AS THE transistors etched into computer chips approach their limits of miniaturisation over the coming years, they could be replaced by devices that use magnetism instead of electricity to store and manipulate digital data.
Magnetic storage devices known as magnetic random access memory (MRAM) already exist. But now Alexandra Imre and colleagues in the nanoengineering department at the University of Notre Dame in Indiana have shown that magnets around 0.1 micrometres across can be arranged to produce the basic building blocks of a processor chip.
These blocks are known as logic gates. Typically a gate takes two inputs, each of which can be a 1 or a 0, and produces a predictable output. For example, what is known as an AND gate, gives an output of 1 when both input A and input B are 1; otherwise it gives an output of 0. An OR gate, in contrast, gives an output of 1 if either A or B is 1, and outputs a 0 only if both inputs are 0. There are also NAND gates, which give the inverse output of an AND gate, and NOR gates that give the inverse of an OR. It is by using combinations of these gates that a microprocessor performs binary calculations.
Advertisement
When transistors are used to make logic gates the inputs and outputs are represented by voltages. The first hint of how magnetic logic gates might work came five years ago from research by Russell Cowburn and colleagues at Imperial College London. They discovered that when rows of magnets each 110 nanometres in diameter were placed in a line, their magnetic fields coupled up so that a north-south-oriented magnet induced an oppositely oriented pole in the adjacent one, and so on down the line. If the orientation of the magnet at one end was flipped, that change was passed along the line from one magnet to the next.
Imre鈥檚 advance has been to adapt this arrangement to form a logic gate called a majority inverter (Science, vol 311, p 183). This takes three inputs, sets an intermediate magnet to represent the majority state of those inputs, and then inverts it to provide the output (see Diagram). So if most of the inputs are 1s, the output is 0. If most of the inputs are 0, the output is 1.
Imre notes that if one input is always held at 1, the gate performs a NOR operation on the other two inputs, and if one is always held at 0 the remaining inputs act as a NAND gate. It so happens that absolutely any logic circuit can be built by combining NAND and NOR gates. So by switching a NOR gate to a NAND or vice versa, processors built from nanomagnets could be reprogrammed on the fly to do different jobs, the researchers say. Simulations show that processing speeds of at least 100 megahertz should be possible using magnets 110 nanometres wide.
鈥淎bsolutely any logic circuit can be built using the nanomagnet arrangement鈥�
The gates made from the 110-nanometre magnets are a similar size to the transistor-based gates on today鈥檚 processors, though these chips run much faster, at speeds of several gigahertz. But the researchers say they can build gates with magnets as small as 5 nanometres wide, opening the door to more powerful chips that operate at higher speeds.
In a nanomagnetic chip the input signals would be created by currents flowing in a grid of fine conductors that would induce magnetic fields where they are needed to change the states of magnets. This arrangement would consume less power than transistors, so nanomagnet-based processors would suffer less from the overheating problems that limit the speed of today鈥檚 chips.
Much work remains to be done before such a technology becomes a reality. Not least amongst the challenges is protecting the delicate magnets from heat and extraneous magnetic fields. Cowburn, who is working with MRAM makers on developing the technology, suggests that an alloy called mu-metal, made up of nickel, iron, copper and molybedenum, could do the job.
