The Incredible Shrinking Finger Factory

Peter will could see he was facing a serious problem. All around him things were shrinking. All kinds of things. When did this start? Why hadn’t he noticed it sooner? Growing up in Scotland in the 1940s, he used to take toys apart to see how they worked, but back then things were hefty—bicycles, typewriters, radios. Electronics started to shrink in the sixties and seventies, when he was a young engineer at ibm; but the machines at least stayed substantial, and the industrial robots he worked on were reassuringly beefy contraptions. By 1988, however, mechanical things were shrinking, too. Researchers had just built an electric motor 60 microns in diameter—60 millionths of a meter, .002 inch, less than the width of a human hair. When charged with static electricity, the rotor would actually spin. Pretty impressive, Will thought, but one thing bothered him. To build even one complicated machine that small takes a herculean effort. How would anybody ever manage to assemble such machines en masse? At the Silverado Country Club in Napa, California, Will posed his question to a gathering of engineers who specialized in making these tiny machines. He had to worry about it himself now because he had recently taken a job as director of manufacturing research at Hewlett-Packard in Palo Alto. He popped a tape into the videocassette player. A giant hand appeared on-screen, holding needle-tipped lab tweezers. Below the tips, a mere speck of a part—a transistor some 2 millimeters square, used in microwave instruments—rested on a table. This component had to be lifted. The matchup looked hopeless. As in anyone with a pulse, the hand trembled ever so slightly. The tweezers magnified this shaking, so that their tips waved in arcs just above the speck. Bravely they closed in and nabbed it, most likely between heartbeats, the same way a target shooter times the squeeze of a trigger. Ah, sighed the audience in relief. But now the speck was stuck on the tweezers. The needle tips parted, but the component would not let go; static electricity made it cling. The hand tried to shake it off, harder and harder. Ooh, moaned the audience. When the speck finally did drop, it landed not in its proper place, where it was to be soldered, but back on the table. The hand started over. This time the tweezers bore down a little too hard. One tip slid off the edge of the speck like a tiddlywinks shooter, and what happened next was truly amazing. The speck leaped away, completely out of the picture, like a homer off Babe Ruth’s bat. Oooooh, the audience gasped. Will’s videotape pretty much scotched the notion that human hands could ever assemble tiny mechanical contraptions out of a pile of gears, flaps, wheels, and rotors. But what about getting some kind of tiny robot to do it? Given what he knew about robotics, Will thought that unlikely. At ibm in the sixties, he built the rs/1, the first robotic arm able to select electronic components and insert them into circuit boards. He worked for decades to get it and robots like it to handle even more complex tasks. And though his efforts led to many breakthroughs and made him a legendary figure in robotics, ultimately he and his colleagues failed. Robots are well suited to spraying paint on automobiles or welding seams and even for stuffing electronic chips onto printed circuit boards, but when it comes to putting together anything intricate, they are useless. At Hewlett-Packard and elsewhere, all the delicate assembly work was done by hand. You still see photographs of rows and rows of people, often in the Far East, assembling things on production lines, says Will. You see it in Silicon Valley as well. Basically, we roboticists haven’t shown we have the stuff to make this easy. Will’s message to the assembled engineers was clear: shrink anything in the lab as much as you want, build the most wonderful tiny machine, but sooner or later you will want to manufacture it. All their wonderful early breakthroughs would come to naught, he said, because we have no technology to build ’em. This drawback hadn’t stopped the attempt—a growing number of researchers were building not only microparts, not only micromotors, but far more complicated machines: a working car the size of a grain of rice, an entire working lathe. Most of these miniature wonders were made in Japan, where industrial research teams tended to follow a strategy of shrinking conventional machines to small dimensions and assembling the parts with superhuman patience. In the United States, by contrast, researchers tended to make simpler, flattish microdevices all of one piece right on the surface of silicon, with electronics embedded in the same chip. No assembly required. That approach dates to the 1970s, when researchers at Stanford and ibm built the first commercially useful micromechanical device: a simple cantilever, a flexible flap like a diving board. At present, the minuscule silicon flap plays an important role as an accelerometer in automobiles. It won’t bend much when a car slows down, but the sudden deceleration of a crash will bend it far enough to switch on a microcircuit that activates an air bag. An equally simple silicon machine came next—thin, microscopic diaphragms that flex in response to pressure. You can find them today in disposable sensors that monitor blood pressure in a hospital patient’s intravenous line. Not least among its virtues, this micromonitor reverses an ugly trend in high-tech medicine. It replaces, for about $10, a much larger instrument that costs $6,000. Another useful diaphragm, the micropump, works much the same way, but instead of flexing in response to fluid pressure, it actively pushes the fluid. Engineers have gone far with these 2-D machines, and in recent years applications have burgeoned. Virtually all modern cars now have a device to gauge airflow through an engine’s intake manifold, figure out precisely how much fuel should go with it, and squirt just enough to ensure efficient, low-polluting combustion. And there’s more to do. The cantilever that launches air bags, for instance, can also be used as an exquisite chemical sensor. Its tip, if coated with a substance that attracts mercury vapor, say, will bend enough to set off an alarm in the presence of that deadly gas. Likewise, the membrane that monitors blood pressure could just as well track any other kind of pressure, like air pressure in tires. If all the tires in the United States were properly inflated, that alone would reduce fuel consumption by 10 percent. Researchers are also working on bench-top chemistry labs reduced to chips the size of a fingernail, portable enough to take to a crime scene, as well as computers with much smaller hard drives. But there is a growing sense among many researchers that we will soon be chafing at the limitations of 2-D devices. More and more, the really exciting ideas for microscopic machines seem to suggest 3-D assembly. And even five or six years ago it was becoming clear that the paths of the holistic, 2-D etchers from the United States were beginning to converge with those of the piecemeal 3-D shrinkers from Japan. Engineers, in other words, were thinking about joining 3-D micromechanical parts with 2-D devices to make little single-minded robots. A complete system would have three components: a sensor like the cantilever, a microprocessor (the brain), and an actuator like a pump or motor. These tiny machines would sense a change in the environment, compare it with a programmed ideal, and kick out the appropriate action. Engineers coined a name for such compound smart machines—microelectromechanical systems, which they mercifully shrank to the acronym mems. The idea behind mems is simple: shrink and integrate. Shrink sensors, shrink actuators, shrink tiny electronic brains; cram the works into micropackages that could replace a whole shelf-load of big stuff. The hardware controlling a guided missile is big stuff; mems people talk about guided bullets. Surgery is big stuff; mems people envision smart pills that might cruise through blood vessels like the submarine in Fantastic Voyage, dosing tumors or reaming clogged arteries. Fat video screens would be made obsolete by inch-square projectors with millions of swiveling micromirrors, each reflecting one image pixel. Even a nonmobile chemical sensor coupled with a pump could work wonders as a microsize medical implant. In a diabetic, it might drip insulin into the bloodstream whenever blood sugar begins to rise; in a cancer patient, it could maintain a steady concentration of therapeutic drugs precisely at the site of a tumor. Some surgical techniques that already require miniaturized tools seem to cry out for mems treatment. For instance, to get at a blood clot in a stroke victim quickly enough to head off brain damage, doctors inject a clot-busting drug through a tiny catheter and then snare out the remains with a minuscule loop of wire. Why not send a mems robot into the artery to do the dirty work? Surgeons have already tested miniature robotic arms for performing coronary bypass surgery. The arms wield scissors, needle holders, grippers, and cameras yet are small enough to fit through a one-millimeter catheter. A mems robot might be 50 or 100 times smaller still. There was one other mems device on Will’s wish list: an assembly robot. Without one, how else could all those other devices be manufactured in the first place? But even if he knew how to design one, he would need hordes of them to even begin doing any useful assembly, and he could not imagine building these hordes by hand, even with the best tweezers money could buy. One prototype, perhaps, might be possible—with painstaking, heroic effort. Researchers who traveled that road had spent years and small fortunes to make a single working model. Will compares the approach to writing the Bible on the head of a pin. People do that, he marvels. But it didn’t seem appropriate to me, after thinking about it for five minutes. We had to find a better way. When attacking a problem, Will says, what I try to do is solve it, but at the same time abstract some essence of it. As his videotape showed, the problem was to manipulate parts that tended to fight back, leaping or clinging unpredictably. To find the essence of misbehaving miniparts, he began looking and thinking and playing with things—big as well as small. He shot tiddlywinks across his desk to imitate leaping parts, and he rubbed things against surfaces to build up static electricity. He quickly found a common thread. I realized, he says, that a lot of the problems were due to scale. In one sense that statement is trivial. Scale means relative size, and obviously there’s a mismatch between human-size tools and gnat-size machines. But in another sense Will had nailed the essential problem. Engineers have long known that scaling things up or down changes their physical behavior. Most of us non-engineers also know this intuitively. A grasshopper and a hippopotamus have vastly different athletic abilities, not to mention physiques. Although the grasshopper can leap many times higher than its body size, the hippo, were it so inclined, would be hard put to hurdle anything much higher than its ankles. That’s because as you scale up from bugs to beasts, weight increases as the cube of the animal’s size (its volume), while muscle strength increases only as the square of its size (its cross-sectional area). In general, things seem heavier in proportion to their size as you scale up and lighter as you scale down. When you descend to realms a thousand times smaller than a grasshopper, the leap is incredible. It’s as if you were playing tiddlywinks, Will says, and instead of a thing ending up close to where you tiddle it from, it’ll leap hundreds of feet. None of the scaling effects that Will observed surprised him. Scaling is a basic concept in engineering, he points out, though not always remembered. Engineers have formulas to adjust their designs for scaling effects. Yet micromachines inhabit a world so much smaller than ours that scaling effects become hard to imagine, even for engineers. You get bizarre things happening, Will says. They don’t violate any laws of physics, but from a human experiential point of view, they’re bizarre. Not only is it easy to knock a microscopic component out of the ballpark, but at the same time other forces conspire to pin down its feeble bulk. We don’t have much trouble peeling off a sweater that’s crackling with static electricity, but a miniaturized component sucks up to static charge like matter to a black hole. Fluids are just as bad—microparts can’t break the surface tension. In a microscopic motor, oil would gum up the works, resisting motion like hardened tar. Yet the wheels on an automobile smaller than a grain of rice would generate too little friction to move the car. The vehicle’s weight would be so insignificant, and so lightly would it push rubber against road, that the wheels would spin as if on ice. As for assembly robots, Will knew that the effects of scaling them down would completely sabotage their actions. You couldn’t just miniaturize the hell out of them, he says. If they were ever to become reality, it seemed, he would have to reinvent everything. In May 1992, as Will was still pondering the problem, he left Hewlett-Packard for a research job at the Information Sciences Institute, a laboratory run by the University of Southern California in Marina del Rey, a shoreline suburb of Los Angeles. Having spent his entire career in industrial labs, he relished the view from his office window of the peaceful harbor below. And as he gazed out the window, he repeatedly turned over in his mind the idea of miniature 3-D robots that would assemble microdevices from an array of parts spread out on a tiny table. It’s only natural to think that way. Most of us, as kids, built things by dumping parts on a table and putting them together. When we grew older, that was our paradigm for assembly. It was Will’s too. At the age of 57, he had already spent a career trying to make robots that would do all the work of assembly. But what if you just opened a box, dumped parts on a table, and the table itself built your stuff? Automatically. Built whatever contraption you programmed it to make. Instead of having this tiny robot man, this homunculus, this arm that has little joints—maybe the best way to do this, he thought, is to upend the problem. Instead of putting the intelligence in the robot and having a dumb table, maybe a better thing to do was to have a very smart table and a very dumb robot. Or maybe no robot. He eventually dismissed the homunculus. The table itself would assemble things. Will had come up with a fantasy worthy of children. This vision, of course, lacked important details. How can a table be smart and physically active? Will would need some sort of mechanism, embedded in the surface, that could manipulate things. He envisioned little pushers, or maybe grippers, all over the table, with intelligence distributed among them—thousands of microsize fingers, say, waving on the surface like wheat in a field. Each of them could be programmed. Acting alone, none would be very capable. Together, they could sort and assemble parts. But microscopic fingers would be hard to build. Will wanted something simpler. So he put aside the notion of fingers and returned to a plain, smooth surface, his tabula rasa that would build stuff. He thought of how objects slide on a surface when you shake it. He knew he could move things that way, but it didn’t seem good for assembly. You can shake bread crumbs onto a casserole, but imagine trying to piece together a jigsaw puzzle by shaking the table. Will needed more control. He thought about covering a surface with fur or velvet. The nap in these materials makes it harder for objects to slide in one direction, easier in the other. By manipulating the nap, perhaps he could control the motion. And it struck me, Will says, that this is very much like ciliated things. Microorganisms have mobile, hairlike projections, or cilia, that propel them through water. Tiny sea animals have cilia that wave bits of food toward their mouths. Cilia coat the surface of air passages deep inside our lungs, churning like conveyor belts to expel dust and mucus. A mechanical version, Will thought, might push small objects across a surface. With the help of computing circuitry embedded in the same surface, the cilia might be programmed to sort things by size and shape, bring the right parts together, turn or align them correctly, and present the assembled pieces for one final joining operation, like the final soldering on circuit boards. The more Will thought about this, the more he convinced himself that it wasn’t a crazy idea. Thousands of identical cilia, press-gangs of dedicated cilia, could be made to cover the surface of a standard computer chip. And these factories-on-chips would usher the next wave of mems into everyday life. Such an Intelligent Motion Surface, as Will dubbed his hypothetical robotic table, would assemble not the current, simple mems devices like air-bag accelerometers but the next generation of fantastic, complex microscopic machines. Today, five years into Will’s mems project, Murilo Coutinho sits in front of a desktop computer at isi, playing games—cilia games. He is testing strategies for manipulating objects on a ciliated surface. As he works the controls—directing cilia to push this way or that, shaping the flow of boxlike objects that lumber across his screen—he looks engrossed, like a kid marshaling toy trains in a railroad yard. The cilia themselves don’t appear on-screen. Instead, bold arrows represent the direction in which a large field of cilia is pushing: up, down, left, or right. If Coutinho sets up a series of fields all pushing toward the right (represented as >>>>), he can drop a part anywhere on the surface and it will chug toward the right. If he sets up opposing fields (>>>> <<<<), they will shove a part toward the center, where it stops. He can then reconfigure fields to make the part rotate, say, by pushing its right side toward the bottom of the screen and its left side toward the top. Or he can bring in other parts to bully this one into place, like tugboats helping dock the Queen Mary. (That idea, Will says, came from watching boats in the harbor.) His teeny tugboats surround and nudge recalcitrant parts. Unlike tweezers, they can’t tiddle. Coutinho now has a library of cilia-field configurations he can combine to do amazing things. He can, for instance, program into a field a hole (where no cilia move) that’s the shape of a particular part, leaving it stone-still while fields elsewhere move other parts to other spots. Perhaps they freeze squares in square holes, dump circular parts off the left edge, and then dump malformed squares (the ones that didn’t fit into holes) off to the right. The surface does indeed seem intelligent. It recognizes parts: it can toss out bad ones and coax good ones to their proper places and orientations. And that’s the essence of starting to build, Will says—not to mention a remarkable feat for sensorless robotics. Will’s group is ready to download custom-made programs into real microchips. Of course, in simulation it’s easy to make rapid progress. Using his computer to reconfigure fields, Coutinho can whip up a new make-believe ciliated surface in several minutes. Real microchips, on the other hand, require about 14 weeks to design and build. Obviously, says Will, the software is further along than the hardware. As it turns out, a cilium that can push an object in any direction is a very complicated appendage, mechanically speaking. Will’s first design was ambitious—a hinged teepeelike structure, supported on two legs, with a motor at the base of each. It was rather clever from a mechanical point of view. I was fully going in that direction, he says. But that’s a very hard thing to build. You would have to etch material from solid silicon until you’re left with a joint with three hinges attached to two motors. I started building that horrible mechanism, Will says. And it was no good. It was too complex. Fortunately, MEMS researchers were eagerly sharing information. At one meeting, when Will described how he wanted his cilia to work, someone retorted, But Fujita’s already done stuff like that. Will checked the references—Hiroyuki Fujita, Tokyo University—and they turned out to be crucial. In the midst of a 3-D engineering culture, a brilliant 2-D strategy had emerged. Fujita had done a really cool thing, Will says. He had built a simple mems device, a flat rectangular strip that would bend when heated. That’s all it did: curl up like the cover of a paperback book. It was made of two materials—a plastic coated with metal—so when Fujita passed an electric current through the metal layer, causing it to heat, expand, and relax, this rectangular flap drooped down. When the current stopped, the metal would cool and contract like a tense muscle, and the flap would curl up again. Curling up and curling down, these flaps resembled waving fields of cilia. But they could push in only one direction; they couldn’t twist or swivel. Nevertheless, Fujita had lined up his little flaps in rows, and he’d demonstrated that they could indeed walk tiny objects over a surface, in a more or less straight line. With no joints or motors, Will realized, such flaps would be much, much simpler to build than the contraption he was struggling to design. Really, this flap was little more than the original mems cantilever, made active instead of passive. It wiggled and it waggled. How much simpler, and dumber, could a robot be? Will felt he was halfway home. Now all he needed to figure out was how to use Fujita’s linear, one-way ciliary motion to push things any which way on a surface. The solution—a notion inspired by color TV, he says—came quickly. Television screens have image pixels, each containing three dots: red, blue, and green. A broadcast signal dictates which dots to light up in each pixel, and the colors change as the signal changes. Will decided to make motion pixels. On Coutinho’s computer screen, the arrows that simulate cilia fields point up, down, left, or right. Those are his only choices. That’s because Will had decided each motion pixel on the hardware would contain four flaps, one oriented to push up, another down, another left, and another right. A controller, part of the circuitry etched into the chip, would select which one (or conceivably, which ones) to activate at any given time as the software ran through its commands. In the office next door, Adam Cohen, Coutinho’s counterpart on the hardware side, frets his way through another day. Cohen, a physicist and a 3-D sort of guy, has the unenviable task of figuring out how to make the actual ciliated chips with the same technique engineers use for making 2-D electronic circuits. It has an evil-sounding name, sea moss, written cmos, which stands for complementary metal-oxide semiconductors. Conceptually, making an electronic chip is like making an etching for art class: First you project a pattern of lines onto a slab of crystallized silicon, and then you use a caustic chemical to cut away the areas where you don’t want material. Then you deposit a layer of material and repeat this etching process with layer after layer until you have an elaborate 3-D structure of trenches and columns and basins filled with various types of metals and semiconductors. And yet, no matter how intricate these electric circuits become, each layer remains essentially a 2-D pattern. Cohen, by contrast, has to somehow coax a third dimension out of these patterns. He has to keep etching away material until he’s left with mechanical parts of appreciable thickness—that’s the source of his headaches. It’s like taking a slab etched lightly with a nice, flat Dürer engraving, exposing it to etchants for a longer time, and hoping to pull out a three-dimensional, sculptured relief worthy of Michelangelo. To make a flap on the surface of a silicon chip, Cohen and his colleagues etch around three sides of a rectangle. The etchant gas also carves silicon out from under the flap, while the top, being masked, remains intact. A layer of aluminum is then laid above the silicon, forming the complete two-layered flap, which measures 400 by 100 microns on current isi chips, roughly the size of a chin whisker. We can go a lot smaller, Cohen says. Chip makers today can, in fact, etch quarter-micron features. But in mems, people don’t operate on the cutting edge. If you’re at ten times greater, he points out, it’s still a thousand times better than what could be done before. Will and his colleagues have had to experiment with many variations on the fine-tuned cmos techniques of circuit manufacturers. The semiconductor processes we use, particularly the etching, are unpredictable, Cohen explains. You may need the etchant to remove 100 microns of material on this particular location, and it removes 200. If your flaps are only 2 microns thick, that’s nasty. In fact, the first time they tried to make flaps, the chip’s surface came back pocked with craters where the flaps should have been. So what are you going to do? Cohen shrugs. Well, you’re going to go through the whole cycle again, change your design to compensate, or change the processing—all of which eats up several months. It’s a very difficult, frustrating, and chaotic way of doing engineering, he complains. But for now he has no choice. There’s a reason, of course, for this madness. The beauty of using cmos to make cilia, Will says, is that it is the same technology used to make ever-shrinking integrated circuits, which means you can incorporate both cilia and electronics on the same chip with no extra steps. For instance, while Will is putting aluminum on top of the silicon flaps, he can also put aluminum onto other parts of the chip to make up the electric circuitry that controls the flaps; all he has to do is etch out a wiring pattern. There’s another virtue in tying mechanical parts to the same technology as electronic circuits: as the electronics continues to shrink, so do Will’s little flat robots. If you can pack more transistors on a chip, it means you can etch circuits in finer detail. Therefore you can also make cilia smaller, and they’ll manipulate smaller parts. That’s why Cohen will stay chained to cmos etching—with its charming clean rooms, bunny suits, and poisonous gases—until he invents a better method that chains electronics to his technique. He’s working hard on that now. Characteristically, after having painted this shining vision of electronics and mechanics marching in lockstep toward the micro-horizon, Will feels the need to tone it down. This is all dreams, you know. Easier said than done. Recently, the team backed away from an earlier goal of getting circuitry and cilia to work all at once. They started building, instead, a series of limited-purpose experimental chips, with flaps controlled by external equipment. Chip number one in this series looked like hell, Cohen says, and we still don’t know why. Few of the flaps worked. Chip two was better: the flaps worked, but not all worked well. Cohen picked the best flap design and used it for the next round. Chip three made it into Will’s film archive. Cohen designed it to move parts straight across the surface, conveyor-belt style. The film shows a blocky speck of silicon, standing in for some mems component yet to be invented, swaying and wambling as it moves across the surface—not in a very controlled way, Cohen admits. But it moves. They might eliminate the swaying, Cohen says, by making the cilia smaller, so that the objects they are pushing always have more than two or three flaps underneath them. The speed of motion is about five to ten flap lengths—a few millimeters—per second. That too must improve. Chip four, wired to instruments in a room down the hall, is state of the art. Smaller than a pinkie fingernail, it shimmers like a tiny pond as the metallic flaps curl up and down. Cohen made this chip to rotate an object. With four fields of cilia laid out in four quadrants, its design resembles, conceptually, the Four Corners intersection of state borders in the southwestern United States. Imagine all the cilia in Arizona pushing north, in Utah east, Colorado south, and New Mexico west. A part plunked down in the middle of this chip, on the four-corner intersection, will pinwheel around on the surface. And since Will makes the cilia in opposite-facing pairs, the same setup can make the part pinwheel in the opposite direction. That’s another stride forward but still no match for the choreography Coutinho performs on his computer screen. Clearly, the hardware must improve enormously to catch up with the software. For one thing, Will must make something closer to the motion-pixel idea, in which the cilia would act in groups of four, rather than in four quadrants. He needs to add the circuitry to control the cilia. And it may be time to add some kind of sensor. We’ve gone about as far as we can without sensors, Will says. It’s stupid not to. A camera mounted above the chip, he suggests, could let the intelligent surface know when parts have arrived in holes, or strayed off a pushing field. A microprocessor, given this feedback, could adjust the program’s speed on the fly, or modify the shapes of fields to round up stray components. There’s still time to experiment, because mems manufacturers don’t yet have a burning need for a ciliated microchip. It was a difficult challenge, and I’m not sure it’s one we’ve solved, Will says. There are lots and lots of options still. Friendly rivals, as Will calls them, starting around the same time he did, have made robotic surfaces from materials other than silicon, and they’ve built cilia as well as other kinds of flaps or pushers that mimic muscle fiber, accordions, inchworms. Flexures are good on this scale, Will says. It’s too soon to say, though, which flexure, if any, might eventually work. But it seems clear by now which one has launched a new field of robotics. The cilia was the first idea, says Will, and it catches people’s imaginations. The Future in Cilia Peter Will calls it the Intelligent Motion Surface. Other researchers call it vector field robotics. The idea is the same: to use the technology of semiconductors to make a whole new kind of robot—a flat, smart, ciliated surface. Although conceived originally to manipulate objects too small for fingers and tweezers, some researchers are finding that it applies handily to the world of human-scale objects as well. Here is what’s in store for ciliated robots. Microvalves. Andrew Berlin at Xerox Palo Alto Research Center in California is developing microscopic valves for exceedingly accurate printers. He has built a prototype printer that has 50 microvalves that move the paper through the printer by shooting jets of air. Multi-pedes. Bruce Donald at Dartmouth is literally turning ciliated surfaces upside down. Rather than moving objects over a surface, his cilia would serve as tiny legs or fins that make microbots crawl or swim. These robots might become mobile smart pills that dispense drugs deep within the human body, or tiny repair-bots for machinery. Donald hasn’t given up on vector field robotics, however. He has developed a mathematical proof that any shape can be sorted and assembled using the technique. A colleague is currently working on a similar smart surface for United Parcel Service to sort real packages. Microflaps. Archimedes supposedly boasted that he could move the Earth with a lever, given a suitable fulcrum. In a similar vein, Chih-Ming Ho of ucla wants to steer airplanes with flaps too small to be seen with the naked eye. Microflaps covering the wings and fuselage would push around the thin layer of air that flows over these surfaces, to the same effect as bigger flaps. So far he’s demonstrated the concept on models, and he’s convinced it will scale up to full-size airplanes. Plastic microbots. Gregory Kovacs and John Suh at Stanford have eschewed the silicon of traditional electronics and have instead concentrated on making arrays of cilia out of plastic, which they say has better mechanical properties. The first application may be for positioning samples in electron microscopes, but assembly might be a long-term goal. Micromirrors. In 1996 Texas Instruments brought out its Digital Micromirror Device, a chip with half a million mirrors fixed onto moving flaps similar to the cilia made by Will and others. The dmd is used for ultra-small projectors. Although it’s a big jump from reflecting light to manipulating solid objects, ti’s commitment to mass-produce the flaps promised refinements in manufacturing and gave mems research a fillip. Meng-Hsiung Kiang, a researcher at DiCon Fiberoptics, and her colleagues at Berkeley are building micromirrors for tiny low-cost lasers that might eventually be used for communications, optical computing, or inertial guidance devices.

The Incredible Shrinking Finger Factory

For years engineers have touted the future wonders of microscopic machines. But they have ignored a real-world-size problem: If machine parts dwindle to tiny specks, how will human or even robotic hands ever assemble them? One man's ingenious solution:

Newsletter

Will Hively

Published in:

Machines of the Very Very Small

New Cannibalistic Robots Consume Other Machines to Grow and Heal on Their Own

Shark Skin and Cicada Wings Inspire New Antimicrobial Cutting Boards

A Swarm of Cyborg Insects Might Save You From Disaster

Robots are Coming to the Kitchen − What That Could Mean for Society and Culture

These 4 Humanoid Robots Use Soft Touch to Hug or Connect with Humans

An Ancient Stingray Lookalike Inspires Underwater Soft Robotics

Advanced Quantum Material Curves the Fabric of Space

4 Robots That Look Like Humans

Robots and Artificial Intelligence Have Ancient Mythology Origins

Scientists Are Trying to Give Robots Social Skills

Some Researchers Debate if we are Living in a Computer Simulation

How Will We Know When Artificial Intelligence Is Sentient?

Study Suggests What the Human-Robot Relationship Looks Like

Tiny Robots Are Fishing For Microplastics

Robots of the Future Might Walk Like Emus

Stay Curious

JoinOur List

SubscribeTo The Magazine