EVEN if you’re not normally fazed by mathematics, Dirac’s relativistic wave equation is likely to leave you speechless. Apart from a zero and an equals sign, it doesn’t seem to contain any run-of-the-mill mathematical symbols. There’s i, representing the square root of minus one, and some weird squiggle that looks vaguely like something out of the calculus, but they don’t do much to explain what’s going on.
The equation, first published by the British physicist Paul Dirac exactly 70 years ago, is the mathematical offspring of the marriage of quantum theory, the physics of matter on the very small scale, and Einstein’s special theory of relativity, the science of motion. It is one of the crown jewels of theoretical physics. But buried within its esoteric symbolism lies the foundation of a computing revolution. Dirac’s equation shows that electrons-the fundamental constituents of electricity-are more than merely charge-carrying specks of matter. It shows that they also possess a uniquely quantum-mechanical property known as spin.
Of all the weird properties of matter revealed by quantum theory, spin is one of the strangest. Dirac used the idea that particles have a kind of internal rotation to help him to make sense of the quantum world. But it was immediately clear that this spin is like nothing that exists in the everyday world (see “The spin doctors”). Now scientists are using spin to control electrons in an entirely new way. Their work is part of a new field of electronics called spintronics, and the tiny devices they are building have the potential to change forever the way computer memories and processors work.
Advertisement
Conventional electronics exploits the fact that when atoms of semiconducting materials like silicon are brought together, the electron energy levels become blurred and form two bands separated by an energy gap. When electrons jump across this gap, they leave behind vacancies or holes, which behave like positively charged particles. Electronic devices work by controlling the electric field produced by the ebb and flow of these two types of charge carrier. And since the invention of the transistor half a century ago, this principle has worked brilliantly well.
But now there is a new way to create two types of charge carrier: by exploiting the spin of electrons. Ordinary electric current is made up of a mix of electrons with spins oriented at random. But in a magnetic field, the spins become oriented either up or down.
By dictating the distribution of spin-up and spin-down electrons, it is possible to control currents in a similar way to ordinary electronic switches and transistors, says Mark Johnson of the US Naval Research Laboratory in Washington DC, one of the leading lights in the emerging field of spintronics.
Crucially, it’s a trick that doesn’t just work in fancy semiconductors. Pioneering experiments performed in the mid-1980s by Johnson and his doctoral thesis adviser Robert Silsbee at Cornell University in Ithaca, New York, proved that it is possible to alter the relative populations of electrons in the two spin states in common-or-garden metals such as copper, silver and aluminium. And that, says Johnson, is one of the things that makes spintronics so exciting, because it promises to solve one of the biggest hurdles that conventional electronics will face-what happens when we start making devices so small that they contain only a handful of electrons?
Close to the limit
This hurdle is closer than you might think, says Johnson. “The density of charge carriers in semiconductors is around 1019 per cubic centimetre-which is a factor of around 100 to 100 000 lower than in metals,” he says, doing a little back-of-the-envelope calculation to show the implications. “So suppose you want to build a conventional electronic device with dimensions around 0.01 micrometres-a tenth of what they do today. That’s around 10-18 cubic centimetres, so at most it will contain just 10 or so charge carriers. That’s obviously close to the limit, since you can’t make a device with less than one electron. But with metal-based devices, you could make them smaller by at least a factor of 10 and still have enough carriers.”
It gets better. Magnetic fields tend to become concentrated as components are shrunk. And since spintronic devices use magnetic fields to create and hang on to the two populations of charge carriers, their performance actually improves the smaller you make them. Johnson says spin transistors could have a volume of less than 0.01 cubic micrometres. “That’s a packing density about 100 times greater than state-of-the-art silicon.”
The prospect of boosting the power of computer chips by a factor of 100 on technology no fancier than ultrathin strips of metal is certainly appealing, but is it more than a theoretical pipe dream? Yes, says Johnson-and he should know. He has already built a crude spintronic transistor, and has persuaded others to start taking spintronics seriously. With funding from the US Department of Defense Advanced Research Projects Agency (DARPA), some electronics companies are starting to put the theoretical advantages of spintronics to practical use.
In the early 1990s, Johnson worked at the laboratories of Bellcore in New Jersey. There he showed that one way to alter the balance of spin-up and spin-down electrons was to pass them through a film of permanently magnetic, ferromagnetic material, such as iron or cobalt just a few hundred atoms thick. “Roughly speaking, ferromagnetic materials are magnetic because the number of spin-up and spin-down electrons in them is not equal,” he explains. When a current flows through a thin ferromagnetic film, the film’s magnetic field makes the incoming electrons take up the same spin-up and spin-down distribution. “And when they come out the other side, they carry this polarisation with them.”
This discovery opened the way to the construction of spintronic analogues of familiar electronic devices. The most basic is the switch-the on/off, one/zero component that underlies the whole of digital electronics. A spintronic switch could hardly be simpler, comprising of a layer of highly conductive gold sandwiched between two thin films of ferromagnetic material.
Unpolarised electrons sent into the first thin film adopt its mix of spin-up/spin-down varieties, and then zoom across the “filling” of gold before running into the second ferromagnetic film on the other side. Whether they get any further depends on their spin. Only those with spins that are aligned to the magnetic field of the film can continue on their way. Effectively, the second film behaves like a spin filter, with the size of the current that can pass depending on the relative numbers of spin-up and spin-down electrons created by the first strip. “It is similar to crossing and uncrossing polarised light filters, but the analogy is not strictly the same,” says Johnson (see Diagram) .
Flash memory
Flipping the switch from off to on-or, in digital terms, from 0 to 1-is then simply a matter of altering the magnetic orientation of the second strip from up to down. This can be done by passing a current through a small wire on top of the ferromagnetic film. The current generates a magnetic field strong enough to flip the orientation of the ferromagnet.
These switches are causing particular excitement among electronics researchers because they can store information. Ferromagnetic materials behave effectively like small permanent magnets, and once their field orientation has been flipped, they stay flipped-giving a way to store 1s or 0s without the need for any external power. They can be used to make incredibly fast yet “nonvolatile” electronic memories that faithfully retain their content even after the power has been switched off. Reading the data is simply a matter of passing current through a switch to see whether it is set to on or off. “The only silicon device that can do this is the so-called flash memory cell,” says Johnson. “But that takes milliseconds to store data-a million times longer than spintronic devices.”
Spintronic memories are even able to cope with intense radiation, because the magnetisation of their ferromagnetic films is unaffected by charged particles that would trash the components of semiconductor devices. What’s more, says Johnson, spintronic switches consume far less power than semiconductor switches, allowing far more of them to be crammed together without fear of overheating. Even the semiconductors in ordinary desktop PCs are liable to overheat.
It is this unique combination of abilities that has made the US Department of Defense take spintronics very seriously indeed. Over the past 18 months, DARPA has poured more than $50 million into spintronics research at commercial and academic centres throughout the US. “Satellites were the primary motivation, but there are many other military systems that need nonvolatile, extremely environmentally robust memories,” says Stuart Wolf, who manages the DARPA programme. “Their commercial potential as faster, lower-power replacements for flash memory also makes them attractive to us.” If spintronic chips can be produced commercially they will be cheaper for everyone.
The objective of the DARPA programme is to make spintronic memory devices that have the speed and storage capacity of the very best conventional semiconductor memories. The bulk of the funding has gone into exploiting the phenomenon known as giant magnetoresistance (GMR), in which devices with many ferromagnetic layers are used to control the flow of the spinning electrons through them (“Giants in their field”, 91av, 10 February 1996, p 34). According to Wolf, progress so far has been rapid. “Honeywell has demonstrated a fully functional 16-kilobit memory with an access time of less than 100 nanoseconds, and is well on its way to a 256-kilobit memory,” he says. “Motorola’s and IBM’s research efforts are also proceeding very well, and they’re on target to have some devices to demonstrate by the end of this summer.”
Semiconductor sandwich
PC users may start to benefit from this startling progress early in the next century. “Spintronic devices would allow us to replace the redundancy and cost of having two memory systems-RAM [random access memory] and hard discs-by a single, fast spintronic device that does the work of both,” says Johnson. Accessing data on a hard disc is a time-consuming business. This is why computers take so long to boot up. With a spintronic memory, they would be ready to work almost immediately. “Making these devices is pretty simple, and people are talking about making 1 gigabyte of nonvolatile RAM for the same cost as today’s 1-gigabyte hard drives,” says Johnson. That is, for well under $200.
Permanent memories look set to be the first commercial spintronic devices, but some research teams are investigating other ways to exploit electron spin. At the California Institute of Technology in Pasadena, Michael Roukes and his colleagues are experimenting with spintronic devices made from both metals and semiconductors.
The Caltech approach is to put a semiconductor “filling” between the two ferromagnetic films in the spintronic sandwich. The plan is to exploit the fact that electrons trapped inside the specially constructed semiconductor layer encounter very few impurities, and thus behave like the particles of a gas. Being able to move around more freely, this should make them much more responsive to incoming signals, allowing them to be made smaller.
In particular, it should make possible hybrid spintronic transistors made from metals and semiconductors that can be packed together far more tightly than they can in today’s equivalents. This is hugely significant, since spintronic devices made from semiconducting materials could easily be manufactured in today’s chip factories. New manufacturing plants would cost billions of dollars to build.
“To begin with, spintronic transistors will probably be most interesting as nonvolatile memory devices integrated into microprocessors,” says team member Franklin Monzon. “Later on, they might find applications as logic elements as well.” But just how much further spin-based electronics will go is still unclear, says Monzon.
One worry is that much of the work on semiconducting spintronic devices has been carried out at temperatures only a few degrees above absolute zero. “The spin of the electron is an inherently quantum phenomenon, and it’s more easily manipulated at low temperatures, where effects that would otherwise interfere with spin dynamics, such as scattering, are frozen out,” explains Monzon. “A device barely operating at 4.2 kelvin would likely just not work at all at room temperature, he says. The big challenge ahead for researchers is to find a way of applying what they have learnt at low temperatures to devices that will function at higher temperatures. The fact that the metal spintronic memories retain their impressive abilities at room temperature provides grounds for hope.
Great expectations
Even so, Monzon sees other problems ahead. “The biggest are material and interface problems,” he says. “For example, it’s still not clear what combination of materials we need to best control the behaviour of the electrons inside these devices.”
Despite being the father of spintronics, Johnson is also keen to play down the hype. His own experiences lead him to suggest that the major barrier preventing spintronics really taking off is not technical, but psychological. “For my devices, the biggest problem has been a reluctance among the semiconductor community to get involved in what they see as `disruptive’ technology.” He also believes that there is a reluctance among other companies to get involved in technologies they haven’t seen before. Johnson calls it “the `not invented here’ syndrome.”
But Johnson thinks the rapid progress made in spintronics speaks volumes for their potential. “The devices we’re working on have a research investment of only a few person-years, yet we already have prototypes that are competitive in terms of size and speed with silicon devices which have had tens if not hundreds of thousands of person-years of research invested. For a fledgling field, we think we’re doing well.”
* * *
The spin doctors
WHEN the idea of electron spin was first mooted in 1925 by Ralph Kronig, a postgraduate in the US, even the brilliant quantum theorist Wolfgang Pauli could not get his head around it: “Very clever,” he declared. “But of course it has nothing to do with reality.”
Pauli’s response seems offhand, but the idea of a spinning electron is weirder than it appears at first glance. First, its rate of spin cannot take on just any value. Quantum mechanics demands that it can only come in “packets” of angular momentum equal to h (pronounced “h bar”), where h is Planck’s constant divided by 2π. And that instantly raises another problem: a simple calculation reveals that the electron must be gyrating at more than 100 times faster than the speed of light.
There is no real paradox here, simply a warning that electrons are not tiny spinning balls. Like everything else in the subatomic world, they are something far more ineffable. Over the years, physicists have learnt to abandon any mental image of spin, putting their trust instead in mathematics. It is a policy that has brought rich rewards.
Spin is one of the basic properties of electrons governed by the Pauli exclusion principle, which states that no two electrons in an atom can share the same mix of quantum properties. This is the fundamental principle behind the periodic table, one of the cornerstones of chemistry.
Spin has in fact proved to be a fundamental property of all subatomic particles, which can be classed as either fermions, which have an amount of spin measured in half integers of h (1/2, 3/2, and so on), or as bosons, measured in whole integers of spin units (0, 1, 2 and so on). And, intriguingly, it turns out that all the particles of matter- electrons, protons, neutrons-are fermions, while all the particles linked to fundamental forces-such as the photon, which carries the electromagnetic force-are bosons.
This has convinced those searching for a Theory of Everything that there must be a way to bridge the great spin schism, and thus unify all particles. (“Are superparticles poised to make their debut?” 91av, 2 September 1995, p 17). Current fashion favours “supersymmetry”, whose mathematics puts all particles on the same footing, regardless of spin. So far, however, solid evidence that the schism really is only spin-deep has yet to emerge.
