As all popular electronics textbooks will verify, any passive circuit can be created with a combination of just three standard components: a resistor, which opposes charge flow; an inductor, which opposes any change in the flow of charge; and a capacitor, which stores charge. But, if research by physicists in the US is anything to go by, the textbooks may have to be appended with a fourth standard component: a “memristor”.

Memristance will herald a paradigm shift not only in engineering, but also in the sciences Leon Chua, University of California at Berkeley

In simple terms, a memristor “remembers” the amount of charge that has flowed through it and as a result changes its resistance. The effect was predicted in 1971 by electronics engineer Leon Chua, but the only clues that it actually exists have been in the reports of strange “hysterisis” loops in the current–voltage relationships of thin-film devices. This means that when the voltage increases the current follows a different relationship to when the voltage decreases.

“[Scientists] essentially reported this behaviour as a curiosity or a mystery with some attempt at a physical explanation for the observation,” says Stanley Williams of Hewlett Packard Labs in Palo Alto, California. “No-one connected what they were seeing to memrisistance.” Now, Williams and colleagues from Hewlett Packard have made the first memristors.

Model behaviour

To make their memristors, Williams’s team first considered how memristance might originate at the atomic level. They came up with an analytical model of a memristor that consists of a thin piece semiconductor containing two different regions: a highly doped region, which has a low resistance, and a zero-doped region, which has a high resistance. When a voltage is applied across the semiconductor, it causes some of the dopants to drift so that the combined resistance changes, thereby producing the characteristic hysterisis effect of memristance.

To put this model into practice, Williams’s team attached a layer of doped titanium dioxide to a layer of undoped titanium dioxide. Through current–voltage measurements, they found that it did indeed exhibit the hysterisis effect of memrisistance (Nature 453 80).

Chua told that he is “truly impressed” that the Californian team has proved his theory. “Most of the anomalous behaviours that have been widely reported in the nano-electronics literature over the last decade can now be understood as simply the manifestation of memristive dynamics,” he says. “It will herald a paradigm shift not only in engineering, but also in the sciences.”

Williams says his team has already made and tested thousands of memristors, and has even used them in circuits containing integrated circuits. Because the hysterisis of memristors makes them able to operate like a switch, the team are now looking at how they can exploit memristance for tasks normally reserved for digital-logic electronics. These include a new form of non-volatile random access memory, or even a device that can simulate synapses — that is, junctions between neurons — in the brain.