Things Fall Together: A Guide to the New Materials Revolution PDF

P R A I S E FO R Things Fall Together “In Skylar Tibbits’s ideal world, roads, buildings, and ob- jects are tingling, made of active materials whose parti- cles and units bind and unbind and recombine in mes- merizing harmony. There is little to no waste, an endless trove of new forms and solu...

P R A I S E FO R Things Fall Together “In Skylar Tibbits’s ideal world, roads, buildings, and ob- jects are tingling, made of active materials whose parti- cles and units bind and unbind and recombine in mes- merizing harmony. There is little to no waste, an endless trove of new forms and solutions, and the ability to test and perfect along the way. I want to go there.” Paola Antonelli, senior curator of architecture and design and director of research and development, Museum of Modern Art “In this book, Tibbits proposes a future where artificial intelligence is not an end in itself but an embodied feature of the products that we make. It is a future that is more humane precisely because of the shared tactility and ma- teriality of stuff. There is no doubt in my mind that the future of materials science lies in the development of the types of animate matter described in this book.” Mark Miodownik, author of Stuff Matters “Things Fall Together is a revolutionary book that helps us see into the future. Skylar Tibbits provides new design possibilities that rely on biological principles to activate materials into self-assembly. His pioneering approach is exactly what we need for Mars exploration and other space missions.” Dava Newman, Apollo Professor of Aeronautics and Astronautics at MIT and former NASA Deputy Administrator Things Fall Together A Guide to the New Materials Revolution Skylar Tibbits P R I NC ETO N UNIVERSIT Y PRE SS P R I NC ETO N A ND OXFO RD Copyright © 2021 by Princeton University Press Princeton University Press is committed to the protection of copyright and the intellectual property our authors entrust to us. Copyright promotes the progress and integrity of knowledge. Thank you for supporting free speech and the global exchange of ideas by purchasing an authorized edition of this book. If you wish to reproduce or distribute any part of it in any form, please obtain permission. Requests for permission to reproduce material from this work should be sent to [email protected] Published by Princeton University Press 41 William Street, Princeton, New Jersey 08540 6 Oxford Street, Woodstock, Oxfordshire OX20 1TR press.princeton.edu All Rights Reserved Library of Congress Cataloging-in-Publication Data Names: Tibbits, Skylar, author. Title: Things fall together : a guide to the new materials revolution / Skylar Tibbits. Description: Princeton : Princeton University Press, | Includes bibliographical references and index. Identifiers: LCCN 2020036342 (print) | LCCN 2020036343 (ebook) | ISBN 9780691170336 (hardcover) | ISBN 9780691189710 (ebook) Subjects: LCSH: Programmable materials. Classification: LCC TA403.6.T53 2021 (print) | LCC TA403.6 (ebook) | DDC 620.1/1—dc23 LC record available at https://lccn.loc.gov/2020036342 LC ebook record available at https://lccn.loc.gov/2020036343 British Library Cataloging-in-Publication Data is available Printed on acid-free paper. ∞ Printed in China 10 9 8 7 6 5 4 3 2 1 4Z Contents Acknowledgments IX Programming Matter 1 Computing Is Physical 15 Order from Chaos 41 Less Is Smart 57 Robots without Robots 77 Build from the Bottom Up 97 Design from the Bottom Up 117 Reverse, Reuse, Recycle 135 The Future of Matter Is Evolving 155 Notes 175 References 181 Index 193 Image Credits 211 Acknowledgments to those who have contributed to A S I N C E R E T HANK YOU this book directly or indirectly, everyone who collabo- rated with us on research, and the many people who in- spired and supported me throughout the process! This book would not have been possible without the amazing dedication from Princeton University Press, specifically Jessica Yao, Eric Henney, Chris Ferrante, and Madeleine Adams. Thanks for believing in the book, continuing to push me throughout, and helping turn ideas into reality. Thank you to Patsy Baudoin for your continual advice and guidance over the years on all of my books. Thank you to my MIT community, Terry Knight, Andreea O’Con- nell, Nicholas de Monchaux, Hashim Sarkis, Andrew Scott, Meejin Yoon, Ana Miljacki, Leila Kinney, Patrick Winston, Erik Demaine, and everyone in the Department of Architecture, the School of Architecture and Planning, the International Design Center, CAST, and many others. Thank you to everyone who contributed to the book: Tal Danino, Manu Prakash, Peng Yin, Fiorenzo Omen- etto, Rob Wood, Jennifer Lewis, Radhika Nagpal, Fabio Gramazio, Matthias Kohler, Marcelo Coelho, Casey Reas, Ben Fry, Daniela Rus, and Suzanne Lee. I’m forever grateful to our entire team of amazing researchers at the Self-Assembly Lab, including Bjorn Sparrman, Athina Papadopoulou, and my codirectors, Jared Laucks and Schendy Kernizan. Without our incred- ible team, the work would not be possible or anywhere X ACKN OwlED gMENTS close to as awesome! Thanks to our nonhuman material collaborators, because you are the true designers, keep- ing us on our toes and surprising us with realities we couldn’t have foreseen. Thank you to our human collabo- rators: Christophe Guberan, Hassan Maniku, Sarah Dole, Doug Holmes, Art Olson, Marcelo Coelho, Neil Gershen- feld, Tom Claypool, Gihan Amarasiriwardena, Gramazio Kohler Research, Patrick Parrish Gallery, ICD, Ministry of Supply, AFFOA, Steelcase, AWTC, Ferrero, Airbus, Carbi- tex, Native, BMW, Stratasys, Autodesk, Google, Tencate, and many others. Last but not least, to my parents, D and J, and my family, V and Z and R: thank you for all of your support over the years and putting up with this never- ending book project! Things Fall Together Programming Matter IN THE EARlY 1700S, the English carpenter and clockmaker John Harrison solved one of the most vexing puzzles that sailors faced at the time: how to calculate longitude while at sea. This challenge was so important for navigation— and had been so confounding up to that point—that the British Parliament offered a substantial cash reward to anyone who could find a practical solution. As trade in- creased, and ships sailed around the world with increas- ing regularity, it was critical for the crew to understand where exactly their ship was along the earth’s horizon- tal axis. Disrupted by the challenging conditions at sea, timekeeping and way-finding devices were inconsistent and unreliable. Consequently, navigation at the time was notoriously imprecise and shipwrecks were far too com- mon as a result of ships losing their way. While scientists and many others looked to astronomy, mathematics, or even magic in their quest to unlock an answer to the riddle, Harrison’s solution was amazingly simple and elegant. From wood, metal, and other simple material components, he crafted a “sea clock” that could keep reliable track of the time in relation to a given refer- ence location, which would allow sailors to calculate their position based on the difference from their local time. Earlier attempts at such clocks had been thwarted by the motion of the sea, changes in the environment, and 2 T H IN GS FA LL TOGETHER accumulating errors in the mechanical clockwork. But Harrison’s design, by accounting for the ways in which materials would expand and contract, enabled his mech- anism to adapt naturally to even the most minor fluctu- ations in temperature, pressure, moisture, and physical movement. As a master craftsman, Harrison understood that the dynamic and adaptive properties of his materials were the keys to a sea clock that could keep perfect time for long intervals, no matter the weather, the conditions of the sea, or the movement of the device.1 His invention became known as the marine chronom- eter, and it revolutionized not only sea navigation but also the way we think about materials and their ability to adapt in intelligent ways. Harrison demonstrated how material properties could be exploited to solve no- toriously challenging design and engineering problems. Since that time, similar material-based mechanisms have been applied to a number of novel devices that are abundant in our everyday lives. Thermostats, for exam- ple, take advantage of a bimetallic structure to regulate the temperature in our houses or maintain safe operat- ing temperatures in an engine. Orthodontic devices are made from Nitinol, a nickel titanium alloy that can move teeth into precise locations based on a response to body temperature. Lifesaving medical devices like stents use similar bimetallic structures to morph from one shape into another. This behavior has been “preprogrammed” in the material through heating and molding it at high temperatures. When a stent is placed in the body, for example, it is collapsed to fit through small spaces, and then activated by body temperature, allowing it to morph into the memorized shape and open the vessel. Yet this way of working with materials to craft elegant, simple, and transformative solutions is still largely con- tained to a few niche applications, and not widely used today. Since Harrison’s time, we have moved from a so-  P ROg RA M M INg M AT T E R 3 ciety that produced goods with localized crafts-based knowledge—one in which products and environments were intimately and intrinsically linked with material properties—to a system of industrially standardized mass production. The Industrial Revolution effectively ignored the intimate material knowledge of previous generations. Instead of taking advantage of the inherent material properties within wood or metal, for example, factories started to create standardized components that attempted to limit the amount of heterogeneity and differentiation. We attempted to standardize the trades and create repeatable outputs that did not rely on a single person’s skill set or knowledge in the craft—with some good reason: it was much more difficult to make a house out of logs and branches, or a stone wall out of geometrically unique elements, than it is to construct anything with repeatable components like bricks or two- by-fours. Similarly, at an environmental scale, humans shifted from an intimate relationship working with the earth and the natural forces of rain, sun, storms, tidal shifts, or sediment movement to a top-down, brute-force dictation through the use of machines. We could build anywhere, create land, dredge, redirect water flows, and artificially construct nearly any environment. Most of this standardization in manufacturing, construction, and land use was attempting to fight the dynamics of materials, minimizing their movement, and resisting the forces of the environment (gravity, temperature changes, moisture changes, vibration, natural disasters, and so on). The goal was to produce more, and to do it faster, cheaper, and better. This alienation from materials has only been exac- erbated in recent times by the rise of computing and the digital revolution. Digitalization and virtualization have tended to disconnect the average person from ma- teriality and led us to believe that creating something 4 T H IN GS FA LL TOGETHER “intelligent” means either a human being or a digital system with software/hardware that simulates human intelligence. But all of our own human and biological intelligence is ultimately built from simple materials, not computer chips or robotic components. We have lost touch with our appreciation for material intelligence. I often think of Harrison and his marine chronometer and wonder: if society were challenged with the same problem today, would we come up with the same ele- gantly simple solution? Hundreds of years later, simple devices like this can encourage all of us to take a fresh look at the way we design with materials, even as new re- search and technologies have us poised to surpass tradi- tional craft-based production methods. The emergence of digital fabrication technologies and the rapid advance of new research in synthetic biology, materials science, and other fields are making it possible not just to tap into, but also to create material properties in a new way, bringing the possibility of a new industrial revolution into view—a materials revolution. In this book, I offer you a glimpse inside this emerg- ing materials revolution, from my vantage point as founder and codirector of MIT’s Self-Assembly Lab.2 The Self-Assembly Lab is a group of architects, design- ers, artists, engineers, scientists, computer scientists, and many others who work on a variety of research topics from self-assembly to new material behaviors or new fabrication processes. Through this work, we explore applications in product design, manufacturing, construction, and large-scale environments. Sitting at the intersection of design, science, and engineering, we are an academic research lab that blends creativity with exploration, elegant design aesthetics with tech- nical performance, and the design principles needed to make those ideas reality. At its core, our work is moti- vated by the conviction that smarter, higher-performing  P ROg RA M M INg M AT T E R 5 products and sustainable environments don’t require complicated, expensive, device- centered solutions to achieve. Instead, we seek to use simple materials and their relationships with environmental forces to de- sign and create a more active, adaptive, lifelike world around us. In this work, we are part of a broader community of scientists, engineers, and designers across research and industry who are finding ways to design, create, and program physical materials that can do more than even Harrison could have dreamed. These materials can take in information, perform logical operations, sense, react, and much more. Unique behaviors often seen only in liv- ing natural systems—like the ability to correct errors, re- configure, replicate, assemble themselves, grow, evolve, and so on—can now emerge in innate material objects. At the Self-Assembly Lab, for instance, we have explored phenomena where physical components assemble and self-organize to build structures from objects, furniture, electronic devices, and even land formations. By under- standing and utilizing material capabilities, we can give simple materials and environments new functionality— going beyond mass production or even mass customi- zation, into material programmability with behavioral intelligence built into our products. As we will explore throughout this book, recent mate- rial advances are influencing various fields from robotics to apparel, furniture, medical devices, manufacturing, construction, and even coastal engineering. With novel material functions embedded within fibers, we are now creating clothing and textiles that can adapt to tempera- ture or moisture fluctuations and keep you cool or dry on the fly. Furniture and products can transform in size, shape, or function and assemble themselves after being shipped flat. Novel medical devices are emerging that can be quickly multimaterial printed to be customized to 6 T H IN GS FA LL TOGETHER the individual’s body. When they are inserted, they adapt to the person’s internal environment, expanding arteries or air passages without complex behaviors. At the largest of scales, a simple material like sand becomes a medium to promote the self-organization of new islands or coast- lines by tapping into the energy of the ocean. These and many other material-driven performances are coming into reality where simple products are becoming more active and static things are becoming more lifelike and playful. This kind of work ultimately requires a new way of col- laborating with materials in our broader environment, new relationships with our products, a different mind- set, and a fresh way of looking at the world. This book describes that new mindset through simple design prin- ciples that offer new ways to think about traditionally “static” mechanisms, products, and environments—as well as a different definition of what makes a product “smart.” The world is crying out for highly intelligent, ac- tive, and “smart” products, yet far too often we see smart products that are expensive, complex, battery-powered devices that are prone to failure. The principles in this book point to a different path forward. My hope is that they will make you stop and think, and wonder why some “smart products” might not be quite that smart after all. The aim is to show how we can take advantage of these hidden possibilities inherent in our physical world—and uncover a new relationship with materials, tapping into their built-in intelligence. What do we mean when we talk about programming ma- terials, and how has this reality emerged? We can start with a general definition: to program something is to create a set of executable instructions that an intended medium can perform or process. This is, obviously, a very general definition of programming—I’m using medium  P ROg RA M M INg M AT T E R 7 instead of computer, because, as I will explain, we can embed a program into any medium. Any time we perform a set of instructions, we’re executing a type of program. When we program materials, we’re embedding such in- structions into a physical material, such that the material can make logical decisions and can sense and respond to its environment. Thus, we can define a programmable material as a physical material structure that is embedded with infor- mation and physical capabilities like logic, actuation, or sensing. A related term, active matter, is used through- out this book to describe the expanded field of research- ers that are programming materials from the smallest to the largest of scales to create highly active structures that can self-assemble or physically transform.3 I will both describe the ways to program a material and explore the applications of its active behavior. In essence, these emerging material systems are all based on the ability to take simple material components, activate them with energy, and then have them assemble, transform, and create new physical behaviors. A diagram showing the key ingredients for programmable materials: materials, geometry, and energy to create physical transformations. Credit: Self-Assembly Lab, MIT 8 T H IN GS FA LL TOGETHER The idea of matter that can be programmed is a fairly old concept, but our understanding and realization of the idea has changed. People have been dreaming of pro- gramming matter since at least the Star Trek Replica- tor, with a machine that could instantly create anything.4 There are many early examples from science fiction that dream of infinitely small programmable material units that can be easily fabricated and set free to live, grow, and transform.5 This dream has a long history of over- promising and underdelivering, however, most likely due to the lack of material and fabrication capabilities, until very recently. Of course, from another perspective, there is a sense in which matter has always been “programmed.” Every- thing around us is programmed to sense something or function based on built-in information. The most ob- vious examples come from living systems: just think of our DNA, which encodes the instructions to build a human, or how a plant grows toward sunlight. But our everyday life is replete with materials that transform in this manner, according to built-in information. In addi- tion to complex living things, we can also see physical transformation in natural, yet nonbiological materials, or even synthetic materials that sense and respond to the ambient environment. For example, crystals that grow and morph, or chopped wood, which is no longer alive yet will still warp in response to changes in humidity, and plastics that expand or contract based on temperature. All of these materials are nonbiological, coming from both natural and synthetic systems, and all demonstrate lifelike, information-rich behavior. Craftspeople, master builders, or anyone with an intimate, hands-on relationship to materials like John Harrison are the forerunners of today’s “matter pro- grammers,” having long taken advantage of the inherent characteristics of materials. For example, craftspeople  P ROg RA M M INg M AT T E R 9 have used wood’s inherent properties when making furniture or building joints, ship hulls, or whiskey bar- rels, forging tighter and stronger joints by changing the amount of moisture in the environment in which they were made. Metalsmiths often use the expansion and contraction of metal based on temperature to make pre- cise and strong connections. Or engineers design a metal component for an engine to be able to operate uniformly with ever-changing environmental fluctuations. Textile manufacturers often use temperature and moisture to control the contraction of a garment to create finely tuned shapes and sizes. Today, however, new digital fabrication technologies can produce at speed and scale while also customizing material properties, giving us greater capabilities than ever before. Computing, fabrication, and materials share deep and long-standing links. The Jacquard loom, in- vented in the 1800s and considered one of the earliest examples of computing, read punch cards as an analog program to create intricate and beautifully complex woven textiles.6 More recently, not long after the mod- ern computer was born with the invention of the transis- tor in 1947, scientists at MIT first linked a modern-day computer with a milling machine in 1952.7 This paved the way for the first computer-aided design (CAD) tool in 1963 and today’s computer numerically controlled (CNC) machining with CAD to CAM (computer-aided manu- facturing) design workflows. This allowed computers to be programmed to run fabrication equipment that produced material parts.8 Nearly every manufactured product today is made in some way with this sequence of technology—CAD to CAM to CNC—from electronic devices to cars, clothing, buildings, infrastructure, airplanes, and even children’s toys. In the twenty-first century, we have achieved an ever-increasing level of sophistication with digital fabrication capabilities using laser cutters 10 T H IN GS FA LL TOGETHER and water jets, 3D printers, milling machines, industrial robot arms, and many other technologies. These are ac- quainting more and more people with the properties of materials and machines, as well as eroding traditional boundaries between design and fabrication, materials and information. This development in computing, fabrication, and materials research has led to the growing materials revolution and enabled programmable materials. Not only can we take advantage of the hidden abilities within materials to sense and transform, but we can customize the material with these rapidly advancing fabrication techniques. In the same way that we can alter the “in- structions” coded into DNA using principles of synthetic biology and other technological advances like gene edit- ing, we can now customize and produce complex com- positions of many different materials, from scratch. We can go beyond the evolutionary mutations that have led to specific genes or material properties to now fabricate embedded material codes. For example, we can now produce synthetic wood that responds to moisture with customized grain patterns that would never be found naturally, complex metal components that adaptively tune engines, high-performance composites that morph for aerodynamics, and multimaterial printed structures for smart medical devices. All of these examples have both geometric complexity and a diversity of material properties, designed according to a set of instructions for tunable and adaptive performance. This entire progression, from naturally evolved mate- rials to synthetically designed and generated materials toward fully programmable materials, can be seen in the simple example of the continuing evolution of fash- ion and footwear. We can trace the lineage from tradi- tional natural spun-cotton garments to programmably controlled textile production, and now synthetic fibers  P ROg RA M M INg M AT T E R 11 and high-performance textile products. More recently, the industry is turning the corner toward smarter gar- ments that have sensors and actuators embedded within the textile to inform and act on our body’s every move. These robotlike articles of “smart” clothing are quickly evolving from bulky garments with bulky devices to simple and smarter materials. The Self-Assembly Lab has worked with emerging companies, like Ministry of Supply, to develop highly active garments that can be made from material properties intelligently knit into in- tricate garments, functioning through materials, rather than complicated devices.9 Garments can become po- rous and breathe when the person is hot, or get thicker to insulate them when they get cold. The garment can morph to the person’s body shape and create the perfect fit, or change aesthetics for different occasions. Not only are we using novel materials, but we can now fabricate garments and other products in this new way, enabling active performance in everyday clothing. One might assume that a programmable material would be more electronic or robotic, less human, static, and less active—just sitting there waiting to be pro- grammed. But as I hope to show you, today’s digital capa- bilities have actually reintroduced the human perspec- tive and the craft of materials. Intimate knowledge of a material’s properties brings surprise and intuition back into discovery and invention. Programming materials is more about opening our eyes and designing in col- laboration with materials rather than forcing them into place. The ideas within this book seek to illuminate the sur- prising, yet still mostly untapped, capabilities of mate- rials through novel approaches to design and fabrica- tion. We will uncover ways to seemingly reverse entropy, create simple material “robots,” and program everyday physical objects or environments to come alive. We will 12 T H IN GS FA LL TOGETHER challenge the conventional idea that things fall apart: objects can get better with time, and we can program materials to become more active, to adapt, and to evolve on their own. We will question why so many objects and environments are designed to be static and why human- made things typically don’t have lifelike properties— for example, why they can’t grow, transform, or repair themselves. Why does a “strong” structure usually mean it requires more material, more rigidity in its composi- tion? Think of a plant or a tree, whose strength usually comes not from bulk or excess material, but from effi- cient distribution, flexibility, and the ability to adapt to different forces, to error correct, or regrow when needed. We will discuss the reasons why we’ve become so com- fortable with the notion of what a robot or a computer looks like and how it behaves, yet why that is rapidly changing. In this way, we arrive at the new reality of active matter. These ideas have taken shape after years of play, ex- perimentation, collaboration, failure, and some happy accidents at the Self-Assembly Lab, yet they go far be- yond our own work, crossing many academic disciplines and offering surprising applications in many different fields. This emerging field is based on blending rigorous science and engineering with creativity, play, and imagi- nation. Progress requires not only the solution of techni- cal problems but also the freedom to explore creatively and to take big risks, tackle big questions, and propose radical ideas. Accordingly, throughout the book we’ll ex- plore both concrete examples of technological advances being made today by talented designers, scientists, and engineers across different fields, as well as near-term thought experiments and possible futures. While this emerging field is rapidly growing and has shown prom- ising advances, it is still early days. We are in just the beginning of this materials revolution, and much of the  P ROg RA M M INg M AT T E R 13 potential impact or applications have yet to be realized. At this exciting juncture, I am hoping to create purpose- ful visions for the future to help catalyze these advances, inspire applications and new collaborations, and ener- gize the field of active matter. Computing Is Physical w H E N wE lOOK AT A C OMPUT E R , we see a machine that translates the physical to the digital. Our busy keystrokes are transformed into digital information (0s and 1s), making its way through silicon, transistors, chips, and drives into computation and communication. This ma- chine translates our conceptual and abstract thoughts into digital information through physical buttons and then patterns of 0s and 1s. It takes instructions, executes those instructions, performs logical operations, and magically transforms information from physical input to digital data. Consider now moving from the digital to the physical again: we can connect our computer to a printer to pro- duce physical paper documents or, nowadays, we may connect our computer to a 3D printer and fabricate physi- cal objects. Or we could connect a computer to a motor, and send code to tell the motor to move something. An ever-growing world of software and hardware eases this translation from the digital to the physical. Platforms like Arduino, Makey Makey, and Little Bits translate computer code into the physical environment, allowing anyone to easily make interactive electronic devices.1 Blurring the boundaries between physical/digital and human/ machine has exploded in recent years due to emerging fields like human-computer interfaces (HCI), haptics, 16 T H IN GS FA LL TOGETHER user experience and interface design (UX/UI), interac- tion design, physical computing, and many others. Yet, early forms of computation involved no such boundaries. Computers were only physical—in other words, there was no translation process when it came to the actual computation. For centuries, computing and calculating meant using an abacus, the knots of a quipu, or a handful of pebbles. Likewise, some of the first “com- puters” were actually the brilliant women at NASA who played a vital role in making complex calculations to help land humans on the moon.2 Early computing machines—from the Jacquard loom and Ada Lovelace and Charles Babbage’s Difference and Analytical Engines in the first half of the nineteenth cen- tury to Vannevar Bush’s Differential Analyzer a century later—were built with gears, pulleys, belts, or vacuum tubes, each meant to translate and calculate. We con- sider these machines “analog” when we compare them to today’s computers because they were noisy—not just in the literal sense of the word, but in terms of their ma- terial tolerance, faulty components, and mechanisms. The performance of these machines decreased as they got larger and more complex. The more gears and com- plex components, the more likely they were to break. The longer the machine ran—the more likely it was to break. As the length of time for a specific computa- tion increased, the accumulated tolerance and errors increased because the small physical errors in a gear or pulley, for example, could cumulatively make things become way off hundreds or thousands of cycles later. These early computers were affected by the temperature or moisture of the room, literally expanding and con- tracting, causing unintentional changes in performance or introducing maintenance errors. Early computers were so very material that the first computer “bug” was said to have been found inside Harvard’s Mark II,  C OM PU T INg IS PHYS ICA l 17 which was a literal moth trapped within the machine that caused an error.3 The concept of “digital” computation came into being around the 1940s, when the first notions of digital infor- mation and communication were laid out in one of the most influential theses of all time by Claude Shannon, showing how to communicate reliably with unreliable devices.4 Shannon’s work outlined principles for how to translate information from one place to another in a re- liable fashion. These methods can be applied to physi- cal, macroscale materials like mechanical computers, or translated to electronics in silicon today. Most relevant at the time, however, was how we communicated across noisy telephone lines in the 1940s. At the core of his theory was the idea that as information is sent from point A to point B, just as in a game of Whisper Down the Lane, er- rors can easily creep in. But if you create a system of error correction, like checks and balances, where you send multiple copies of the same information, and then cross- check the information on the other side, you can recon- struct the information and ensure it is perfectly accurate. This became the essence of being “digital” versus analog. Even the first computers that were considered “digi- tal,” incorporating Shannon’s principles of error correc- tion and programmability, were still made mostly from mechanical parts with the addition of electrical switches. The computers used by Alan Turing and the code breakers at Bletchley Park, like the wartime Enigma- breaking machine, Colossus, and John Von Neumann’s ENIAC (Electronic Numerical Integrator and Computer), were made with vacuum tubes and still relied on paper tape or manual cable switching as input.5 These digital machines were slow, difficult to program, and physically enormous, eventually leading to their replacement with ever more modern forms of computers, moving from gears and mechanisms to silicon and transistors with 18 T H IN GS FA LL TOGETHER the invention of the transistor at Bell Labs in 1947.6 These smaller and faster transistor-based computers eventu- ally led to desktop computers in the early 1980s. Our modern computers today maintain accuracy in compu- tation and communication even as time passes or as the distance increases between the sender and the receiver, in part thanks to Shannon’s work in the 1940s. However, as a result of the miniaturization and reliability of com- puter hardware, our idea of computing as inherently physical has simply disappeared. Alongside the development of computers, the first CNC machine at MIT in 1952 paved the way for computers to run personalized fabrication equipment. This led to the development of CNC milling machines, laser cutters, water jets, 3D printers, and many other tools that have become easily accessible for individuals or factories.7 Just as code became a new tool to design and create, code has enabled new forms of physical making to turn ideas into reality. In the past few decades, as the digital fabrica- tion and maker movements have exploded, we are seeing more and more physical making in fab labs, tech shops, and maker fairs, where people have become acquainted with the properties of materials. This development led in part to the materials revolution that is upon us. With the growing interest in materials simultaneously coinciding with new biological and synthetic material discoveries, it is now possible to create digital and programmable material just like their computer-machine counterparts. In recent years, a digital material has been defined as a set of material parts that can be precisely, yet reversibly, assembled and disassembled.8 This can apply to DNA, Lego building blocks, press-fit assemblies, or many other discrete building blocks. This definition of digital mate- rials doesn’t fully address material properties directly, however. The evolutionary path that enabled material programmability certainly involves the building blocks,  C OM PU T INg IS PHYS ICA l 19 but it is even more ingrained in the material’s properties. Programmable materials go beyond digital information in a material and are more than just the discrete assem- bly or disassembly of components. The extraordinary properties inherent in any physical material can contain information, sense a stimulus, and respond in sophisti- cated ways. This takes us beyond digital material to fully programmable materials. Today, we tend to associate computing solely with a laptop or a phone, losing sight of the material properties that make computation possible. Perhaps it is because we cannot see the inner workings of today’s digital com- puters: the components are so small, and the invisible electron-based processes on which they rely are im- mensely complex and hard for everyday users to grasp. Most of us don’t really understand how it works; we take it for granted, almost as if it were magic. Or perhaps it is the sheer ease and speed of modern computation. We type a line of code or send a quick email and forget that very specific material properties—of silicon in the transistor, of solder in the circuit board—and intricate geometries within the circuit are enabling this simple computation. As Moore’s law has increased the number and decreased the size of the components on a chip, a profound disconnect has grown between the physical computer that we type away at and the tiny components that enable computation under the surface. It is easy to understand why we were once fooled into thinking that the future is digital and not physical, as our world becomes increasingly virtual with more computing in our pockets, daily video conferences, and virtual reality surrounding us.9 Yet, as we all feel strangely less and less physically connected while being more and more digi- tally connected, we are realizing that the digital does not necessarily need to be less physical. The loss of physical connection and tangible realities from our devices has 20 T H IN GS FA LL TOGETHER actually created a renewed need for the physical. If we look at the COVID-19 pandemic, for example, where so many of our interactions were relegated to only digital connections, it became all too clear how important the physical and material aspects of our daily lives are. So as we become more and more digital, we actually need to be- come more and more physically connected with one an- other, with our products, and with our physical environ- ments. We are now realizing that the most elegant future is when these two worlds are symbiotically connected. Rather than virtual/digital and analog/physical being separate and isolated, we can create this blended and interconnected world of digital and physical materials. Computing is still inherently physical, just as it always has been. It still relies on precise material properties like conductance, capacitance, density, stiffness, opacity, and even the physical inputs from our own hands. Activating materials, at its core, is a rediscovery of these material properties and an extension of this essential fact. It seeks to imagine and develop different forms of computing that could take all of the awesome qualities of today’s invisible digital computing, yet utilize human-scale physical prop- erties that we can see, feel, touch, and play with. We typi- cally think of computers as static and electronic. It’s time we adopt a new perspective, moving beyond the bound- aries of our computer screens into a vision of computing that places the symbiosis between digital and physical at its center, and brings the capabilities of computers and materials together under one umbrella. How might this change the ways in which we view the world around us? Material as Computation It’s easy to find examples of physical embodiments of computing if we use the broadest definition of the term: transforming information from one form to another. For  C OM PU T INg IS PHYS ICA l 21 example, a simple mechanism that demonstrates infor- mation transfer from a code into a physical mechanism is the combination lock, which only opens with the right combination entered into the system. Material comput- ing imagines embedding some of the more complex properties of today’s digital forms of computing into ma- terial objects, such as modularity or reconfiguration (like copy/paste function), or replication and error correction (as with computer viruses). Conversely, we might even imagine one day incorporating the unique behaviors of the physical world (stretchability, bounciness, or trans- parency, for example) into the cold and static computers of today. These are just a few imagined realities that may seem far off but could be closer than we think if we can take advantage of material-based computing. One pathway to material computing that has emerged is called morphological computation, which studies the relationship between geometry, materials, and comput- ing.10 This form of computing takes advantage of physi- cal interactions and the interplay of objects to perform computational tasks. Some of this work emerged in the robotics communities, where you can imagine trying to build a walking robot with all of the complexities that might entail in both the software and hardware. Imagine designing something that could walk down a ramp on its own. You might try to design a robotic mechanism and software program that could give instructions on how to precisely walk down the ramp. You would need to program every single motor, gear, and joint at every single time step. This could quickly get very difficult. Alternatively, you could design a wheel or some type of multilegged structure that could step its way down the ramp with forward momentum, or create some wobbling structure that moves down the ramp by almost falling, catching itself, and then falling again. Most of the work would be done by the ramp and gravity, helping to guide 22 T H IN GS FA LL TOGETHER it from the top of the ramp down to the bottom. This is morphological computation—using physical materials and interactions to solve some of these difficult compu- tational programs. Creating the right physical material structure that wants to walk itself down the ramp solves many of the difficult software and electrical problems building the robot entails. There would still likely need to be software to do other things, but the material struc- tures can automatically “compute” some of the mechan- ics by figuring out how to walk down the ramp, simply through their interaction with one another and the forces in the environment (in this case, gravity). The notion of morphological computing could be ap- plicable more generally to say that materials and their geometric configurations—how they are put together, with different material properties, or different mechani- cal and geometric structures—can form the basis of ma- terial computing. We can see material computing emerge in various scenarios, some of them for robotic applica- tions, others just about information storage/processing, and others even about physical change, how something goes from one form to another. Material computing al- lows us to explore all of these scenarios, something that would be challenging with conventional notions of sili- con and electronic computing. As a framework to understand how different forms of material computing compare with traditional com- puting, the concept of “Turing completeness” is helpful. Named after Alan Turing, a computer is said to be Turing complete or computationally universal if it can simulate any other computer.11 More specifically, for something to be computationally universal it needs to show that it has both conditional branching (if this, then that) and a way to read and write memory, or to give input, change the program, and get a new output. From a computer science perspective, all real-world computers today are  C OM PU T INg IS PHYS ICA l 23 effectively universal, other than the fact that they don’t have infinite memory. When we talk about materials as a computational medium it is helpful to think about what type of computational significance they may have. Will these material computers be universal, will they be able to perform all types of computation, like any other com- puter or programming language? Will they be able to per- form other types of computation? Or will they be limited in their capacity? The lock mechanism that we discussed previously has a type of conditional statement—if this combination, then open the door, but if it is not the right combination, then do not open the door. The lock does not have the ability to change its own program, however. It can read the program but not alter it; therefore, the lock example would obviously not be a universal computer. Recently, DNA computing and DNA “hard drives” have been developed that compute much like our traditional computers yet have a number of unique qualities.12 DNA is considered computationally universal because it can have conditional programming—if these base pairs, then do this, but if these base pairs, do something else—and it has read-write functionality in its program.13 As we see in mutations and evolutionary biology, the genetic code can be written, changed, and replicated any number of times through the relationship of RNA, DNA, and the ribosome. These mechanisms allow DNA to be read and written and then perform specific tasks or create copies of strands. Researchers have shown that this proverbial code of life can be repurposed to store and retrieve other types of information, in base pairs rather than 0s and 1s. In his book Regenesis, molecular engineer and geneticist George Church explains how he took the entire text of his book, encoded it into a custom sequence of base pairs of DNA, produced synthetic strands of DNA with this code, and then read back the DNA through sequencing to extract the precise text of the book again.14 Church demonstrated 24 T H IN GS FA LL TOGETHER that DNA could be used like a hard drive for information storage and retrieval, and this demonstration has opened up further avenues of inquiry into how we might take ad- vantage of DNA’s unique properties for computing. As one scenario, there’s the potential that scientists could create DNA hard drives in the near future that correct errors, repair themselves, and self-replicate. In any one of Church’s test tubes there may be millions of strands of the DNA hard drive containing the text of his book. Upon retrieval, he gets not just one copy but mil- lions of copies. And, if a strand has an error in the code, that error can be detected and fixed either physically, by changing the order of the base pairs, or digitally, by com- paring it with the other strands when extracting the in- formation. The text of the book could even self-replicate, producing millions of copies of the same book, similar in many ways to a digital file of the book. This DNA copy of the book could perhaps survive thousands of years, like the DNA we extract from fossils—which is likely much longer than a paper copy or an electronic version. Just think of the fate of CDs, zip drives, VHS tapes, and cas- sette tapes, all of which remind us of the fragility of our contemporary technologies. Practically speaking, though, biological computing and information storage may not be seen as a replace- ment for traditional computing devices; instead, it is the vision to tap into the innate ability of biological materi- als while enhancing and augmenting their functionality through computation. If we can utilize biological material for computing, then we can allow biomaterial computers to functionally target diseases or cancerous growth. In many ways biological material already does this, but by augmenting their functionality through computational means, we can create networks of logic and sophisticated programs that wouldn’t naturally be found. For example, researchers such as Tal Danino have recently been able  C OM PU T INg IS PHYS ICA l 25 Programmable bacteria printed in precise patterns such that the bacteria can evolve and grow to become collaborators of the artwork. Programmable bacteria are also being used for cancer therapies. Credit: Tal Danino to do exactly this by programming synthetic biological material like bacteria to target and deliver cancer thera- pies directly within the body.15 In this way, they are using the functionality of biological material, but augmenting it with new information and agency to create powerful therapies that work with the materials of our body rather than placing foreign devices or drugs in our system. Other researchers have been building biological circuits that embody machine learning while tapping into the bioma- terial’s ability to learn and adapt.16 This approach takes naturally intelligent material from biology and combines 26 T H IN GS FA LL TOGETHER it with principles from artificial intelligence to evolve more sophisticated programs. These hybrid computa- tional and biological mechanisms could have incredible value in medicine and other biomedical applications. But these approaches also highlight a rapidly emerging ex- ample where a different model of computation was cre- ated, one that is purely biological-material-based rather than electron- and device-based, where we tap into the material properties to provide solutions that aren’t tra- ditionally possible in our classical forms of computers. As materials are increasingly used for more compu- tational capabilities, they may not always be designed to be computationally universal or the most efficient forms of computing. Some material computers may be slower than traditional computers, or they may have a smaller capacity than today’s supercomputers. Yet there are many other reasons why material computing may be a great way to compute. Designers, for example, have been exploring aspects of material computation focused mostly on a material’s ability to sense and respond dy- namically in its environment.17 Or as we saw in the case of DNA hard drives, scientists are developing methods where material computation can be more flexible, more adaptable, self-repairable, and self-replicating. For ex- ample, if we want to create a computational device to place within the body, it might make sense to trade super high-speed, high-capacity traditional forms of computing for a biological-based computational medium. Some of the traditional devices that we place in the body today—a pacemaker, for example—may start to seem archaic and brutal compared to emerging biological material devices and computers. These biomaterial computers exhibit particular properties that could make them far more robust than traditional devices because the medium al- lows them to adapt to fluctuations in the body and they don’t require traditional electronics, bulky enclosures, or  C OM PU T INg IS PHYS ICA l 27 dangerous failure modes. These are properties that are often missing in today’s version of electronic devices. These capabilities are not only found in biological ma- terials, however; many innate natural and synthetic ma- terials, such as metals, plastics, and even various fluids, error correct and adapt. Physicist and bioengineer Manu Prakash at Stanford University has demonstrated that air bubbles can com- pute digital logic by moving themselves around physical circuits and performing logical operations. Or similarly, Microfluidic bubble logic, a ring oscillator demonstrating cascadability in fluidic logic. For details, see Prakash and Gershenfeld 2007. Credit: Manu Prakash et al. 28 T H IN GS FA LL TOGETHER he showed that droplets of water can actually function as a synchronous and universal computer, rather than electronics and transistors in a circuit.18 Prakash ex- plains that computing is intrinsically linked to the laws of physics because bits are physical entities and that we can use computers to physically manipulate matter just as we can use computers to manipulate information.19 They can build all of the fundamental mechanisms of a computer by making any logical operation with hundreds of drop- lets that move around metallic traces that are roughly the size of a postage stamp. He says, “It’s not about manipu- lating information faster, it is about manipulating matter faster.” This work is particularly exciting because it could lead to a droplet-based computing medium where we can build up or change the physical material based on both information and the ambient environment at extremely fast speeds. A change of ambient temperature for ex- ample, could cause a transition in both the computing state and the physical output state, altering how informa- tion is visualized or how computing translates to human interaction through literally fluid interfaces. With the example of water computing, and various other forms of material-based computing, the digital realm is the physical realm, and our environment influ- ences the way matter computes as well as the physical form that it takes. The power required to compute with materials could be based on temperature or moisture, sun- light, sound, or other widely abundant and underutilized energy sources. In this way, perhaps material computing could one day offer alternatives to our energy challenges and our ever-increasing demand on electric or fossil-fuel energy sources. Or imagine the world’s supply of water as a computational medium. Our physical resources could provide ample computing capacity and storage space, and perhaps an abundant free flow of computational power. Maybe these dreams are a bit far-fetched at the moment,  C OM PU T INg IS PHYS ICA l 29 but if we can turn any material into a computational me- dium, it may change the way we see, interact with, and communicate with the world around us. Communication I’ve outlined the ways in which physical materials can compute; communication and connectivity are other key components of any computing platform. Beyond computing information internally, a medium can also pass information externally. One of the ways that digital electronics and computation have dramatically changed our world is through global communication. The fact that we can communicate with another person (or an- other device) from any point in the world to nearly any other point in a matter of seconds, without any physical proximity or wires, is an incredible achievement. With information and communication happening wirelessly, it certainly doesn’t feel physical. Yet information and communication are rooted in the physical. It’s more obvious how materials can communicate locally—two materials in physical proximity can liter- ally push, pull, or otherwise interact with one another, passing information, sensing, and activating each other, to create a communication platform. Think of two people tapping each other on the shoulder and pointing. It is a simple physical movement that communicates informa- tion from one person to the other. Or think of pool balls that bounce into one another, passing information in the form of a physical force, causing a physical reaction and translating that information into a new piece of infor- mation, the new state of the balls. This simple physical behavior can be translated into information by repre- senting each contact as a 0 or 1, or some type of symbol. An entire game of pool then could represent a poem, or a calculation, or a piece of music. In fact, it has been Physical building blocks that embody functional Boolean logic through geometry and the sequence of assembly. NAND Gate top [1,1] = 0 and bottom [0,0] = 1. Credit: Skylar Tibbits  C OM PU T INg IS PHYS ICA l 31 demonstrated that pool balls can actually embody digital logic and perform as a universal functional computer.20 In my master’s thesis at MIT, I developed a series of physical building blocks called Logic Matter, which em- bodied functional Boolean logic through its geometry.21 The way that you put the building blocks together would represent 0s or 1s as input; the building blocks would then guide or block future connections and result in a three-dimensional structure that represented a logic cir- cuit or some simple computation being performed. In this way, you could assemble the building blocks based on some predetermined code, or you could use them to compute something almost like a macroscale calculator or abacus, or even store information like a primitive hard drive. Perhaps these examples aren’t the most efficient forms of computing, but they demonstrate the transla- tion of information through local, physical, contact. Material communication is more challenging as we move to a global level, however, since a piece of material sitting on one side of a room may not be able to see, feel, or touch another piece of material sitting on the other side of the room. As a result, we might conclude that it’s simply impossible for materials to communicate globally or remotely. There are two perspectives to consider. The first is that global communication can be a hybrid “digital-to- physical” interface. Rather than thinking of physical material and digital devices as being at odds, imagine them working together. We know that materials sense the physical environment around them (moisture, tem- perature, light, pressure, and so on) and electronics easily communicate globally, so perhaps they can work together. For example, it’s difficult to get a Wi-Fi module on a circuit board to interact physically with the envi- ronment or move across the table. And it’s challenging to have a piece of raw material communicate wirelessly 32 T H IN GS FA LL TOGETHER with someone on the other side of the globe. So it makes more sense to use the Wi-Fi chip for communication, and let the raw material act as the physical interface, sensor, or actuator with the local environment. A simple sheet of wood, for example, could sense the amount of moisture in the environment and activate a Wi-Fi chip to send out communication about the local humidity throughout the day. In this scenario, we can set up a collaboration with both the materials and the electronics: humans act as de- signers, who program functionality, intention, and per- formance goals; the materials perform in their natural environment, acting as physical sensors and actuators; and the electronic components communicate globally. The collaboration among the three is seamless, taking advantage of their individual strengths. The second way to consider global communication is to realize that materials already communicate globally if we include line of sight where they can remotely sense, actuate, and influence another material. Recently, it has been shown that plants communicate through physical contact when leaves and branches touch; however, they can also communicate remotely by secreting chemicals in the soil or in the air or by growing fungi off their roots, communicating to their neighbors warnings of things like aphid attacks.22 Even more surprising, scientists at MIT have recently developed a way for plants and hu- mans to communicate whereby spinach plants can sense and detect explosive devices and then emit a fluorescent signal that can be detected with a device for humans.23 In this way, the plants can sense and communicate not only to themselves but also to the outside world. Ants, slime molds, and many other species communicate through distributed chemical signals rather than through audi- tory signals like most human communication. Humans and many other species also communicate through non- verbal/auditory pathways, as seen in sign language, body  C OM PU T INg IS PHYS ICA l 33 language, facial expressions, and physical contact. There are many ways to create both local and global communi- cation. In fact, before today’s modern forms of commu- nication through landlines or wireless signals, humans communicated through many remote techniques such as smoke signals and Morse code, which transferred in- formation across long distances and translated it into physical patterns, visual graphics, and ultimately com- munication. We can similarly utilize these simple, yet so- phisticated physical techniques for local and global com- munication in the design of material communication. Recently, there have been a few examples where researchers have developed remote communication through simple material behaviors. A team of scientists created a printed lattice structure that functioned as a morphable antenna. They could change the resonant frequency of the antenna by changing the temperature of the environment, therefore controlling the sending and receiving of information between two points across different frequencies.24 As an everyday example, think of a speaker and a microphone: they are the same mecha- nism but with reversed functionality—a speaker turns electricity into sound, and a microphone turns sound into electrical signals. Both the speaker’s sound and the microphone’s signals happen as a result of a physical material—a thin membrane that vibrates to produce or receive sound waves. The vibrating physical material can produce sound, transmitting a signal from one location to another, which is received through the vibration of another material. To continue the sonic analogy, a vinyl record conveys information through physical geome- try, bumps and grooves, which store the information of a song and translate these geometric complexities into beautiful sound through physical, vibrating speakers. As a thought experiment, if we had a thin film that could be activated by infrared (IR) light, causing it to 34 T H IN GS FA LL TOGETHER undulate, we could use this to create semiglobal com- munication. We could place a piece of this material on one side of a room and subject it to pulses of IR light, on/off patterns, causing it to move. This would behave like the speaker, creating a vibration. On the other side of the room we could place another piece of the same material, acting as the microphone, which would receive the vibration of the speaker. This would cause the micro- phone material to vibrate, which could then be read out as information. The frequency of the vibration in the microphone material would relate to the activation from light on the speaker material. This hypothetical example demonstrates how two physical materials could commu- nicate and create a wireless IR gauge showing that mate- rials may be able to communicate in unusual ways and translate information from one location to another. This example is a reminder that our current means of infor- mation and communications are actually quite physi- cal, and with a more careful lens turned to investigat- ing the inherent capabilities of materials, we can create physical-digital hybrids. These hybrids serve as models of computing and communicating platforms that create elegant solutions by tapping into material properties. Computing without Efficiency Some computer scientists argue that the fundamental aspect of computing comes down to what can be (effi- ciently) automated.25 But what if computing were not only about efficiency, automation, or optimization? What if computing were about inefficiencies, creativity, and all of the other messiness in our lives? Why did computing get relegated to being only pragmatic, optimal, or efficient? The first applications of desktop computers were thought to be spreadsheets and accounting, but luckily that wasn’t the end. Personal computers quickly came  C OM PU T INg IS PHYS ICA l 35 along that began to address all of our weird and person- alized interests through today’s music, photography, gaming, video editing, apps, and so on. Perhaps the most interesting emergent domain in computing today is the creativity and aesthetics of artificial intelligence (AI) and machine learning. For example, Google’s image gener- ator created unimaginable scenes, ones that humans could hardly dream of creating, rather than optimizing or solving problems. We should be promoting computa- tion for creativity and its generative qualities not only for optimization. Efficiency and automation invoke speed, performance, and optimal solutions. At the other end of the spectrum, however, there is creative computing, in which we can explore the intersection of the creative arts, design, and computing. Seymour Papert, Mitch Resnick, Muriel Cooper, John Maeda, and many others have pioneered playful computing and creative coding in an effort to promote curiosity and exploration.26 In Processing, a software language developed by Ben Fry and Casey Reas, each file is called a sketch, and it is pre- cisely the idea of creatively sketching with code that is desired—making mistakes, being surprised, discovering something, rather than optimizing or trying to find a sin- gular, efficient solution, as the primary goal.27 Computa- tion itself can be used as an idea generator, just as a quick sketch allows for misinterpretation and creativity, or as watercolor makes it possible to blend and blur the lines between detail and fuzziness. Creative coding allows for new ideas to emerge. Similarly, by computing with physical materials, we can embody these inefficiencies of computing and put them to creative use. Materials have unusual traits that may not increase speed or capacity. They may not be faster than silicon at flipping a bit and may not have unlimited storage capacity (although DNA hard drives might get pretty close in some aspects). Materials are 36 T H IN GS FA LL TOGETHER traditionally nuanced, weird, and dynamic; we should tap into these traits and do something interesting with them. With programmable materials, we can make materials active: we can build a simple set of building blocks that can be combined to make materials behave in surprising ways—sometimes with higher performance, sometimes with storage or efficiency in mind, and at other times with creativity, playfulness, or lifelike qualities in mind. Materials can also embody polymorphism to its fullest. Polymorphism, a term used in both biology and com- puter science, is about a single code that can manifest as many different outputs.28 This creates a very interesting conundrum: if you have a linear set of instructions (like an algorithm written in code), or a set of physical behav- iors in a material (sense, actuate, fold, curl, twist), or a series of choreographed rules for a dance, or a procedure in Brian Eno’s algorithmic music, or the instructions for realizing a Sol LeWitt artwork—how is it possible that unexpected things can emerge?29 If the rules are well understood, and the operations or procedures among those rules are clear, then how could anything surpris- ing happen? Polymorphism is one way to explain that. Polymorphism is the ability for difference to arise from the same embedded information. Polymorphism insists that different results can emerge, even with the same information and the same procedures. And this is the surprising and awesome place that materials can take us in computing. As a thought experiment, imagine that we can create a material composite that has certain parts activated by moisture and other parts activated by heat, in an al- ternating pattern, heat-moisture-heat-moisture-heat- moisture. When the moisture zones are activated, they will bend so that the overall structure makes a square with four corners. If the material is heated (without mois- ture), however, the previously straight segments will curl  C OM PU T INg IS PHYS ICA l 37 in the opposite direction, creating a circular shape. And if we activate this material structure with hot water, it will do something completely different—the moisture zones will bend downward, the heat zones will curl upward, and you’ll get a sinusoidal-type shape. For example, if we put the wet material on the bottom and the heated material on the top, it will behave one way with the sun shining from above, and in a completely different way if you set it in a puddle. Both the physical structure of the material (where the materials are in relationship to one another) and the changes in the environment (different amounts of energy, different times of activation, and so on) will conspire to create different results even though we are using the same material composition. The exe- cuted program comprises the material’s geometry and the patterned placement of the material properties. This leads to a logical operation for the material: If water, then do X, or if heat, then do Y, and if water AND heat, then do Z. If we make even more complex material structures based on this with fluctuating physical environments, we could get extremely sophisticated, complex, and surpris- ingly brilliant material computers that aren’t just about efficiently solving a problem or automating a human task. They are about discovery, design, or performance. Another way to think about creating differentiation through simple rules is to consider the physical embod- iment of chemical morphogenesis, a topic that fascinated Alan Turing and was the focus of his work just before he died in 1954.30 Morphogenesis is the differentiation of chemical, biological, or physical processes, which can grow from homogenous building blocks (such as cells) to complex patterns like zebra stripes, cheetah spots, or all of the complexities of us humans. These patterns emerge from instabilities triggered by random disturbances in the homogeneous system, activating and deactivating the system. The instabilities and random fluctuations 38 T H IN GS FA LL TOGETHER of our physical environment combined with simple material-based rules can lead to incredible complexity. Entire fields of study such as chaos theory and complex systems focus on exactly these types of complexities that emerge from simple rule sets. Complex systems are often defined by very important initial conditions (such as the genetic code or programmed behavior) and a feedback mechanism, both of which can lead to dramatically dif- ferent outcomes from even simple changes in the envi- ronment. The common thought experiment is a marble that is dropped precisely on the tip of a mountain. Very minor changes on one side or the other side of the moun- tain ridge will lead to dramatically different paths for where the marble will end up at the bottom of the moun- tain. Even with the same marble and the same mountain, minor changes can create very different outcomes. Our physical environment is a perfect petri dish for a computational material medium to flourish, offering complex and creative outcomes. This possibility is ripe for use in design and engineering. Designers can imag- ine a product made of coded-material rule sets and with random fluctuations in the environment: a new pattern or structure could emerge—like a chameleon— ever- changing yet always made of the same physical material. In cellular automata—a classic computational model that has white and black colored cells with embedded rules for their color based on their relationships with their neighbors—researchers have studied how patterns emerge from simple sets of rules. In his book A New Kind of Science, Stephen Wolfram outlines four types of pat- terns: growth/death, where patterns either take over or completely die off; repetitive patterns endlessly mak- ing the same thing; chaotic patterns that have no dis- cernable structure and look like pure noise; and, most interesting, patterns that oscillate between chaos and order.31 I believe it is this fourth type of pattern that we  C OM PU T INg IS PHYS ICA l 39 should strive for when working with programmable ma- terials. We want to build active systems by embedding simple rules into materials, enabling them to transform themselves, interact with one another, and enhance their function, performance, or aesthetics. We want them to enable truly unique and surprising results even from simple embedded rules. These capabilities shouldn’t be just repetitive tasks, or behaviors that eventually die off; nor do we want them to be random glitches or spasms of behavior. We want interesting and useful behaviors, much more like the patterns of human or natural sys- tems that are both chaotic and repetitive in all of their delightful and surprising ways. Order from Chaos MOST MANMADE THINgS seem to fall apart. Beautiful new products eventually deteriorate, buildings need main- tenance, food rots, clothing rips, cars eventually die. It is sad how things typically move from order to disorder. Yet, nearly everything about the natural world starts out by growing from seemingly nothing into something amazingly complex, functional, self-repairing, and self- replicating. All of life strangely appears to resist this, building order from disorder. As the second law of ther- modynamics states, the entropy of an isolated system must always increase (or sometimes remain constant). Yet many things seem to go against entropy and scien- tists sometimes even describe life as the fundamental pursuit of negative entropy. Most systems that appear to lose entropy and create order from disorder are not isolated systems; there are fluctuations in the en- vironment or the amount of external energy input, or we can even design the components to promote their self-organization. Globally, however, the environment surrounding these systems must increase its entropy at least the same amount, in which case the global entropy never decreases. In everyday terms, what is amazing is that we can see spontaneous patterns emerge in crys- tallization, or swarms of insects, or a single cell growing into a human that can regenerate and self-heal when injured. The question is why this happens and how to utilize it. 42 T H IN GS FA LL TOGETHER Active matter can allow us to tap into these incredible behaviors often only believed to be possible in biological or chemical domains. We can imagine a near-term future when everyday physical materials will build themselves, grow, adapt, transform, and improve over time. The prop- erty of self-repair, for example, no longer applies solely to living things; it is starting to be embedded within our built environment. Researchers can now create self-repairing materials like self-healing concrete, polymers, and com- posites.1 As we will explore throughout this chapter, many examples are emerging with animate material behaviors in traditionally static material systems. There are more precise descriptions of entropy that can help explain this complex phenomenon and the pro- cess of growth and decay. The most common illustration of entropy involves two chambers with different tem- peratures, one hot and one cold. When the two chambers are allowed to mix, the entire system will move toward an equilibrium temperature. In 1867, James Clerk Maxwell described a thought experiment as a hypothetical viola- tion of the second law of thermodynamics that describes entropy, now called Maxwell’s demon, where a small door could be opened and closed quickly to allow only the fast molecules (in this case, hot air) to enter into a second chamber. This process would separate the two chambers into one hot and one cold, effectively reversing entropy by creating order and seemingly violating the second law. In Maxwell’s chamber, after the two sides are allowed to mix with hot and cold molecules bouncing around the tank, it is theoretically possible that for one moment in time, all of the hot molecules would be momentarily on one side of the chamber and all of the cold molecules on the other. It is possible, yet incredibly unlikely—because the most likely scenario is that the two types of mole- cules will move to be a mixed distribution of hot and cold molecules at any given time. So in order to make the op-  ORD E R F RO M CHAO S 43 posite happen, we need to coax the molecules to make them comfortable in the seemingly impossible condition where they are separated. One simple example where we can manipulate the local environment (seemingly violating the second law) would be to use the vertical axis to separate temperature zones. We know that hot air rises, so we know that the hot air molecules will rise to the top of a space while the cold air molecules will tend to stay toward the bottom. If we were able to design the chamber in just the right way (making it very tall, for example), or if we were able to rotate the chamber in different orientations, or if we could add a heat source outside the chamber (all of which emphasize that this is not an isolated system) we could promote the hot and cold molecules to separate above and below one another. Another way to describe entropy is a system seeking equilibrium: any object or system, will tend to move to- ward a position that is the lowest-energy state. For ex- ample, a ball will roll down a hill, moving from a higher position with greater potential energy to a lower one with less, and then naturally come to rest in the lowest-energy state. Yet, as in the example of a Rube Goldberg machine, you can design the “hill” (or machine) in such a way that the ball can do extraordinary things like fly through the air, or perfectly balance on top of objects, fall into precise locations, or set off entire chain reactions, seemingly on its own. Or if you imagine dropping a number of pieces of paper, they will likely scatter into a random pile on the floor. But if you could fold each piece into a uniquely de- signed paper airplane, you might be able to drop them and have them fly into very specific and surprising patterns. I like to think of this as designing the system so that the lowest-energy state is a very useful and interesting one. In this way, the equilibrium state does not necessarily mean a destructive, degenerative, or disordered state. Instead, think about how we can use the concept of entropy to 44 T H IN GS FA LL TOGETHER design non-isolated systems that move themselves to- ward states that get better with time. The main challenge is how to create the lowest-energy state to become the most functional, ordered, or even surprising condition. The phenomenon of self-assembly can be described as individual components that spontaneously assemble or- dered structures without human or machine interven- tion. This is the underlying principle within biology and chemistry that accounts for everything from how humans are built from DNA or how ice crystals are formed from water to how planets are formed in the solar system. We can think of self-assembling systems as individual parts moving toward a final configuration, or an equilibrium point. This is slightly different from self-organizing sys- tems, in which components don’t necessarily move to- ward equilibrium, but can move between multiple states, oscillate, and may never come to rest in a final configura- tion. Schools of fish, flocks of birds, sand ripple patterns, and traffic jams that grow, organize, change, and dissolve repeatedly are all examples of self-organization. In biology, self-assembly is the primary approach for construction. There aren’t many other techniques to build with biological material, given how small and com- plex the environment is. DNA assembles itself through complementary base pairs, proteins fold themselves, and higher-level molecular structures come together to make functioning capsids like viruses or other biomolecular structures where order and functionality build them- selves. Recently, scientists have been able to tap into this phenomenon and design DNA structures that can self-fold and self-assemble into nearly any two-dimensional or three-dimensional shape. Peng Yin’s group at Harvard’s Wyss Institute has pioneered an approach now called DNA origami to use DNA as a building block, much like a Lego, for construction at the nanoscale.2 With this technique,  ORD E R F RO M CHAO S 45 many other researchers around the world can now pro- gram DNA with specific base-pair sequences to promote self-assembly, forming precise nano-architectures from 2D and 3D bricklike modules. They have demonstrated hundreds of 2D shapes such as letters of the alphabet, emojis, and symbols. Three- dimensionally, they have also created hundreds of demonstration objects using a canvas of DNA building blocks to create volumes like DNA spaceships, drug delivery capsules, and other geometry. DNA origami structures formed from the self-assembly of programmed DNA strands forming precise nano-architectures. Credit: Wyss Institute 46 T H IN GS FA LL TOGETHER This approach may very well translate into bottom-up assembly and manufacturing approaches for various applications of nanotechnology in the near future. To create self-assembling systems, we first need to understand why self-assembly works and how things move toward equilibrium. Then, we need to find ways to take advantage of this equilibrium seeking to promote order out of chaos. At the Self-Assembly Lab, we define self-assembly through its three core ingredients—energy, geometry, and interactions.3 The first ingredient we look for in any self-assembling system is energy. The energy imparted into a system needs to be just right to promote assembly, and the amount of time or frequency of this energy activation needs to be designed to promote the structure to easily find equilib- rium. It is also important to think about where the en- ergy is coming from—where you can find abundant energy sources, whether that is vibration in a system, or changes in temperature, pressure, or wave energy—these are often common and overlooked sources of energy that we can work with when designing a self-assembling system. A number of years ago, the Self-Assembly Lab devel- oped a project in collaboration with the molecular biol- ogist Arthur Olson, around a macroscale and tangible demonstration of self-assembly.4 We based the project on a poliovirus, a tobacco plant virus, and other bio- molecular structures that naturally self-assemble. We produced a number of glass containers that had simple plastic parts inside. When you shake the container, the parts inside spontaneously come together, building order from disorder. If you do not shake the container hard enough, the parts don’t have enough energy to come together. Conversely, if it’s shaken too hard, the parts will separate themselves. If you try to shake it with intention, almost like a puzzle or game bouncing them into place, it is often worse than if you shake it randomly. The trick is  ORD E R F RO M CHAO S 47 A tangible demonstration of self-assembly based on bimolecular structures like the polio- virus or a tobacco plant virus. When the container is shaken with the right amount of en- ergy, the parts inside spontaneously come together, building order from disorder. Credit: Skylar Tibbits, Arthur Olson, and Autodesk to find just the right amount of energy so that the parts can move around freely, find one another, and connect. This simple experiment demonstrates that physical ob- jects can easily transition from order to disorder and back again when given just the right external conditions. The perfect amount of energy for self-assembly is similar to Brownian motion, where the components are able to move around, bump into one another, and con- nect easily. This amount of energy input can vary from system to system, depending on the given environment, material property, and bonding characteristics. For ex- ample, the amount of energy and the type of energy needed to move objects underwater is very different from objects tumbling around in a container or flying around in the wind. Underwater you may need waves, air pumps, or propellers to produce turbulence in the water, whereas with a tumbling system you may need a motor to drive a rotating chamber. If the components underwater 48 T H IN GS FA LL TOGETHER are neutrally buoyant, they will need a smaller force to move them around the tank, whereas if they are made of metal and sink to the bottom, they will need a much greater force to keep them moving freely in space. In a tumbling scenario, if the components are made of rubber versus clay, the energy landscape will look very differ- ent. The rubber materials will have a very delicate bal- ance of forces because if the forces are too strong, they will bounce off one another, while the clay components can easily stick to one another or mold around one an- other. Finally, the bonding strength and property of the connections can change the amount of energy in order to promote successful connections and weed out incor- rect bonds. If the connections have significant bonding strength, then you will need less energy to promote the connection of the units, but you will need greater energy to ensure that incorrect connections break off and even greater energy to disassemble the entire structure. The amount of energy input is subject to a very del- icate balance: if there is slightly too much energy, the components can bump into one another with too much force and either bounce off one another or break apart existing connections. With too little energy, the com- ponents won’t be able to find one another and will not move around and connect with one another. Similarly, there needs to be enough energy to break apart incorrect connections, yet not so much energy as to break correct bonds. It is like the Goldilocks principle for activation energy in any self-assembling system—with just the right environment, the structure moves toward equilibrium. The second ingredient is the geometry of the physical parts in the system—its individual components, materi- als, and connections. For living systems, the components may include DNA, proteins, cells, and other materials, all of which have physical properties such as size, shape, density, and bonding strength, which will influence their A series of 36" diameter weather balloons filled with helium inside a fiberglass frame with Velcro nodes. The balloon structures float around in the courtyard and self-assemble into various lattice structures. Credit: Self-Assembly Lab, MIT, Autodesk The balloon structures self-assembled into a large-scale cubic lattice. After the helium fades, the balloons float back to the ground and the self-assembled lightweight structural lattices remain. Credit: Self-Assembly Lab, MIT, Autodesk 50 T H IN GS FA LL TOGETHER interactions with one another. The geometry of a compo- nent is obviously important because it has to effectively interact with surrounding components and easily come together to build a precise structure. Certain geometries will promote two-dimensional structures, and others will promote three- dimensional structures. Some geome- tries will promote the linear growth or aggregation of materials, like lattices, and yet others will promote the formation of closed-loop structures. Think of lipids and bilayers that are formed from hydrophobic and hydro- philic interactions. The geometry and orientation of the lipids create very different structures in different envi- ronments. The key is to craft a local geometry that will aggregate to achieve the desired global structure. The size and density of the materials influences how they mix or separate when we add energy. If they are all similar, the components will mix in a uniform manner, but if they are vastly different from one another, they will separate into different regions. This can be seen in a principle called granular convection, and it comes into play when we’re separating elements into constituent parts.5 A coin-sorting machine is one example— coins separate due to their differences in size and density. Conversely, we can use this principle to promote close proximity and increase the chances that components will assemble on their own. For example, mixing mate- rials with different densities in a tank of water will cause some to rise while others will fall. This rising and falling behavior could enable objects of similar density to con- nect with one another. Some parts that are very buoyant might assemble into 2D sheets on the surface of the water and less buoyant parts will assemble at the bottom. De- signing their interactions requires close attention to a material’s characteristics such as elasticity, stickiness, or friction, which will dictate the types of interactions that occur among the components themselves.  ORD E R F RO M CHAO S 51 Finally, after ensuring that a system has just the right amount of energy and the components are designed to promote assembly, you have to take a look at the connec- tivity and interactions of the components. The goal of any self-assembly system is to promote the precise assembly of something useful, such as a product assembled from a number of pieces—not to just end up with a random mess of parts. The stickiness between parts is important, and this can be created by using adhesives, Velcro, magnets, surface tension, Van der Waals forces, or various other approaches to creating physical connections. If everything is extremely “sticky,” however, the system wouldn’t de- liver an ordered structure. The components need to have just the right amount of stickiness, in just the right places. The strength of the connection is important as well. If it’s too strong, the components will stick but won’t be ca- pable of breaking apart. This sounds like a good idea, but what happens if the parts come together in the wrong way? They will be forever stuck in the wrong place. If the strength of the connection is too weak, however, the parts will always fall apart, even when connected correctly. With just the right amount of connection strength, the correct parts can connect, stay connected, and con- tinue to grow stronger, while the incorrect and weaker parts will fall off. In other words, the parts themselves correct their own misguided attempts at connecting. For example, if there are many “male” and “female” parts tumbling around in a container, the similar connections (male to male or female to female) will have very little or no strength when they meet one another, whereas the male to female connections will be much stronger and will remain connected as the part continues to tumble. This is a simple technique to design error correction into a system based only on the strength of the connections. Patterned connections are another example of error correction. One way of thinking about this is a A number of unique components tumble around within a tank of turbulent water, eventu- ally self-assembling into a chair. This process was designed by using custom geometries and nodes that encode the correct assembly sequence. The process was filmed over seven hours, after which the fully assembled chair was complete. Credit: Self-Assembly Lab, MIT  ORD E R F RO M CHAO S 53 lock-and-key type of mechanism. A lock-and-key type joint is useful because it can dictate that only the exact components can connect with one another, while all other components can’t—they simply don’t fit geometri- cally. This is particularly useful when you’re designing structures that have a number of unique connections. You can create an infinite number of geometries that only fit with the complementary pair. Just like car keys or house keys that have many variations, you can create a number of unique connections that allow for only com- plementary connections. The lock-and-key patterning offers a wide variety of possible combinations (limited only by our geometric imagination) and can allow for geometric error correction. We demonstrated this fea- ture in our Fluid Assembly Chair project, where unique components had lock-and-key connections that pro- moted the self-assembly of a three-dimensional chair when tumbling around in a water-filled tank.6 Similarly, we showed the successful self-assembly of a cell phone from a few simple building blocks, the front enclosure, the rear enclosure, and the circuit board/battery core. These components were designed with precise male/ female connectors and a series of polarity patterns that promoted the complete self-assembly of the functional phone after being tumbled in a mixing chamber. Another example of patterned error correction comes in the form of polarity, which can be magnetic polarity, polarity in static charge, hydrophobic or hydrophilic po- larity, or a variety of other types. Magnetic polarity is the most common: a positive side attracts a negative side and vice versa, whereas two similar sides repel each other. This allows for the precise patterning of components, connecting only complementary polar neighbors. For example, a string of magnets in a positive-negative- positive-negative pattern will attract a complementary strand (similar to DNA, yet DNA has four base pairs), 54 T H IN GS FA LL TOGETHER with the pattern negative-positive-negative-positive. A similar patterning technique can be used with surface tension, for example: a hydrophobic material will repel a hydrophilic component but attract similar materials to itself. The “Cheerios Effect” is one example of this con- cept: Cheerios tend to cluster into hexagonal patterns in your cereal bowl due to the surface tension created by the milk and the circular shape of the cheerios.7 To take this one step further and start designing things that come together on their own, it may not be enough to sim- ply use polarity because there are usually only two poles (positive and negative), and in some cases a third, neutral (like a piece of steel that can connect to either magnetic pole). Wish as we might for an infinite number of poles, they don’t normally exist—unless we design them! Self-similar modules were released in a 500-gallon tank of turbulent water and self- organized into lattice structures based on the interactions of the elements and movement of water in the tank. The complete structures can then be removed or disassembled and thrown back into the chamber to self-assemble once again. Credit: Self-Assembly Lab, MIT  ORD E R F RO M CHAO S 55 By designing patterns of complementary matches, we can actually create an unlimited