You don't have to be a chip engineer to be sick of Moore's Law, that hoary old observation that about every two years the number of transistors on a chip doubles. I'd like to propose Goodwins' Corollary: every two years the number of times Moore's Law is mentioned at chip conferences is itself doubled. Nonetheless, the whole IT industry depends on regular and revolutionary advances in basic chip technology -- and now, many physicists are getting excited about a very different kind of electronics called spintronics. This takes the fundamental tool of electronics -- the electron -- and uses it in a very different way.
For a familiar friend, the electron is a slippery customer. It lures us into a sense of security by having commonplace attributes like mass, electric charge and spin. What it doesn't have is a body -- mathematically, an electron is an infinitely small dot spreading out through space. Any comparisons made with objects we can imagine are just convenient metaphors: like all subatomic particles, electrons are more alien than anything you'll ever find in science fiction.
Fortunately for physicists and the rest of us, while we don't know what electrons are we do know what they do. Electrons can be persuaded to move through wires by changing the charge at its ends: that's the basis for electricity, and by connecting wires in various ways through switches the electrons can move power from place to place. Semiconductors take small numbers of electrons and use them to switch much larger numbers: thus signals change other signals, and data can be used to create other data.
So far, the electron's spin has been ignored. The electron can only spin at one speed, and then only in one of two directions -- spin up and spin down -- and it doesn't really spin at all. It's just that the mathematics of that property of the electron behaves like the mathematics of a spinning top, and so the idea makes a handy analogy.
However, electron spin is very familiar to us in another way -- it's at the heart of magnetism. Most materials have an equal mix of electrons spinning in both directions, and thus have no magnetic field. If you have a preponderance of electrons spinning in a particular direction, then you get a magnetic field -- and conversely, if you magnetise something you change the ratio of spins. In that sense, all magnetic media from the very first audio recorders to the latest half-terabyte hard disks use spin engineering. Hard disks that use giant magnetoresistive heads -- these days, that's almost all of them -- go a step further and use an explicit spintronic effect, where the different spins of electrons create a changing resistance in a detector.
The most exciting thing about spin is that it sticks around. If you build an electronic circuit based on the flow of electrons -- current, in other words -- it only works while that flow continues. That needs a constant input of energy, and is almost always somewhat inefficient. Circuits that rely on stored charge have areas with an excess of electrons and areas with a deficit, and over time the electrons themselves will flow to even out the difference. Once an electron is spinning in a particular direction, though, it stays spinning until something happens to disturb it. This is exploited to the extreme in paleomagnetism, where scientists read magnetic signatures laid down by the Earth's field in rocks. Sequences have been calibrated back to the Jurassic period and useful results obtained from basement rocks laid down four billion years ago: sufficiently non-volatile for most purposes.
IBM, Motorola and other companies have developed MRAM -- magnetic memory -- that electronically manipulates spin for storage purposes, and these can combine non-volatility with high densities and DRAM-like speeds. We may see MRAM in mobile phones and the like sometime in 2004: previous attempts at commercialisation have hit production difficulties, but the companies involved this time around are being extremely bullish.
The second most exciting thing about spin is that it can take vanishingly small amounts of power and time to set up. An electron doesn't need to move to have its spin changed, whereas traditional circuits rely on vast electronic movement. Regions of a conductor with electrons spin-aligned with each other tend to reject currents of electrons with a different spin, which can be a very efficient way to let information modulate signals -- the basic way computers represent and act on data. One of the most active areas of research at the moment is how to integrate spin injection and detection with existing techniques: numerous theoretical and practical designs exist for various forms of spintronic transistors and other components, but the basic building block for spintronic logic is yet to come. When it does, it promises much faster switching, much higher densities and much lower power consumption -- purely because there's much less electron manipulation needed. At the logical lower limit, you can build a transistor that works on a single electron, and SEDs -- Single Electron Devices -- that use spin have been demonstrated under lab conditions.
There's also an aspect of electron spin that may in time overshadow the rest. As with all subatomic particles, quantum physics applies directly to the electron. It can spin in either direction, but until the direction is known by measuring the electron can be said to be spinning in both directions simultaneously -- an uncollapsed state. Quantum computing works by encoding multiple bits of information onto just such a state, forming a quantum bit or qubit. A spin transistor can theoretically form a physical framework within which qubits encoded onto uncollapsed spin can be handled. This may open the way for enormously effective parallel computations, where very complex problems are solved by the near-instantaneous decomposition of chains of qubits, but in principle one must remain uncertain.
This is a long way away from flash memory and faster ways to run Microsoft Office. However, one of the most exciting aspects of spintronics is that it can use many of the same techniques as ordinary solid-state semiconductor physics. If the right materials can be found that include magnetic properties alongside normal charge-based transistor techniques, then spin can be introduced to devices very rapidly. Researchers at the Royal Institute of Technology in Stockholm say that they've created just such a material -- zinc oxide with manganese doping. Zinc oxide is a widely used semiconductor and optically active material; manganese is a magnetic metal that is just the right size to fit into the zinc oxide structure without distorting it.
There is a consensus among researchers that practical spintronic devices will be in production in five to 10 years. Consensus has been wrong before, but the field is being intensely investigated and while many problems remain none seem to be showstoppers. By adding an entirely new dimension to electronics, spintronics may create as big a step change in our technologies as did the invention of the transistor.