In the fast-shrinking realm of electronics, micro is already retro. Today’s smallest transistors—the fundamental components of information processing—are about one-thousandth the width of a human hair and operate at several billion cycles per second, or gigahertz. That’s fine for listening to CDs or playing The Sims. But it’s way too bulky and nowhere near fast enough for the next generation of devices.
The most dependable way to increase the speed at which electric charges work is to decrease the distance they have to travel. And there’s the technological rub. By the end of this decade, many experts predict, the current method of making microcircuits—done by shining high-frequency light through stencil templates to etch connections onto a semiconductor chip—will have reached its physical limits, with individual elements bottoming out at something like one-fifth the size of today’s best.
At those dimensions, troublesome quantum behaviors start to pop up. Electric charges can bleed through their insulation and tunnel into neighboring lines. Heat becomes a huge problem. Not surprisingly, many experts feel it’s time to dump the transistor altogether and move on.
Where? Ideas abound. One, recently developed by a Hewlett-Packard team, is a “crossbar latch” circuit made of platinum and titanium wires that are only a few dozen atoms in diameter. They lie across one another like the strings of a tennis racket. Other groups are looking at using a single organic molecule, strung between two electrodes, as a transistor. But whole molecules are still too big for those scientists furiously at work on what they regard as the supersmall Next Big Thing: spintronics.
Conventional transistors, small as they are, work by controlling the bulk movement of myriad electric charges through differently “doped” sections of silicon. A silicon crystal is doped by adding trace amounts of some other element that has either more free electrons than silicon’s four (producing an electron surplus and making the result negative, or n-type) or fewer free electrons (producing a net electron deficit, and hence a positive, or p-type). Charges are herded between two regions of one type by passing through an intervening “gate” of the other type. The gate is opened or closed by applying a voltage. It’s magnificent technology. But it’s not fundamentally different in principle from Edison’s lightbulb—it achieves its effect by moving vast hordes of charges around.
Charge, however, is not the electron’s only talent. Each also has a weird property called spin: It behaves as if it were an infinitesimal revolving sphere with a magnetic field that aligns with the spin axis. Each electron is either “spin up” or “spin down,” a property that can be reversed with a magnetic field. Because binary computing only recognizes two states—0 or 1, on or off—the two spins can be used in a similar way. Doing so offers several advantages. For one thing, virtually no energy is needed to “flip” an electron’s spin. The flip transpires in dramatically less time than it takes to get a herd of electrons moving toward a target.
And it works. In recent years, electronics researchers have made light-year advances in their ability to manipulate electron spin. In the past two years David Awschalom and his colleagues at the University of California at Santa Barbara have demonstrated new and ultrafast means for generating, transporting, and manipulating spins in semiconductors at up to 100 GHz using electric fields. That could make the technology more attractive to chipmakers, who have already invested billions in plants to build electrical connections on silicon.
Even more tantalizing, it may be possible to separate spins without either an electric current or a magnetic field. Awschalom’s team recently discovered an effect predicted 35 years ago, called the spin Hall effect: By introducing certain chemical defects into a semiconductor, electrons with opposite spins can be induced to move in opposite directions and line up on the sides of a chip.
Spintronics research is rapidly developing, says Awschalom, including the intriguing new field of molecular spintronics. Researchers want to use molecules with controllable properties to take the place of transistors in many applications. Because even a fairly chubby molecule is hundreds of times as small as today’s tiniest transistor, it’s an appealing idea.
Awschalom’s group is looking at ways in which molecules can be turned into “spin channels,” much as metal wires serve as charge channels. “By twisting and controlling the molecular bonds with light,” Awschalom says, “it is possible to operate on the electron spins as they move through the chemical structure.”
No matter how it turns out, one thing is clear: Spin control is no longer exclusively a Washington phenomenon.
Gordon Moore wrote an essay in Electronics magazine 40 years ago titled “Cramming More Components Onto Integrated Circuits.” The trend he predicted then—the number of transistors on a single chip would double every 18 months, later revised to two years—proved astonishingly accurate. Moore went on to become the cofounder of chip giant Intel and now, at 76, is chairman emeritus of its board.
How’s Moore’s law holding up?
M: It still has quite a ways to go. I’ve never been able to see more than three generations of the technology ahead. By a generation, we usually mean the point when we shrink the dimensions [of individual chip components] by a factor of 0.7. That was happening every three years. Recently, it’s closer to every two years. So the rate of evolution has actually increased, like the expansion of the universe. On that basis, there are three or four more generations to go at the current pace—at least another decade or so.
But isn’t the conventional silicon transistor doomed by fabrication problems as sizes shrink?
M: I don’t think so. I don’t see anything coming along that’s likely to replace silicon integrated circuits. I’d say the trend is going the other way: that the [silicon lithography] technology that has developed around the integrated circuit is now being adopted in several other areas, like microelectromechanical systems, microfluidic devices, chemical labs on a chip, and more. As for the limits of lithography, that limit keeps getting pushed away. We’re now talking about using extreme ultraviolet light [as a lithography beam], which would give us a 13-nanometer-wide light source, smaller than what we’re using now by more than a factor of 10. Eventually, there has to be a limit. Two or three molecular layers is all we’re dealing with now. Heat is a problem, but we’ve found ways to deal with it.
What will be on our desktops in a decade?
M: It’ll certainly be a computer of some sort, but it’ll have tremendous processing power and hopefully the software to take advantage of that. I wouldn’t be surprised if it were approaching a teraflop [a trillion mathematical operations per second, about 1,000 times faster than a high-powered PC in 2005]. What microelectronics has done to the cost of electronics has made gadgets like cell phones possible, and there’s still a lot of time for improvement. One place you’re going to see a big benefit is in home entertainment systems. They’re going to get much more integrated and drop in price a lot.
How about wireless technology?
M: Underdeveloped countries are going to benefit tremendously; that’s the area that’s really going to blossom. In this country, the main issue is how the spectrum gets allocated.
Has your personal life been changed by high-speed Web connections?
M: Well, I can’t get broadband in my house. I have a DSL line.