4.13.22-Build.nano-Marcus-Buehler

All right. So good afternoon or good morning, wherever you are. I think it's really late in the day. So hopefully we'll be a little more entertaining now. I show you a bunch of pictures of spiders and construction in nature. So hopefully you don't you're not scared of spiders of course.

So we'll talk about how materials are built in nature. You've heard a lot about amazing things we can do in nano, machine nano, and elsewhere at MIT. But we'll ask the question, what is happening in nature already? Can we learn from insects, perhaps, that build stuff, actually, in fact, from the nanoscale upwards? So we'll talk about biologically-inspired nanotechnology and how we can approach this topic today.

So we are interested in my lab in engineering materials across the scales. We call it material mix. We're looking at, from the nano to the macro. And that's why Vladimir and I had long discussions for many years on the exciting opportunities, actually, in scaling up nanotube make it really big. And this conference really has been an amazing show of exciting research we're doing here at MIT in making this a reality.

And I think, going back to the nanoscale, now, and asking how living organisms actually organize nanoscopic building blocks is a tremendous opportunity. We use theory, computation, and experiment. We combine them. And I only have a 10 minutes or so, so I won't go into much detail. But that is an exciting area.

We have, of course, tremendous advances in computation to really understand how quantum mechanics ultimately relates to macroscopic properties, and mold features, and designs across scales into the kinds of solutions we want in engineering. And particularly, we're interested in it's mechanical properties. And you saw in the previous slide, interested in fracture and impact-resistant materials-- making materials that are very lightweight, very tough, very strong. And you can imagine it has lots of applications in many different industries, from health care, to transportation, to construction.

So biology builds quite incredible nano stuff. and we always have to remember that we have invented nano we think. But actually, of course, biology has been building nanoscale machinery, in fact, for billions of years. And so that's what we're asking-- questions-- can we understand that? And how do we model these systems?

So this is a spider. I won't put it up for much longer, but this is really an incredible builder that we see in nature. And the spider is interesting, because a spider will basically build, from DNA machinery, things you can see in touch, like a spider web like you see here. And so that's fascinating sort of crossing in scales from nano to the macro.

It all begins with DNA and proteins, and this is what you see here. So a lot of the work in my lab is understanding how biology builds living materials. And that's oftentimes made from proteins. And so we have lots of activity.

Proteins are kind of like LEGO building blocks. And the amazing feat that the spider accomplishes, and many other organisms, including ourselves, is that we eat-- maybe you're eating a cookie right now, and your body will translate that cookie into hopefully muscle tissue. And that is this transformation that the spider does as well. So spider will eat the fly and takes these LEGO building blocks in your belly that you're just seeing the cookie and the food. And now, the spider will build this amazing structure, which, in this case, is a spider web.

In one example it might be bones, and tissues, and cells, and all sorts of energy transformation opportunities. So there's lots happening there, and we can't quite replicate this. So that's part of the current state of technology is that we can begin to understand these systems, but we can't really exactly replicate what nature does. And that's what we're trying to change.

Observing spiders building spider webs, like seen here, we really trying to understand, how does it work? When the spider eats the fly, and the genes are activated, and the spider silks are produced, how will they actually put together? And this is an example where construction could learn from this. We can learn how we can build complex structures like this without any scaffolding.

How do we direct design? What kind of structures emerge depending on whether the spider is hungry or has just eaten a whole bunch of flies, whether the spider is interested in mating or doing other things-- catching prey. So this is all in the structure in the architecture, and it's all kind of created by this natural neural network, the spider, by sensing the environment. Brian gave an amazing talk on sensors. And this is what spiders do as well-- they sense the environment. They decide what to do next based on what they feel mainly through vibrations.

We can use the natural design cues, sort of observing nature. Just like Brian was talking about observing patients and understanding how healthy they are, we can observe nature, and take design cues, and create things that are our own engineering creations according to certain design objectives. We can make structural materials, lightweight materials, architected materials-- I think you heard a talk earlier today. And these are actually synthetically created spider webs that look like spider webs, that have all the design features, but they're made by an artificial intelligence algorithm that has learned the mechanisms by which spiders actually construct these webs structures.

And we can then build computer models of these 3D architectures. We can mold them in all sorts of shapes and forms, and that's what's interesting. We're not limited to the corner in your house. We can make them into any kind of material shape that we want. And then we can build them.

We have 3D models. And we can then translate these three-dimensional models, of course, using 3D printing. And in fact that's an area where we use MIT nanofacilities quite a bit is to create, from the nanoscale upwards, these very small-scale architected materials.

As you can see here, that's more of a natural spider web. On the right-hand side you see engineering solutions that are taking design cues from spider webs, playing with irregularities and defects to create, for example, in this case, optimally mechanically-resistant materials, but not following exactly what the spider does. And this is where the ingenuity as an engineer comes in. Of course, we don't have to replicate exactly what the spider does. We can fine-tune the solution so that it meets certain engineering demands. But we can merge the two ideas.

Now, a little bit more on the nanoscale. So the spider webs are interesting, and they find lots of applications in structural materials. But how do they create it? So I mentioned proteins earlier, and we'll talk a little more about proteins. The proteins are sort of these incredible self-assembling structures. And this is where we can yet see another scale of observation now. It's not the spider making decisions by sensing the environment, but it's here, actually, nature is kind of taking its cause and organizing nanoscopic building blocks by itself and folding the protein, for example, into these three-dimensional structures purely based on what's encoded by the language of life DNA, language of protein or amino acid languages.

And so this question, of course, is a language question, where we're trying to understand how a language of protein sequences relate to structure and function ultimately. And this is very similar to other languages, like sound, music, vibrations. Brian talked about that. So that's an area in my lab as well. We're trying to understand how we can cross correlate different types of signals, and extract information from them, and understand why certain sequences of amino acids, for example, create hair, spider silk, cells, and viruses.

These are all made from exactly the same stuff chemically, but obviously they're very different. And so, for those of you who biologists, very obvious to you, but those of you who haven't thought about this, this is quite amazing actually. We have one basic building block of material, protein, and you can make all these different things out of them. You can make motors, you can make solar converters, you can make catalysts, and the list goes on and on. So our entire life really is based around this idea.

Now, we as engineers, we're trying to take this forward. So once we understand how these systems work, we can create synthetic materials. We can copy some of those cues. We can add additional materials. We can limit it by proteins. We can make them out of polymers, nanoparticles.

We can add minerals, like nature does as well, in conch shells. And we can create incredibly strong materials. And this is a lot of times what we look at, are properties like toughness. And so, in this case, you can see how we're trying to make materials that have internal structure-- architected materials, that have perhaps internal porosities or air. And by creating these internal structures, we can actually make it much more difficult for these materials to break, and hence we can use less materials. We save material with less expenses, less carbon footprint, and so on.

And so, the idea is essentially creating clever structures to make it hard for cracks to propagate. And that's a lot of times, in fact, what you want to do. You want to make sure that your crack doesn't just go straight through. But you want to make it difficult-- create a torturous path for the crack, if you wish.

Now, the last part I want to talk about is if you think a little bit deeper about the problem, it's not just looking at structures in nature, or looking at the spider building the web over a few days. It's actually the question of where these structures really come from originally. Evolutionary time scales-- you have billions of years of time scales during which this natural nanotechnology has evolved, and we see it today. But there are these different timescales of observation.

Do you have a material that's made like the spider web's being repaired, our skin is repaired, our bones are repaired, but the actual structures have evolved by themselves over billions of years-- how species interact with other species, and the ecology, and the environment. So those are all questions that really have remained intractable for many years. And now they become tractable using artificial intelligence, where we can actually begin to learn relationships.

Sort of the traditional way we looked at physical problems in a sense is that we created, let's say, a force field for an atomic simulation to understand plasticity in metals, like you see on the left-hand side. And we came up with laws-- quantum mechanical laws, force fields, and we can simulate these systems. And those are complex problems, but they're reasonably simple compared to a question of how structures emerge in spider species or in human skin or human cells. And for that we need a little bit better machinery, I think, as we're not going to find it in a simple mathematical equation, perhaps like Newton's laws.

We're going to need to have some other ways to do that. And this is where we use deep learning, especially attention-based networks, that can really understand long-range interactions-- long-range both in time and time and space-- that we have no idea about the origin of them. We observe them directly from the data, and then the model will tell us what's happening.

And so, a great example for these applications of this is protein folding. So this is sort of the baseline model of a simple problem in some way. We want to predict the structure, how the protein looks like, just based on DNA sequence. And this problem can be solved using such attention-based models. We can predict, very accurately, how the protein looks like and how it folds, and we can observe this from the simulation.

So we don't have to make the protein in the laboratory for tens of thousands of dollars or wait for months, but actually we can make it in the simulation, and we can do it very quickly. We can make thousands of these within a click of a mouse. And now we can assess them. We can predict properties.

And this is where the really exciting opportunity lie is not only predict infrastructure structure, but actually asking the model, what kind of properties do these proteins have? How stable are they? Are they biologically active? How do they interact with cells, with food-- working on projecting with the USDA in the food space.

And so, there's lots of opportunities where you can think about engineering these systems which today we cannot engineer. We can tweak them, but every iteration will take weeks or months to do. Now we can make billions of iterations within a few days, even on my laptop. I can run this code on my laptop once it's trained.

So this is very exciting. We can discover relationships that we had just not even an opportunity to see. We called it the banana curve. Essentially, very non-linear relationships, as you play with the sequences, language of life, and you begin to tweak these amino acid compositions-- a DNA code essentially. You can create some unique properties of these materials, which we had absolutely no idea.

You could not discover that, because each data point in one of those graphs would take, like I said, weeks or months to make, costs lots of money. So this is really the only way today computationally we can screen the space. And then we can limit ourselves to making maybe a dozen of these structures and then test them in the laboratory.

So lots of things we can do. We can engineer enzymes for example. And this is an area, if you're interested in sustainability, well, we might want to have a new enzyme to make concrete less carbon emissive for example. We want to make stronger materials-- we talked about that. We want to make maybe health applications.

So engineering existing proteins that nature has created that are in our body today, but changing them up. And so, we're not going to reinvent the wheel from scratch. Here, in this case, we're using excessive protein lysozyme as a model. It's a very well-studied protein in our cells. And we can change that. We can tweak that.

And so, these models are very accurate in saying, well, we want to have a little more beta [INAUDIBLE], a little more alpha helix, a little more this or that-- less stable, more stable. And we're really beginning to engineer as a partner with these biologically-evolutionary timescales. And again, this is a space where computation and deep learning in particular is really the only way to solve the problem. There's no theory. The data space is too large.

It would take us billions of years just to compute possible simulations for composite design. So there's no way we can actually do that either on paper or using conventional computing. So deep learning has really changed the way we can approach these problems. We can solve complex factor problems with deep learning now and predict how things break. That's a big deal.

For those of you in the industry, you might have a building, and you might want to inspect that building and understand, will this crack grow or not? And so, these kind of models are able to predict this just by taking an image of photos. So lots of work being done on image-based modeling, also working a lot on, as I mentioned earlier, biological inspiration. We're trying to figure out, can we take design cues from nature, like in this case leaves, and thinking about alternative sources for material construction.

We talked about silk already, but there's lots of other biomass in the world. Leaves, one of them-- wood, lignin, and so on. And I think you heard a little bit from others earlier today as well about this. It is an exciting space. There's lots of opportunities in thinking about future lightweight materials from leaves or wood and other biomass sources.

We can design them-- create these architected materials. And so, all of nature works with these really complex geometries, and now we can actually work with these geometries in three-dimensional space very easily using these deep learning techniques. This is something we just couldn't do a couple of years ago. And we can manufacture them. Thanks to MIT Nano, we can make them and test them.

So we can go from a concept, and designing gradients, and all sorts of design cues that improve, in this case, mechanical integrity, and lightheadedness, and toughness. We can then make inks that incorporate, for example, biomass, lignin, waste materials, sewage sludge, and other materials. And these liquid resins can then be assembled using UV light from the nanoscale [INAUDIBLE] into these very interesting architected materials, which have, now, features across all sorts of length scales and nanoscale upwards. And you can kind of see this here on this picture.

And the reason why we make them is we want to make them fracture resistant. We want to make sure they don't break us easily. So this is an example for a very tortuous way to prevent this crack from propagating. If you wouldn't have the structure in there, the material would be very brittle and very weak.

Last thing I want to show-- the spider, of course, is-- sorry, one more picture of the spider. One of the things that the spider's build materials-- and they're living essentially. The web is a living structure. And we can also go to that end. And that's an area called biomanufacturing, which is an area that is very interesting, where you're not just printing dead material, you're actually printing living organisms. For example, you can work with fungi, and fungi will grow mycelium, which is the root structure essentially in your fungi.

And these, together with designing the structure, the architect material now in a clever way, you can incorporate filament structures that form after you have 3D printed this material. And if you do the design right, you can incorporate really complex architectures in there. and again, you're using what nature already does really well, which is creating nanotechnology. But we're tweaking it in the way that we create materials that we want. We don't want to create just fungi. We want to create a material using the fungi technology or 0 fungi apparatus, but creating a packaging material, or a wrapping material for a building, or maybe a screen for your phone or your computer.

And so, all of these, again, require some really exciting intersections between modeling theory, deep learning, conventional physics modeling-- we use a lot of that as well-- as well as experimentation and manufacturing. And MIT is a wonderful place, where we can actually do all these things. And Nano has been really a game changer for us to being able to do this kind of work and explore how we can make the materials from less, we say-- from resources that are renewable and less expensive, or maybe waste that other people actually throw away. With that, thank you very much.