The future of ultrafast materials and devices

这里是斯坦福工程学院的《万物未来》，我是主持人阿尔特曼。

This is Stanford Engineering's The Future of Everything, and I'm your host, Altman.

自从八年前我们推出这档节目以来，它已经成了记录我的斯坦福同事那些杰出且富有影响力工作的宝库。

Since we started this show eight years ago, it's become an archive of amazing and impactful work by my Stanford colleagues.

研究不只是发生在实验室里，正如你们在节目中将听到的，斯坦福的研究能影响健康、科技、法律、商业，以及许多与日常生活息息相关的领域。

Research is not something that just happens in the lab, and as you'll hear on this show, the research at Stanford can impact areas like health, technology, law, and business, and many other topics that can affect everyday life.

我们希望你们能持续关注，了解科研如何有可能改善你们自己，以及你们家人和社区中重要之人的生活。

We hope you'll tune in to learn more about how research has the potential to help your life and to help the lives of people you care about in your family and your community.

作为材料科学家，我们的核心兴趣是理解材料的工作原理，弄清楚原子和电子的位置，然后从这个微观层面看，这些特性如何转化为功能。

At its basis, we're interested as a material scientist in understanding how materials work and understanding where the atoms are, where the electrons are, and then at this microscopic level, how this amazing leads to functionality.

如果你想用某种材料制造设备，就必须理解电子和离子如何在这些结构中流动。

So if you want to make a device out of a material, you have to understand how electrons and ions flow through these structures.

一旦掌握了这一点，你就能开始利用这些基本过程，制造出太阳能电池、电池这类设备。

And if you can do this, then you can start to kind of harness these kind of fundamental processes and make devices like solar cells or batteries, things This like

这里是斯坦福工程学院的《万物未来》，我是主持人阿尔特曼。

is Stanford Engineering's The Future of Everything, and I'm your host, Altman.

如果你喜欢这个节目，别忘了分享给你在乎的人。

If you're enjoying the show, please it with somebody you care about.

亲朋好友的推荐，往往是传播节目最好的方式。

Personal recommendations are one of the best ways to spread news about the show.

今天，亚伦·林登伯格会告诉我们，原子层面的过程在速度、所需能量和输出稳定性之间，存在一个根本性的权衡。

Today, Aaron Lindenberg will tell us that there's a fundamental trade off in atomic processes between their speed, the energy it takes to make them go and the reliability of their output.

这就是超快材料的未来。

It's the future of ultrafast materials.

在开始之前，先提醒一下，我们稍后还有‘一分钟未来’环节。

Before we get started, a reminder that we have the future in a minute.

在这段对话结束时，我会问亚伦几个快速问题。

At the end of this conversation, I'll ask Aaron some quick questions.

他会给出简短的回答。

He'll give me some quick answers.

那就是‘一分钟未来’。

It'll be the future in a minute.

在开始之前，再提醒一下大家，请把这档节目分享给那些会从了解万物未来中受益的人。

Before we get started, another reminder to please share the show with somebody you care about who would benefit from knowing about the future of everything.

材料其实就是构成我们周围世界的基础。

Well, materials are literally what makes the world around us.

尤其是电子、光学、电池和能源这些材料，它们支撑着我们如今完全依赖的所有科技。

In particular, electronics, optics, batteries, power, these are key materials that fuel our technology that we all have become totally dependent on.

一个关键概念是，这些材料通常被称为非平衡态或动态材料。

A key idea is that these materials are often what we call non equilibrium or dynamic.

也就是说，在原子层面，这些材料会随着外界作用而发生变化。

That is to say at the atomic level, these materials change in response to processes.

比如，一个质子或一束光打到太阳能电池的材料上，就能瞬间把它转化为电能。

So for example, a proton, a ray of light hits a material in a solar cell and it turns that dynamically into electricity.

这跟平衡态材料不同，后者对变化有抵抗力，即使施加外界影响也基本不变——有时候这种稳定反而是好事。

That's distinguished from equilibrium materials where they're resistant to change and actually they don't change when you give them inputs and sometimes that's a positive as well.

但大多数电子设备所依赖的，还是非平衡态或动态材料。

But for most of electronics, we're looking at non equilibrium or dynamic materials.

而我们能够理解这些材料，很大程度上得益于现在可以在超快时间尺度上对它们进行测量和观测。

And our ability to understand these materials has become revolutionized by our ability to measure them and look at them on ultra fast time scales.

林登伯格是斯坦福大学材料科学与光子科学的教授。

Well, Lindenberg is a professor of Material Science and Photon Science at Stanford University.

他通过在最高速度和最小尺度上研究材料，来探索它们的工作原理。

And he studies materials at the highest speeds and in the smallest scales in order to understand how they work.

亚伦会告诉我们，这些测量技术如何带来全新的认知，从而提升我们制造更优材料的能力。

Aaron will tell us how these measurement technologies are enabling new understanding that promises to increase our ability to create better and better materials.

所以亚伦，先说说看，你是怎么决定投身材料科学这个领域的？我们待会儿再深入。

So Aaron, start out, how did you decide to work in this area of material science, which we'll get to?

是的，其实我的背景是物理学。

Yeah, actually my background is in physics.

有时候，某些经历会深深印在脑子里，彻底改变你看待事物的方式。

Actually, you know, one of sometimes you learn you have experiences where, you know, they kind of imprint on your brain, and they kind of change how you think about things.

对我而言，读研的时候，我清楚地记得自己第一次接触到一个生物学问题。

So for me, actually, in grad school, I remember really clearly actually learning about a problem in biology, actually.

这虽然不是材料科学，但正是这件事让我走上了这条路。

So this is not material science, but this was kind of what got me on this path.

问题是关于血红蛋白和肌红蛋白如何工作，以及它们如何在人体内运输氧气。

And it was a problem about how hemoglobin and myoglobin work and how they carry oxygen through the human body.

当时人们用一种叫X射线衍射的惊人技术，来拍摄这些复杂晶体中原子的位置图像。

And so this was a problem where people used these amazing techniques called X-ray diffraction to kind of make images of where the atoms were in this kind of amazingly complicated crystal.

做这项工作的是马克斯·佩鲁茨，他在上世纪60年代因为这项研究获得了诺贝尔奖。

And when so Max Perutz was the guy who kind of did this and won the Nobel Prize back in the 1960s, I think for this.

我上大学时见过佩鲁茨，就在他去世前，那真是令人震撼。

I met when we Max before he died as an undergraduate and it was a huge Yeah.

据我所知，当他第一次画出这种分子的结构时，复杂得让人难以置信，根本看不出氧原子是怎么穿过这么复杂的结构，最终结合到中心的血红素上的。

So the story, at least as I learned it, was that when he made this picture of this kind of molecule, it was so complicated and there were the structure was so unbelievably complex that there was no way that the oxygen atom could kind of migrate through this complex structure and bind to this central heme group.

一方面，人们普遍认为，只要搞清了结构，就能理解它的功能，诸如此类。

And so they on the one hand, they had this idea that if you could understand the structure of something, then you could this would allow you to understand how it functions and all kinds of things like that.

但在这个案例里，它反而成了一个谜——后来人们才明白，这个分子其实是动态的，它会以一种极其复杂的方式打开，让氧气进入。

And rather in this case, it created a mystery, what people eventually realized was that this molecule was dynamic, it was opening up in this amazingly complex way to allow the oxygen to come in.

所以，正是这段经历让我真正兴奋起来，不仅对材料科学，而且对材料在原子尺度上的动态行为产生了浓厚兴趣。

So that was kind of how I got really excited, not just about material science, but also about kind of the dynamics of of how materials move at the atomic scale.

太棒了，这是个特别好的起点故事。

Great, that's a great origin story.

作为一名研究生物学的人，知道生物学能催生出如此出色的物理和材料科学研究，真的让人欣慰。

And as somebody who studies biology, it's great to know that we can spawn great physics, great material sciences work.

对于平时不常关注材料科学的人，能不能简单介绍一下，材料科学的目标是什么？特别是从你的角度，你们在研究哪些问题？这些研究最终想解决哪些核心问题？

So for people who don't think about material science all the time, can you just give us like a little thumbnail sketch of what is the goals of material science and particularly from your perspective and the problems that you're working on kind of getting us to what are the big questions that like motivate your lab?

是的，我觉得这其实是个根本性的问题，背后涉及大量基础科学，核心就是提出一些最基本的问题。

Yeah, I would say it's a question, there's a lot of fundamental science that goes into it, just asking really fundamental questions.

同时，也有很多与技术应用密切相关的问题值得探索。

Then also there are a lot of kind of really technologically relevant applied questions that come into it.

归根结底，作为材料科学家，我们关心的是材料如何运作——原子和电子在哪里，以及这种微观层面的复杂性如何最终转化为功能。

You know, at its basis, we're interested as a material scientist in in understanding how materials work, in in understanding where the atoms are, where the electrons are, and then at this microscopic level, how this how these this amazing complexity leads to functionality.

如果你想用某种材料制造一个器件，就必须搞清楚电子和离子是怎么在这些结构中流动的。

So if you want to make a device out of this thing, out of a material, right, you have to understand how electrons and ions flow through these structures.

如果你能做到这一点，就能开始利用这些基本的原理，制造出太阳能电池、电池这类设备。

And if you can do this, then you can start to kind of harness these kind of fundamental processes and, you know, make devices like solar cells or batteries, you know, things like that.

这太棒了。

So that's great.

我知道你们领域现在最令人兴奋的一点，就是能够以前所未有的精度观测到原子层面的结构。

And I know that part of what's exciting in your field right now is kind of unbelievable ability to measure and see atoms like at the atomic level.

你之前提到血红蛋白的研究难题，看来技术确实进步巨大。

You described this challenge of understanding hemoglobin and it sounds like the technology has really advanced.

你能给我们描述一下，现在你们手上有哪些强大的工具，能让你们对材料内部的运作机制看得如此精细吗？

And so can you paint a picture for like the kind of power tools that you know have at your disposal that enable super detailed understanding of what's going on in material.

没错。

Yeah, right.

其实很多技术发展，某种程度上都跟当年那个血红蛋白的问题有关。

A lot of it actually in some sense connects to this original hemoglobin problem, right?

马克斯·佩鲁茨当年做这些测量，用的就是X射线散射的方法。

So this was the way that Max made these measurements in the end, Max Perutz made these measurements was by using X-ray scattering approaches.

他们用一束X射线，本质上是波长极短的光，波长和原子间距差不多，然后让这束光照射到这些物质的晶体上。

So they took an X-ray beam, which is essentially a beam of of light with very, very short wavelength, wavelength comparable to the spacing between atoms, and essentially bounced it off of a crystal of these things.

通过测量衍射图样——这其实是一种干涉现象，光束从每个原子上散射出来，相互干涉，就像你看到地面水洼或DVD表面反射光线时产生的干涉效果一样。

And by measuring the diffraction pattern, which is essentially an interference effect, beams of light kind of scatter off of each of the individual atoms and create interference effects in the same way that you see interference effects when you, you know, look at light, you know, reflecting off of a puddle on the ground, or when you look at light reflecting off of a DVD or something like that.

所以，通过这类测量，你可以反过来推算出原子在结构中的具体位置。

So essentially, by making these types of measurements, you could kind of work backwards and reconstruct where the individual atoms were in the in the structure.

而近年来，也就是我刚读研究生时，出现了很多令人兴奋的新工具。

And then more recently, and this is where I kind of started out as a grad student, really exciting new tools have been developed.

其中一些就在斯坦福大学不远处的SLAC国家加速器实验室研发出来的。

Some of them just up the road from Stanford at the Slack National Accelerator Lab.

这些工具不仅能测量结构，还能捕捉这个过程的动态变化，观测它在极短时间尺度上的演变。

And these are tools which allow one to do this type of experiment, but not just measure the structure, but measure the dynamics of this process, how it's changing on amazingly fast timescales.

有些人可能觉得，这么快的时间尺度跟我们的日常生活关系不大，但其实当你放大看材料时，原子的运动速度惊人，快到飞秒级别。

And sometimes people have the sense that, you know, these very fast timescales are things that are, you know, maybe not that important to our everyday lives, but really materials when you zoom in, on them, the atoms are moving on amazingly fast timescales, femtosecond So time

这太棒了。

this is great.

你能给我们打个比方吗？我知道一点这方面的内容，但要想象这么快的事情真的很难。

So give us a sense, because I know a little bit about this and it gets hard to think about things being so fast.

能不能给不太熟悉这个领域的人讲讲，我们说的这种速度到底有多快，最好能跟一些他们平时能感受到的现象做个比较。

So can you lay the landscape for somebody who doesn't think about this all the time about exactly how fast we're doing it maybe in comparison to other phenomenon that they might be roughly familiar with.

是的，没错。

Yeah, yeah.

通常人们会这样形容：我刚才提到一飞秒，也就是十的负十五次方秒。

One way it's oftentimes described is that, so I mentioned one femtosecond, so that's 10 to the minus fifteen seconds.

你可以这么想：一飞秒相对于一秒，就像一秒相对于整个宇宙的年龄。

And so, you know, one way to think about it is that one femtosecond is to one second as one second is to something like the age of the universe.

明白了。

Okay.

这只是帮助我们理解这个时间尺度有多短的一种方式。

So that's just one way to kind of, you know, appreciate how short this time scale is.

从某种角度来说，这种速度真的超出了人类的想象范围。

It's really beyond the human mind to kind of comprehend it in some sense.

但即便如此，材料的运动确实发生在这样的时间尺度上。

But nonetheless, materials, are moving on on these types of these types of timescales.

这就带来了一个挑战，就像你拍照时，如果物体移动得特别特别快，照片就会模糊，拍不清楚。

And so it's a challenge, you know, for the you know, in the same way that if you take a picture, of something, you know, and the object is moving very, very, very fast, it will be blurred out and you'll get a blurry image.

人们后来发明了闪光灯，就是为了应对这个问题。

You know, people developed flash photography as a way to kind of get around this.

对吧？

Right?

闪光灯能极短时间地照亮物体，从而捕捉到这个过程的瞬间画面。

And so this kind of illuminates the object for a very short amount of time and allows you to take a snapshot of this process.

我们很多实验其实也是这么做的——在这些极短的时间尺度上，给这些结构拍下一张张快照。

That's kind of how, what we're doing in a lot of these experiments, we're kind of taking snapshots of these structures as they evolve on these very short time scales.

那我们就来聊聊这些快照，还有你们实际做的实验。

So on these, so let's talk about the snapshots and the actual experiments that you do.

你们会选一种觉得有意思的材料，对吧？

You get a piece of material that you find interesting, I guess, right?

这是个桌面大小的设备，还是个巨大的装置？

Is this a tabletop measurement device or is this a huge device?

你和你的学生、其他合作者是怎么实际开展这些实验的？

And how do you and your students and your other research collaborators, how do these experiments actually get done?

你们最终分析的是哪些数据？

And what are the readouts that you then analyze?

这方面有好几种不同的方法。

There's a lot of different kinds of approaches to this.

有些实验真的可以在桌面上完成，一个学生一天之内就能拍下这类现象的快照。

Some of them are really tabletop type experiments where you can, you know, a single student can really get to work and and in a day, you know, make a snapshot of of of these types of things.

但当你开始往更小的尺度推进，想看清原子级别的细节时，我们通常会用到更大规模的设施。

You know, when when you start to push the lane scales, when you start to really want to zoom in at the atomic scale, we tend to often make use of the larger scale facilities.

我提到过斯坦福国家加速器实验室。

So I mentioned the Slack National Accelerator Laboratory.

那是一个长达好几公里的直线加速器，原本是为粒子物理实验建造的，但过去几十年里已经被改造利用了。

This is a multiple kilometer linear accelerator that, you know, was originally built for kind of particle physics type experiments and has been kind of repurposed over the last few decades.

现在它本质上是一个能产生X射线脉冲的装置，专门用来做这种超快闪光成像。

And now it's essentially a source that allows one to create bursts of X rays to do exactly these types of this type of flash photography.

如果我没理解错的话，你做的不只是拍静态照片，还在拍视频。

And if I understand you're not just taking so called still photographs, you're also making movies.

那你能说说视频是怎么拍的吗？更重要的是，从视频里我们能学到什么，是静态照片无法提供的？

So tell us how we make the movies, but more importantly, like what can we learn from a movie that you might not get from a still photo?

是的。

Yep.

对。

Yeah.

没错。

Yeah.

其实拍视频的方式，就跟我们在电影院看的电影一样，对吧？

So the way essentially you make a movie in the same way that when we, you you make a movie in, you know, that you watch at the movie theater, right?

本质上就是把很多张静态照片按顺序连起来，一张接一张地播放，这样就能呈现出一个动态过程。

Essentially it's a whole bunch of still photos put together in a way that's sequenced such that when you kind of run them from one to the next, you, you turn this into a real movie that captures some dynamical process.

那这些电影里的主角是什么？

And who are the actors in these movies?

你们是在观察单个的碳原子或氧原子吗？

Are you looking at like an individual carbon or oxygen molecule atom?

还是在观察电子？

Or are you looking at the electrons?

我的意思是，这些照片的分辨率有多高？

I mean, how fine grain are these photos?

抱歉用了这么长的比喻。

Excuse the extended analogy.

有时候我们观察的是晶体，比如血红蛋白的例子，里面其实有海量的分子，大概有10的23次方个，它们整齐排列，形成一个肉眼可见的周期性晶体。

Sometimes we're looking at, this might be a crystal, like in the example of hemoglobin, this was many, many copies, you know, something like 10 to the 23 copies of these kind of individual molecules, all kind of locked, you know, together to form an overall kind of periodic crystal that, you know, you could see with your own eyes.

但现在越来越多的研究者开始往纳米尺度推进，比如让一粒纳米晶体、一个量子点正好处在X射线束的焦点上，然后对这个动态变化的微小物体拍下瞬间的照片。

But more and more people are kind of pushing to do these measurements at the nanoscale, where you might imagine, you know, for example, a single nanocrystal, a single quantum dot, in the focus of this of this beam, and, essentially taking a snap at the nanoscale of this evolving object.

太好了，太好了。

Great, great.

好的，所以我们有了这些内容，我觉得这非常有价值，虽然花了一些时间，但搞清楚你的能力边界很重要。

Okay, so we have a, this gives us a, I think a really good, it took some time, but it's important to have this groundwork on what your capabilities are.

那么接下来的问题是，你通常会提出哪些类型的问题？

And so I guess the next question is what are the kinds of questions you ask?

我知道我特别感兴趣，因为你提到过，我们在研究某些电子或光学设备的运行速度极限。

I know just I'm intrigued because you've had, you've made statements about, we're looking at the speed, the fundamental speed limits for some kind of electro, electronic or optical devices.

这听起来对我而言非常重要。

That sounds important to me.

那我们不妨从速度极限开始，你能不能详细说说？

So maybe let's start, tell me about speed limits.

没错。

Right.

是的，这个问题一直是我们长期关注的核心。

Yeah, this is something that has motivated us for a long time.

而且这确实又是一个从基础研究角度看特别有意思的问题。

And it's actually, again, one of these problems that is really interesting from a fundamental perspective.

而且，正如你提到的，这在技术领域也涉及许多非常重要的问题。

And also, as you mentioned, has a lot of really important kind of problems in technology.

对吧？

Right?

从基础角度来看，这其实归结为一些问题，比如：材料的切换速度能有多快？

So from a fundamental perspective, you know, this comes down to questions like how fast can you switch a material?

我们通常通过原子在晶胞中的位置来表示设备中的零或一。

How fast can you you know, we oftentimes encode a zero or a one in a device by the position of an atom within the unit cell.

所以你可以问：这种结构到底能多快地被切换？

And so you could ask questions like how fast can you really switch that structure?

对吧？

Right?

这最终决定了计算机或设备运行速度的上限。

This defines in the end speed limits that, you know, define how fast a computer can operate or how fast a device can function.

过去几年里，理论方面也出现了许多令人惊叹的进展。

And it turns out there's been a lot of kind of really interesting amazing kind of theoretical developments as well over the last couple of years.

本质上，人们正在尝试把一些在高中和大学初期学过的理念应用起来，比如热力学，那是用来理解平衡状态下过程的一种思维方式。

Essentially people trying to take ideas that, know, in high school and kind of early in college, you learn thermodynamics, which is the, you know, kind of a way of thinking about processes at equilibrium.

没错。

Yes.

直到最近，人们才开始把这些想法扩展到非平衡过程，也就是那些存在能量耗散的过程。

And so it's only very recently that people have been able to kind of extend those ideas to nonequilibrium processes where there's dissipation, where energy is lost in these processes.

而我们所使用的这种瞬态实验方法，可以帮助我们把握这些现象，直观地观察这些过程。

And so these types of snapshot experiments that we apply can let us get a handle on these things, visualize these processes.

接下来更重要的一步——这也是我们特别兴奋的地方——就是思考如何控制这些过程。

And then the next important step, and this is again where we're really excited, is kind of thinking about controlling these processes.

对吧？

Right?

如果你能以新颖的方式观察这些过程，就可以开始想：我该怎么设计这个过程？

So if you can see these processes in interesting ways, then you can start to think about, well, how can I engineer this process?

有没有办法控制材料在零和一状态之间的切换，从而降低能耗，或者加快切换速度？

Is there a way for me to control how I switch a material from a zero to a one state in a way that minimizes the energy cost or maximizes the speed by which it occurs?

我不确定这个问题是否恰当，但就计算机的速度极限而言，我们现在是不是已经接近瓶颈了？你我这些年在电子设备上感受到的那些速度提升，是不是就要到头了？还是说其实还有很大空间？当然，物理上有极限，但离那些极限还远得很。

So I don't know if this is a fair question, but with respect to speed limits of things like computers, are we at a position where you can tell me that we are pushing up against it and that the last forty, fifty years of speed ups that you and I have experienced just as a human in our electronic devices, is that era coming to an end or do you see plenty of headway so that, yeah, of course there are physical limits, but we have not gotten anywhere close to them.

总体来看，情况怎么样？

What's the general sense?

这其实挺有意思的。

It's quite interesting.

无论是在速度上限还是能耗方面，我们离理论极限还差好几个数量级。

We're many, many orders of magnitude away in both speed in both the speed limits and in the energy costs.

明白了。

Okay.

对。

Right.

举个例子，普通电脑一般工作在吉赫兹这样的频率上。

So, know, a typical, you know, a typical computer, you know, might operate at gigahertz like like frequencies.

对。

Right.

每秒大约十亿次运算。

That's 10 to the nine kind of operations per second.

对吧？

Right?

我之前提到的一些过程，其实跟信息存储设备里最基础的信息编码方式有关，这些过程理论上可以快到皮秒级别，也就是十的负十二次方秒，而不是现在的十的负九次方秒。

And there are processes that I was talking about, you know, essentially that are related to kind of really fundamental processes and how you encode information in an information storage device where you could push this to picoseconds, 10 to the minus 12 instead of 10, you know, instead of 10 to the minus nine.

那直接就快了一千倍。

That's a thousand times faster right there.

快一千倍。

Thousand times faster.

没错。

Yeah.

还有能耗，我刚才说的，其实更重要，尤其是在现在这个人工智能盛行的时代。

And then energy I mentioned is also, you know, in some sense even more important, You know, in this age of AI and so on, right?

我们所有人

We all

服务器机房，没错。

Server know that farms, yes.

服务器机房的能耗，跟核反应堆差不多，对吧？

Server farms, right, are equivalent of a nuclear reactor in terms of energy costs, right?

那你可以再问一个问题：完成一次操作、一个晶体管的切换，到底需要多少能量？

And so you can ask again this question of like, how much energy do you really need to like make a single operation, like a single transistor?

想象一下，把镜头拉近你的电脑，看看切换一个晶体管到底要消耗多少能量。

Imagine like zooming into your computer, right, and saying, well, how much energy do I really need to kind of switch that transistor?

对吧？

Right?

结果发现，我们离这个基本极限还差了好几个数量级。

And it turns out again, in this case, we are many, many orders of magnitude away from the kind of fundamental limits.

实际上，至少对于平衡态的过程，人们已经对这个基本极限有了比较清晰的认识。

It turns out people actually, at least for an equilibrium process, kind of understand pretty well what that kind of fundamental limit is.

这个极限被称为朗道尔极限。

It's something called the Landauer limit.

它实际上代表了擦除一个比特、也就是把一个1变成0所需的最小能量。

And it and it essentially it kind of encodes the kind of fundamental amount of energy you need to kind of erase or a one or, you know, take a one and transition it into a zero.

我们目前的实际能耗其实可以量化出来，真能算出具体数字。

And it turns out that the energy costs that we're at, you can put some numbers on these things actually, right?

现在大家热议的前沿是阿托焦耳级别，也就是你正在努力达到的水平。

You know, people are pushing a lot talking about something called at the Ato jewel frontier, which is where you're

没错。

trying Yes.

阿托是ATTO，对吧？

To Ato is ATTO, right?

ATTO比皮科还要小得多。

ATTO and it's even smaller than PICO.

比皮科还小，甚至比飞秒还要小，对吧？

Even smaller than PICO, even smaller than Femtosecond, right?

这可是10的负18次方焦耳。

This is 10 to the minus 18 joules.

这就是一个阿托焦耳。

That's one Atojoule.

对吧？

Right?

没错。

Right.

所以，很多人对研发能耗低至这一水平的设备充满热情。

And so, people are there's a lot of excitement about making devices that operate with energy costs associated with that of dissipation.

是的。

Yes.

结果发现，朗道尔极限所定义的基本极限大约是10的负21次方焦耳。

And it turns out that the fundamental limits defined by the Landauer limit are something like 10 to the minus 21 joules.

对吧？

Right?

这实际上是一个zeptojoule（仄托焦耳）。

So this is a Zeptojoule actually.

我给你说个事。

A word for you.

好的。

Okay.

如果我没理解错的话，这可是个好消息，意味着这些数据中心未来可能不需要消耗那么多能量——当然，不是马上就能实现。

If I'm if I'm if I'm hearing you correctly, this is super good news because it means that, these server farms, they might not, I mean, it's not tomorrow.

我知道这还只是基础研究，要转化成实际设备和功能还得一段时间，但我们现在已经有很大的余地了。

I understand that this is basic discovery and there's always time to translate it into devices and capabilities for engineers, but we have plenty of headroom.

就像你刚才提到的运行速度一样，我们在能效方面也有很多潜力，说明我们不该放弃研发超低功耗的设备。现在那些需要大电池或大型数据中心的系统，未来其实大有可为。

Again, just as you were talking about speed, we also have energy efficiency opportunities that indicate we should not give up on very, very low energy versions of the kinds of things that right now we have big batteries or big server farms that there's a lot of potential.

没错。

That's right.

另外还有一个有意思的地方，这些因素其实是相互关联的。

And actually one other interesting point, it turns out these things are interconnected.

速度和能耗这两个概念，实际上以非常有趣的方式紧密联系在一起。

These ideas of speed and energy costs are really kind of interlinked in really interesting ways.

我们之所以还没达到能耗的根本极限，其中一个原因在于，计算机本质上是动态的系统，就像我们刚才说的那样。

Like one of the reasons why we're kind of in terms of these energy costs, not at these kind of fundamental limits, it turns out, is because computers are intrinsically dynamical things like we're talking about.

它们是非平衡过程。

They're non equilibrium processes.

你知道，计算机不像热力学里常假设的那样是准静态的，那种适用于平衡过程的模型并不完全适用。

You know, computers are not quasi static in the way that we kind of think about thermodynamics is kind of applied on an equilibrium process.

所以当你想在极短的时间内完成动态操作时，实际上需要投入更多的能量来驱动它。

And so when you try to make processes happen dynamically over very, very fast time scales, it turns out you need to put more and more energy into them to drive.

你越想加快速度，就需要耗散越多的能量。

And the faster you wanna go, the more energy you need to dissipate.

因此，这两者之间存在一种内在的权衡，这一点我看得非常清楚。

And so there's an intrinsic kind of trade off between these things that is really I see.

那我们是不是可以把这看作是速度与能耗之间的权衡？

So should we think of it as a speed energy trade off?

这基本上就是那个权衡关系，对吧？

Is that basically the trade You off that's

有时候我们称之为一种速度与时间的不确定性原理，某种程度上确实如此。

sometimes call it, there's a kind of a speed kind of time uncertainty principle in some sense that kind of encodes.

明白了。

Okay.

不过这也有好处，因为一旦我们掌握了这种权衡，有些事情就不需要快，但必须高效。

So that's good though, because once that trade off, once we have it all under control, there are things that we don't need to be fast, but we need to be efficient.

而有些事情，比如打电话之类的，就需要快，哪怕多耗点能量我们也愿意。

And there are other things like our phone calls or whatever, that we need to be fast, but perhaps we would be willing to pay a price and efficiency.

没错。

Exactly.

就是这样。

That's right.

另外经常提到的一个问题是可靠性，对吧？

And then another thing that comes up often is reliability, right?

比如你存了一张孩子的照片，想十年后再看，那它可得还在才行。

You know, when you store information, you want that So to be when you, you know, save a photo of your kids and you wanna go back and look at it ten years from now, it better be there.

对吧？

Right?

那确实挺好。

It would be nice.

对吧？

Right?

所以，这种可靠性方面的考量，其实也包含在这个权衡里头。

And so so so this aspect of reliability also is is encoded in this kind of trade off.

你可以让系统没那么可靠，但换来的可能是更快、更高效；或者你选择更可靠，但代价可能是更高的能耗。

You can make things less reliable, but then maybe switch and operate more effect, more fast, or you can make them more reliable, but maybe then there are more energy costs associated with them.

这一切，都是Russ Altman眼中的一切未来。

This is the future of everything with Russ Altman.

接下来，我们会继续与Aaron Lindenberg对话。

We'll have more with Aaron Lindenberg next.

欢迎回到《一切的未来》。

Welcome back to The Future of Everything.

我是拉uss Altman，正在与来自斯坦福大学的亚伦·林登伯格对话。

I'm Russ Altman and I'm speaking with Aaron Lindenberg from Stanford University.

在第一部分，我们简单了解了如何用先进的技术测量材料的原子特性，这些技术在空间和时间尺度上都极其精准，让原本极快的过程看起来像慢动作，因为测量速度实在太快了。

In the first segment, we got a little tutorial on our ability to measure the atomic properties of materials with amazing technologies that give us the scale in both space and time that makes very fast processes look slow because the measurements are so fast.

在这一部分，我想问问亚伦一些大家比较熟悉的话题，比如太阳能电池、随机性，以及人工智能如何帮助他的研究。

In this segment, I'm going to ask Aaron about simple things like solar cells, randomness, and how AI is helping him in his work.

别忘了，本部分结束时我们会进入‘未来一刻’，我会快速提问，他也快速作答。

Don't forget at the end of the segment, we'll have the future in a minute, where I'll ask some quick questions and get some quick answers.

亚伦，这一部分我想从大家比较熟悉的东西开始聊起，比如电池和太阳能电池。

So Aaron, in this section, I wanted to start out with things that people are pretty familiar with, batteries, solar cells.

从材料科学的角度看，它们对你来说意味着什么？我们有哪些机会能更好地理解它们，进而改进它们？

What are they to you as you look at them as a material science and what are the opportunities kind of to make, to understand them better and then to make them better?

对，没错。

Right, Yeah.

比如说，我们拿太阳能电池来举个例子。

So, mean, think about let's take the solar cell as an example.

对吧？

Right?

这是一种把阳光中的光子转化为电能的装置，从本质上讲就是这样。

So this is a device which takes photons, light from the sun, and turns it into electricity, right, at fundamental level.

对吧？

Right?

但如果你放大来看，仔细想想原子层面到底发生了什么，假设你能以完美的分辨率观察这些过程的展开，你会发现，比如一个光子被材料——可能是硅或者其他材料——吸收了。

But if you zoom in, if you think about you know, what's really happening at the atomic scale, you know, and and imagine that you have, you know, perfect resolution to see these processes as they unfold, right, then what you actually see is something where, you know, maybe a single photon of light is absorbed in the material, silicon or or whatever that material is.

对吧？

Right?

它激发出一个电子。

It creates an excited electron.

这个电子开始在材料中移动。

That electron starts to move through the material.

它在材料内部不断与原子碰撞、反弹。

It's scattering and bouncing off of atoms within

激发态就是说它拥有很多能量，对吧？

the excited means it has a lot of energy, right?

它吸收了光带来的所有能量，不可能安安静静地待着。

It has absorbed all that energy from what the light and now it is not gonna just sit there quietly.

没错。

That's right.

对。

Yeah.

所以，归根结底，如果想让这个设备正常工作，就得把这部分能量拿走。

And so in the end, of course, if you want this device to work, have to take that energy.

对吧？

Right?

你已经把光能转化成了电能。

That energy, you've converted that energy from light into electronic.

对吧？

Right?

但现在它得想办法跑出来，还得在两个电极之间形成电流或电压，这样才能给设备供电，对吧？

But now it has to make its way out and there has to be a current or a voltage that is developed, right, you know, across two leads to actually power some some device.

对吧？

Right?

所以，如果你想搞清楚这个过程的终极效率——比如太阳能电池能有多高效？

And so so it turns out that if you want to understand the ultimate efficiencies of this process, right, like how efficient can you make a solar cell?

实际上，这里涉及的那些过程，时间尺度极短，空间尺度也小得惊人。

Can you can well, the processes that that come in here, the time scales and the length scales are amazingly short in time and amazingly small in in length scale.

对吧？

Right?

在皮秒级别的时间内，这些高能电子在弛豫过程中，一部分能量就会损失掉。

So on picosecond time scales, this electronic energy, some of that energy is lost as the kind of this kind of hot electron kind of relaxes.

对。

Right.

如果你想把这些能量捕获下来，就得想办法设计出这样的器件。

If you want to kind of capture that energy, then then you need to find a way to make that device.

你需要在这些极短的时间内把能量提取出来。

You need to find a way to extract that energy on these very short time scales.

对，趁它还没完全变成热能之前。

Yeah, before it basically just turns into heat.

没错。

Exactly.

一旦变成热能，这种能量就很难再用来做功了。从热力学角度看，理解这些过程能帮我们做两件事：一是寻找新材料，二是建立反馈机制，去探索各种经过不同调控的材料体系，通过原位实验实时观察器件的工作状态。

And once it's turned into heat, now suddenly you this is a form of energy that is hard to use for performing work from So a thermodynamics in the end, understanding, being able to see these processes allows one to, number one, think about new find new materials, work in kind of feedback loops where you, you know, can kinda think about a whole bunch of different material systems, where you've tuned the properties of the system in different ways, and kind of run these kind of operando like experiments where you kind of see the device as it operates.

你可以在原子尺度上测量这些过程，同时监测太阳能电池的实际工作情况，比如流过器件的电流大小。

You measure these kind of processes at the atomic scale, and you measure maybe simultaneously the operation of the solar cell, how much current is actually flowing through the device.

通过把这些数据关联起来，你就能把它当作一种设计原则，用来发现新材料、优化材料性能等等。

And you can kind of start to correlate these things and use this as kind of a design principle to kind of discover new materials, to optimize the performance of the material and so on.

是的。

Yes.

这让我想起你最初举的那个例子，氧气试图穿过蛋白质。现在你在我脑子里画出了一个类似的画面：电子正被引导着，连同其他电子一起，汇成一股电子流，朝某个正电荷方向流动，从而形成电流。

It does remind me of your very first founding example where the oxygen was trying to get through that protein and now you've drawn a very similar picture in my head where there's this electron and you're trying to kind of channel it into, with all its other fellow electrons, basically into a stream of electrons that are heading towards presumably some kind of positive charge to create the current.

我能想象，不同的材料在引导这些电子方面表现会有所不同，有的更好，有的更差。

And I can imagine that different materials would be either better or worse at kind of allowing you to channel those electrons.

你之前写过关于随机性的问题。

So you've written about randomness.

整个事情看起来挺随机的，但我觉得你说的随机，其实是可以在理论层面加以控制的那种。

How does that So this whole thing seems kind of random, but I think you mean random in a much more kind of theoretically manageable way.

能跟我讲讲，随机性在你的研究中扮演什么角色吗？

So tell me about the role of the randomness in your work.

没错。

Right.

就拿我们刚才说的太阳能电池来说，当光子被材料吸收并产生电子激发时，还会引发晶格的重构。

So actually, even taking that same example we just talked about, the solar cell, when that photon gets absorbed in the material and creates electronic excitations, you also, it turns out, drive reconstruction of the lattice.

原子会以复杂的方式发生位移。

The atoms move around in complicated ways.

对某些材料而言，这个过程非常关键，甚至有人认为，原子的这种动态运动能起到保护作用，让材料的效率以意想不到的方式提升。

So this is, again, some materials, this is kind of a really important process, and people think even that this dynamical motion of the atoms protects and, you know, makes the material more efficient in in interesting ways.

所以这个过程，在材料科学里我们通常认为是克服一个能量壁垒。

And so this is a this is a process that involves kind of the way the way we think about it in in material science is it's overcoming an energy barrier.

也就是说，存在某种能量障碍。

So there's some there's some kind of energy.

这就像是你在爬一座山一样。

It's it's just like when you're kind of climbing a mountain.

对吧？

Right?

前面有座山，你得想办法爬上去，这当然需要付出很多努力才能越过这个障碍。

And there's a a mountain in front You of kind of have to kinda, you know, try to make your way up, and it takes a lot of work, of course, to kind of overcome that barrier.

同样地，在原子尺度上，要克服这些壁垒、改变材料状态，也需要消耗能量。

In the same way at the atomic scale, there are energy costs that are associated with overcoming barriers and switching materials in that way.

而这个过程，有趣的是，它包含了一种明显的随机性。

And and this process, it turns out, is has a really interesting element of randomness.

我们有时候会用‘随机性’这个词来描述它

Sometimes we use the word stochasticity to

是的

Yep.

来描述这个过程。

To describe this.

我们通常这么想：材料可能会吸收太阳能，然后试图越过这个能量壁垒。

And and the way we think about this is that the the material is kind of you you know, you're maybe trying to take energy from the sun and kind of push this material over this barrier.

对吧？

Right?

但它其实一直在波动，不断尝试冲过壁垒，却又一次次回落。

But it's kind of fluctuating and it's, you know, making attempts trying to get over the barrier, but falling back.

结果发现，在原子尺度、纳米尺度上，这类波动变得极其重要，对吧？

So it turns out that at the atomic scale, at the nanoscale, these types of fluctuations become really, really important, right?

就像你把目光聚焦到室温下的材料上，

In the same way that if you zoom in on a material at room temperature, right?

如果你真能看到原子的运动，它们可不是静止不动的，而是在以复杂而随机的方式不断晃动。

If you can actually see what the atoms are doing, they're not just sitting there steady, they're fluctuating and in really kind of complex random ways.

所以，理解这种随机性如何影响效率，是一个非常关键的过程。

And so that randomness and understanding how that plays a role in the efficiency is a really important process.

这本身也是一个巨大的挑战，因为当我们做这些‘电影’——就像我之前说的——把大量这类过程的快照拼在一起时，我们通常希望以平均的方式来做测量，也就是重复实验很多次，采集大量快照再综合起来。

And it's a really challenging problem as well because when we do these when we make these movies, like I was describing, right, where we kind of make snapshots of of many of these processes and put them together, oftentimes, we we hope to we kind of do these measurements in kind of an averaging mode where you might like run the experiment many, many times and take many different snapshots and put them together.

我们在做这些实验时，默认的前提是，每次的过程都沿着相同的路径进行。

And the assumption that we make when we do these experiments, when we do those experiments, is that kind of the process is kind of always following the same trajectory every time.

但当你放大到纳米尺度时，这个假设其实并不成立。

And this turns out to be a bad approximation when you zoom in at the nanoscale.

这些材料本身就在不断波动，每次你进行化学反应，结果都可能不一样。

These materials are fluctuating, and each time you kind of run a chemical reaction, it might do something different each time.

因此，如何捕捉并处理这种变化，成了核心问题。

And so how to kind of capture that and deal with that is a really central

对，我特别能理解这一点，就像你看到那些河流的照片，摄影师把相机长时间开着，拍出来的水流看起来特别平滑，但任何真正仔细观察过河流的人都知道，水里全是漩涡和各种复杂的运动。

Yeah, find that to be very intuitive because you'd see these pictures that people take of rivers where they leave the camera on for a long time and the river looks like this beautifully smooth, like flow of water, but anybody who's actually watched a river in detail sees that there's all of these Eddies and there's all these other motions.

我听出来你的意思是，你得把这些细节都捕捉下来。

And what I'm hearing you say is you need to capture those.

所以我们现在明白，为什么你们要回到这些超高速的视频和图像了——因为实际上，你们已经能够拍到这种照片，来理解这种随机性，甚至区分哪些材料的随机性更有利于当前的应用，而哪些材料的随机性反而在拖后腿。

And now we can see why we're going back to these ultra fast movies and pictures that you're able to take because I'm gathering that it actually is within range to get these kinds of photos, so to speak, understand that randomness and how you can kind of harness it, or at least distinguish between materials whose randomness seems to be more favorable for the application at hand versus ones that are actually fighting it too much.

没错，河流的例子特别好，因为它引出了另一个关键概念。

Right, the river example is a really good one actually, because it also gives you, it brings us to this ideas.

我们常常用‘异质性’这个词。

We often use the word heterogeneity.

材料具有异质性，意思是如果你看材料的不同部分，它们看起来都不一样。

So materials are heterogeneous in the sense that if you look at different parts of the material, different parts look different.

对吧？

Right?

材料里有缺陷，有各种各样的复杂结构，你不能简单地假设所有地方都在做同样的事。当你做这种平均处理时，就会丢失关于动态行为的信息，就像河流表面看似平滑，实则暗流涌动一样。

There's defects, there are all kinds of kind of complexities, and you can't kind of just assume that everything is doing the same thing and everything, you know, when you make these kinds of averages, you lose information about the dynamics in the same way that a river has this amazing complexity.

刚才最后几分钟，我想问你一下人工智能。

So in the last couple of minutes, I just wanted to ask you about AI.

我知道你们团队也在用AI，我想知道它带来了什么样的影响？

I also know from your work that your group does use AI and I'm wondering like, what has the impact been?

这个影响大不大，还是只是边缘性的？

Is it a significant impact or is it somewhat peripheral to the work?

那有哪些机会呢？

And what are the opportunities?

你看好AI的发展吗？

Are you bullish?

是的，我觉得现在是个特别令人兴奋的时期，AI肯定会发挥核心作用，这点毫无疑问。

Yeah, I think it's a really exciting time and there's no question that there are ways in which, you know, AI is going to play a really central role.

我之前提到过那些衍射实验，比如用X射线照射材料，对吧？

You know, I talked about these diffraction experiments, for example, where you bounce a beam of X rays off of material, right?

这会在探测器上形成极其复杂的图案，我们就是靠这个来测量的，对吧？

This creates an amazingly complex pattern on a detector that we measure, right?

这些图像的生成速度非常快。

And these images are kind of coming in at very high rates.

光是分析这些图像，就是个海量数据的问题。

It's a massive data problem just to kind of analyze these images.

对吧？

Right?

所以大家一直在思考，怎么用人工智能在实验过程中提供实时反馈。

And so people are thinking a lot about, you know, how to use AI to kind of provide kind of real time feedback on an experiment, right?

比如在做实验的时候，能根据反馈调整某个参数，尤其是在研发新材料或新器件的过程中。

Where you might, as you're running the experiment, get some feedback which allows you to change some parameter, you know, as you kind of develop a new material or as you kind of build a new device.

这方面确实让人很兴奋。

So there's a lot of excitement there.

另一个有趣的例子，是我之前提到的工艺控制问题。

Another interesting example is in this idea of control of a process that I talked about previously.

没错。

Yes.

我们现在在研究如何在设备运行时实时监测，比如施加一个电压波形，通过电压切换来改变材料状态。

But we're thinking a lot about ways in which we can, for example, we're trying to measure some device as it operates, we apply some voltage waveform, some voltage switch that switches the material.

对吧？

Right?

你可以想到很多种不同的波形。

And you can think of lots of different kind of waveforms.

关键问题是，哪一种能最大程度降低能耗？

The question is, what is the optimal one that minimizes the energy cost?

就像我们之前聊过的。

Like we were talking about before.

我明白了。

I see.

这又是一个要在海量可能性中搜索的问题。

And so this is again a problem where you're searching through this massive phase space of possibilities.

根本看不出哪种协议才是最优的。

It's totally unobvious what the optimal kind of protocol is.

所以我们觉得，人工智能能帮我们解决这类工程难题。

So AI, we think can help us to kind of solve these types of engineering problems.

没错，我觉得这确实很合适，因为正如你所说，最优波形可能长得特别奇怪。

Yeah, that strikes me as a good match because the optimal waveform, as you say, might be an incredibly weird looking like waveform, right?

它可能不是正弦波或余弦波，而是一些特别古怪的形状。

It might not be a sine or a cosine, it's just some crazy.

AI 就能耐心地一个个试不同的方案，不断调整，最后说：就是这个了，看着挺随机，但可能就它最有效。

And that's the kind of thing that AI can patiently try different options and tune and say, okay, here's the one, it looks random, but this is the one that might work.

这正好弥补了人类的短板，因为我们并不总是擅长想到这么古怪的设计。

And it compliments human capabilities where we're not always the best at thinking of crazy shapes.

没错，就是这样。

Yep, that's exactly right.

太棒了。

Fantastic.

这确实很棒，也很令人兴奋。

Well, that's great and that's exciting.

在结束对话之前，我想进入我们称为‘一分钟未来’的环节。

And before we end our conversation, I wanted to move to our segment that we call the future in a minute.

我会问你五个简单的问题，你只需要简短回答就好。

And this is just where I ask you kind of five straightforward questions and you give me kind of short answers.

我想知道，你准备好进入‘一分钟未来’环节了吗？

And I'm wondering if you're ready for the future in a minute.

我们开始吧。

Let's begin.

好，来吧。

Okay, let's do it.

第一个问题。

First question.

对你来说，什么最让你对未来的前景充满希望？

What is one thing that gives you the most hope about the future?

我觉得，最让我兴奋的是基础科学与应用科学交汇处那些巨大的可能性。

I think for me, a lot of it is about the really exciting opportunities kind of at the boundaries of fundamental and applied science.

这些机会真的能让我们这个世界变得更好。

There's really exciting opportunities that can make our world a better place.

如果可以的话，我再补充一点：那就是我跟年轻人一起做这些项目时的体验，它们都是特别富有协作精神的项目。

And then maybe another quick answer to that, if I'm allowed, is is the you know, my experience kind of working with young people on these projects, these are kind of amazingly collaborative projects.