In 1867, Jules Verne imagined spaceships propelled by the pressure of light. In 1871 James Clerk Maxwell predicted that such pressure actually existed, and in 1900 Pyotr Lebedev confirmed that prediction experimentally. But 109 years later, engineers have yet to find a practical use for this odd force.
Maybe it's just too weak. Light pressure equals beam power divided by c, the speed of light. A one-milliwatt laser pointer, therefore, presses its object with the force of 3.3 piconewtons. You could lift a penny with laser pointers—you'd just need 30 billion of them.
Yet such weakness becomes a strength when you're trying to nudge something of nanometer size and picogram mass. And the optics for directing light on such a scale already exist, in the form of miniature waveguides, couplers, and beam splitters, all of which are now routinely laid down on silicon-on-insulator substrates. We use these nanophotonic devices for optical processing. Why not also use them to exploit the ultralight touch that light itself can provide? Why not use light as an actuator, reaching right into the guts of an integrated circuit to throw tiny switches, either to control electronic circuits or, better yet, to reroute light itself, and the data that it carries?
Such a marriage of nanomechanics and nanophotonics would bring us a giant step closer to optical chips. That's important, because light has a vastly wider bandwidth than electricity, which would enable it to get around the critical bottleneck in computing: the connections between processors. If light could act directly on circuit elements without first being converted to electricity, entire systems would run faster. You could imagine yoking together the multicore processors in a chip, making them run much faster and more efficiently than they can now. And if we really master the technology, optically controlled switches might ultimately supplant transistors, ushering in an era of all-optical computers. The resulting speedup would be stupendous, even by the standards of modern computing.
At the beginning of 2007, shortly after I joined the engineering faculty at Yale, I began to assemble a team to find ways of using light to drive silicon devices on a nanometer scale. We would be building on nearly two decades of research.
The quest had begun with microelectromechanical systems, or MEMS, which as the name implies are built in dimensions measured in micrometers. Today's engineers find it easy to build resonators—tiny tuning forks, basically—at that scale by simply embedding them within electronic circuits. The circuits drive the resonators through electromechanical coupling, typically by pairing them with electrical plates—one fixed, the other on the movable MEMS. In such a scheme, a current applied between the plates alters the gap between the plates, which changes the capacitance and further induces a current that oscillates in response to the motion of the plates. Basically, you feed in a steady current and get an oscillating one in response. In a cellphone, for example, such oscillators are used in filters, picking out the desired signal from the swath of frequencies your antenna pulls in.
The smaller the size of the plates and the gap between them, the faster the oscillation can be and thus the higher the frequency that can be isolated. At the nanoscale, such oscillators attain the frequencies needed for microwave communications. However, the high frequencies give rise to complex impedances both mechanical and electrical. Because the structure is so small, the impedances are often badly mismatched.
How badly are they mismatched? In high-frequency circuitry you normally want every component to have an impedance of around 50 ohms. At these tiny dimensions, though, you're going to end up with an impedance that's millions of times as great, which means that essentially none of your signal will get through.
Scaling down from MEMS to nanoelectromechanical systems, or NEMS, also brings on other, more fundamental problems. First, in the nano realm, the oscillators are so fast that the conventional electronic circuitry they work with can't keep up. Second, the oscillators' signals are so faint that they can get drowned out by the random noise that's endemic in any electronic circuit. A NEMS device can barely make itself heard over noise that's just one-thousandth as strong as what you'd find in a typical IC.