快猫短视频

AS MILA BONCHEVA opens a drawer in a sideboard in her office, you might mistake her for a jeweller showing off her prized wares. She proudly pulls out perfectly intertwined spirals of pearly white beads, spine-like metallic necklaces, and hexagonal, red and green patterned copper twists, each precious object nestled in a felt-lined box.

But these delicate, decorative pieces are not jewellery, and Boncheva wouldn鈥檛 dream of wearing them for fear of causing damage. They are the results of the many experiments she has run to develop methods for manufacturing three-dimensional microchips.

Boncheva is one of many researchers worldwide who are striving to master the production of microscale 3D designs, with the eventual aim of mass-producing micromachines. With complex moving parts and their own embedded circuitry, such devices could have myriad applications, from tiny 鈥渟mart dust鈥� surveillance motes to nanobots that could be sent into the human body to deliver medicine or make repairs.

Boncheva鈥檚 structures are a long way from this. For a start, they are millimetre-sized, not micro- or nanoscale. And they are far from being autonomous. The most complex task any of them has performed is to save a stream of incoming bits and play it back. But it is demonstrating the principle of self-assembly that excites her and colleagues.

Boncheva鈥檚 structures assemble themselves from many millimetre-sized units. And she folds up each of those units in ways that are inspired by the way nature folds amino acid chains, which are effectively 1D, into large 3D protein structures. 鈥淲e look to nature for examples,鈥� she says, holding up a centimetre-wide electronic circuit made of copper-polymer sheets that have been folded into a helix.

Her approach builds on a series of other recent advances towards taking 2D manufacturing into the third dimension. In the early 1990s, several researchers independently hit on the idea of using a technique akin to origami, the traditional Japanese art of paper folding. Take a precision-manufactured 2D object and then fold it into a 3D structure. 鈥淭he reason we are using folding to achieve our goal is that it provides a convenient way to create 3D structures while fully exploiting existing 2D technology,鈥� says George Barbastathis at MIT, who is working on ways of folding up nanoscale patterned silicon sheets to make 3D forms.

The approach could easily piggyback on what computer manufacturers do already. If you can assemble 3D devices from flat structures fabricated using photolithography, the standard technique behind 2D chip manufacture, then much of the equipment that is already out there could be adapted to building 3D microchips, and one day even micromachines or nanobots.

Conventional photolithography begins by coating a silicon wafer with a thin layer of glass, followed by a thin layer of light-sensitive polymer. You shine a pattern of UV light onto the polymer, which dissolves away certain sections to expose the glass underneath. Then you dissolve away the exposed glass and deposit another layer of silicon, so the silicon layers are separated by glass except where the holes in the glass allow them to touch. Finally you dissolve away the glass and the polymer with acid. In this way, you can lay down an intricate but essentially flat pattern on a single silicon wafer.

But photolithography also provides researchers with a way to pre-program a chip to fold itself into a 3D structure. Back in the early 1990s, Kris Pister and colleagues at the University of California, Berkeley, began to take photolithography into the third dimension by finding a way to build a hinge on a chip. Pister has found a way to use the technique to lay down pin-and-bracket hinges (see Graphic). Once you can do this, you can start building much more complex 3D devices (see 鈥淧op-up chips鈥�).

But there is a limit to what you can do with this kind of origami, says Pister鈥檚 former student Elliot Hui. Like a piece of paper, a single sheet of silicon can only be folded a finite number of times, which limits the complexity of your folded structures. So Boncheva is trying to take 3D manufacturing to the next level, by assembling individual folded units into even more complicated structures.

Instead of working with silicon, Boncheva makes her pieces out of hollow copper units, each one folded together from a flatplan. Although Boncheva carefully manufactures each unit by hand, she is confident that they could be mass-produced using techniques that already exist. She is more interested in proving the principle that they can also assemble themselves into objects greater than the sum of their parts.

To assemble her structures she uses solder. This is an older microassembly technique than the silicon hinge, and is based on the work of Richard Syms and colleagues at Imperial College London. Take a copper sheet and deposit a line of solder along a central crease. When you warm the sheet, the solder melts, changing shape into a droplet to reduce its surface area. Just as things stick to your hands when they are wet, the copper sticks to the 鈥渨et鈥� solder as it changes shape, and this pulls the two halves of the sheet together to form a fold.

Syms now says he can use the same trick in silicon, and can control the final angle between the silicon plates to 1/60th of a degree. His secret is to include in his silicon flatplan a second flap, called a 鈥渕echanical limiter鈥�, which is anchored to one side of the main fold by a hinge that lies parallel to it. When the flap on the opposite side of the fold rises up, the free end of the limiter flap slides up it until it hits a carefully positioned kink. This jams the closing fold, fixing the two flaps at a specific angle relative to each other.

Boncheva can use this kind of technology to make very precise copper or silicon flatplans for individual components. In 2002, building on work done by colleagues in her Harvard lab, Boncheva made two common electrical components, a ring oscillator and a shift register, which assembled themselves out of identical hollow copper prisms and then joined together.

A ring oscillator is a circuit that oscillates between high and low voltage output, and components like it are found in today鈥檚 computers. A shift register either saves an incoming stream of bits, or outputs a saved stream of bits. Together a connected ring oscillator and shift register can act as a basic processor, with the shift register saving bits when the oscillator鈥檚 voltage is high, and reading them out when its voltage is low.

Making such a basic component 鈥� which is dirt cheap to make by conventional 2D lithography 鈥� might not blow anyone鈥檚 mind. Except that Boncheva鈥檚 device assembled itself without anyone or any equipment touching the parts. Before assembly begins, the prisms look like building bricks shaped like tiny triangular slices of cake. Each slice of cake has blobs of solder on its sides, and on the top is a green or a red LED, which allows the finished shift register to output the bits it has saved. The blobs and the LEDs are connected by stringing them on a connecting wire, like beads on a necklace.

Boncheva then shakes the prisms together in warm water. The warmth softens the solder enough to make it sticky, so that one prism can stick to another, when the blobs of solder match up. The formation of the structure is guided by the wires, so that each 鈥渂ead鈥� can only solder itself to the one lying next to it. As the solder sticks, it forms extra connections that complete the circuits. By connecting both ends of the wire to a voltage, Boncheva then showed the device was functional.

This proved that the idea could work, but stringing pieces onto a wire is fiddly, and ultimately Boncheva wants the method to work with microscopic rather than millimetre-sized subcomponents. So she has ditched the wire and experimented instead with the shape of the container the pieces are heated in. Instead of prisms, she is now using millimetre-sized rhomboid shapes, which look like cubes that have been squeezed, so that their top and bottom surfaces are equilateral parallelograms. Their faces are patterned by photolithography to form basic circuitry, and three solder blobs are then placed on each parallelogram face in the shape of an equilateral triangle.

Boncheva makes these rhomboids manually from a copper flatplan, but ultimately this could be done on a smaller scale with silicon, using some of the self-assembly techniques developed by the other groups. As before, she puts the rhomboids in a warm solution and sloshes them around, so their solder spots soften and allow the pieces to stick together.

The orientation of the solder triangles means that there are only two ways that two rhomboids can stick together, either so their parallelogram faces are glued on top of each other, or so they are half-overlapping and twisted at 30 degrees to each other. The result is a long chain that has parts running in a straight line, and parts that turn helically. When Boncheva first carried out the experiment, the result was an uncontrolled higgledy-piggledy assortment of linked rhomboids in a twisted chain. 鈥淵ou end up with a messy, ugly structure,鈥� she explains, and that wouldn鈥檛 be much good for making a useful device, since you need to ensure you get the desired shape every time.

In her next experiment, she used a very narrow container that was too cramped to allow any helical sections to form. Sure enough, it produced a straight chain of linked rhomboids with no helical kinks. And when Boncheva made the container exactly wide enough for a helix to form, she found that she got helices and no straight lines 鈥� any straight line sections that started to form were broken apart when they hit the sides of the container. 鈥淭he lines were mechanically unstable in this container,鈥� says Boncheva. 鈥淭he beauty of this method is that we not only control geometric structure, we also control the electrical functionality of the devices.鈥� That鈥檚 because every new rhomboid-rhomboid connection forms new electrical connections too.

Boncheva hopes to combine this with self-folding techniques to produce a two-stage manufacturing process: individual units first self-assemble by folding and then those units join up, like the rhomboids, to form even more complicated 3D structures. Even better, the whole assembly might be condensed into just one step, says Boncheva.

But as the 3D device becomes more complicated, designing the flat starting components could quickly become a headache. 鈥淚t would be great if we could design the 2D structures more intelligently. Right now we use trial and error,鈥� says Boncheva. That is why the end result can sometimes look more like a piece of abstract art or a necklace than part of a shift register. 鈥淚 keep many photos of them at home, for strictly decorative purposes, pictures on the walls, bookmarks,鈥� she says.

Barbastathis is already tackling the design difficulties for his fold-up structures, and has enlisted computer programmer Erik Demaine, also at MIT, to help him write programs that will produce a 2D plan for a desired 3D structure.

Boncheva hopes programmers and designers can help her design self-assembly procedures in the same way. Right now, she is still waiting for manufacturers to come to her with a description of something they want that they cannot assemble, so she can get started. And it is only a matter of time. 鈥淭here鈥檚 a lot you could do with a 3D microscopic structure. There is just a high probability that in the future we are going to need techniques like these,鈥� says Hui. Then today鈥檚 trinkets will be tomorrow鈥檚 microchips.