while now, so we have to do that.
I can tell you that, for what we’re doing, we’ve identified the technology bottlenecks in the current fabrication methods for making a quantum computer. We’ve written some papers on it, and we’re working with people in the semiconductor industry to manufacture this in a much more cost-effective, high-quality way—the way you make these GPUs or something. We think that when we get that to work, we can scale up very rapidly, so on a 10-year timescale, something like that.
In a lot of technically difficult fields, like fusion energy and perhaps even quantum computing, we’re seeing profound acceleration in getting to crazy big goals on these very big technical projects because of AI. Is AI starting to play a role in solving some of the engineering, materials science, scaling, and noise issues that we’ve seen historically? And do you think that there’s an acceleration underway in performance improvements because of AI?
There may be. There are things we can perhaps do with modeling and the like. We also think we can use the quantum computer and AI together to solve problems better. That’s what our theory team is proposing. I used to work with Google Quantum AI; that’s what they’re proposing, so there’s a general feeling of that.
My particular view, though, is that in terms of control, if you don’t build your system cleanly enough and make sure the control is clear enough, you’re not going to get great performance out of it. So I’m a little bit old-school here, working on building it that way. There are certainly some elements where you can use AI, such as in the decoding circuit for error correction and the like.
The one thing to mention is that these qubits are naturally very noisy. You can maybe do 100 operations for bad qubits, and maybe 1,000, maybe a few thousand operations, before they kind of lose their memory. You can think of it as dynamic RAM, where you have to refresh it. Well, you have to refresh it with error correction.
Because of that, you’re talking about a million-qubit quantum computer to be general-purpose and solve really hard problems.
A million-something.
A million is a good round number for it. Maybe a little bit more. Right now, we’re at 100 or a little bit more than that, so we have a ways to go.
What is your view on China and the progress that they’re making in this technology versus the U.S.? This is the topic in every field—industrial fields, computing, and science: Where’s China compared to the U.S.? Everyone’s worried about the progress in China versus the U.S. and what that means.
I can talk about my own field. When I’ve read the papers that duplicated what we did at Google in the quantum supremacy experiment, they know what they’re doing. They go through the theory, and a lot of it is very similar to what we’re doing, but they know what they’re doing and they’re getting great results.
The thing that scares me a little bit is that last December, the Google group published the latest results, which are really much nicer. They made some real improvement, but then China soon afterward published something indicating that they were on par or near par with it. I’m worried that the Chinese government is saying, “You can’t publish anything until it’s in the Western press, and then it’s open and you can talk about it.”
That’s precisely what I’ve heard.
Yeah, I’m a little bit concerned about that. Now, what we’re doing with our company is a new generation of fabrication for the devices. I would consider, in my research, that we had the simple fabrication with the original papers in 1985; then, around 2000, we had more sophisticated fabrication; and for the quantum supremacy experiment, we did something even more complicated. Other groups did, too, but we want to make a similar jump in fabrication.
What’s interesting about this is that we’re going to be using Applied Materials and the modern fabrication processes they have. On 300-millimeter tools, you can’t get that in China, for example. You can get it for CMOS, and we’re developing standard processes with new recipes and new ways to put them together.
We think that by doing that, we can make a huge leap forward and get there faster, in a way that will protect our lead. There are other things we’re doing, too, but that’s a small part of it. We think there’s a way to really lead the field, and we’re happy we have good industrial partners: Applied Materials, Synopsys design tools, Hewlett Packard Enterprise, and some startups that do the theory work. We have a good consortium, and we want to use all that engineering knowledge and expertise to make this happen.
Where were you when you got the news this week that you won the Nobel Prize, and how surprised were you? This is a 40-year-old research effort. Had anyone given you a call—the rumor mill, the gossip mill—saying, “Hey, you’re on the list this year, potentially being considered”?
Let me give you a little bit of the inside story. We’ve known that this was an important experiment from the beginning. We’ve obtained some other prizes that are much less well-known, and we’re really appreciative of all that.
What happens is the Nobel system puts together Nobel symposia, where they get together physicists in a certain field, which is quantum information and this kind of thing, and they have all the scientists give talks. They want to check on the vitality of the field—how big is it? They also want to see whether some of the leaders they’re considering can give a good talk and would be a good representative.
Michel and John and I have been to these symposia before, and we kind of knew what was going on—that at least we were being considered. But I’ll just tell you, as a scientist, being invited to these and being considered is a fantastic honor. Getting the prize is so unbelievable that you shouldn’t think that way.
So, I've known about it for a few years. In fact, to be very honest, in the past, when the dates have come around, it's like, “Oh, is this going to happen?” Then you wake up in the morning and it's like, “Oh, it didn't happen,” and you're kind of down for a day. It didn't happen this year, and that's a very bad attitude. I don't like that at all. You should not covet some insanely difficult prize that only goes to a few people.
So what happened this year is that I kind of worked through this over several years, and this year I just forgot about it. So I went to bed, and we got the call at 3:00. My wife answered the phone and found out what happened, but she didn't wake me up right away because she knew that if the day was going to be hectic, I needed my sleep so I wouldn't be grumpy.
That was nice of her.
You don't want to be grumpy talking about it. So she woke me up at 5:30, and as I looked at the computer, I said, “Oh my God.” Then we had some reporters coming over at 6, who interviewed me right when I had found out—half an hour after I'd found out. It's great. It's a great honor, and it's just been really fun.
I've been getting a lot of emails from people I've worked with or students I've had in the past congratulating me, and you exchange little stories and the like. It's kind of a very special time.
That's great. Any science or technology fields that you've been following outside of your core discipline that you think are really exciting?
To be honest, I'm just so focused on doing this, and especially when you start a company, you better be focused, right? So I'm doing that.
But one of the fields that I find interesting is what Ben Mazin at UC Santa Barbara is doing. He's looking for exoplanets, and they're using superconducting detectors that are somewhat similar to what we're doing. In fact, in the 1990s or so, I helped establish that field with other people and did that for 5, 6, 7 years. He's doing it in a different way. I really like how this instrumentation that we've been working on—these quantum devices—is now able to do these astronomy detectors and look for these.
Of course, there's so much going on in astronomy these days, with gravitational-wave detectors and exoplanet searches, and it's just fascinating to me. Again, it's very much technology-oriented, where people are building good detectors. This is what I like. I like building instruments. So that's particularly interesting.
Yeah, that's great. I mean, it's a very exciting field, and hopefully we'll develop quantum computers that will help us build materials and technology to help us get there one day.
So that's right.
Many rungs on the ladder of human progress. Well, congratulations again on winning the Nobel Prize in physics this year. Very well deserved. It's a fantastic moment. Enjoy it. Enjoy the ceremony, and we're excited for your continued work in the field of materials and quantum computing. And thank you.
Yeah, and thank you. I really enjoyed the questions and the flow, where you were asking questions to explain it at the right level for people. I really appreciate that. This is a great podcast.
Great. Thank you.