Titanium oxide film ends four-decade quest for basic circuit component
US-based scientists have used nanoscale solid oxide films to create a new circuit element, a memristor, which researchers have been hunting for almost four decades.
The device could help to shrink silicon chips even further and spawn new types of memory storage that allow computers to start without booting up.
The memristor is designed so that its resistance to electricity flow depends on the amount of charge that has recently passed through it. It works because its internal structure rearranges as charge flows. That makes it rather like a resistor with memory - hence the contraction, ’memristor’.
’This will launch a decade of research into solid-state chemistry, studying the connection between the movement of ions in the solid state and electronics,’ says project leader Stan Williams of Hewlett-Packard labs, California.
As Williams explains, the working prototype is made of a five-nanometre thick film of titanium dioxide (TiO2), sandwiched between platinum electrodes. The film is divided into two parts: a lower layer of pure titanium dioxide, which has a very high resistance, and an upper layer of titanium oxide with positively-charged oxygen vacancies as dopants in the structure.
Applying a positive voltage to the upper platinum electrode drives some charged dopants into the lower layer. This re-arrangement of the film’s internal structure allows current to flow through. The dopants can also be pulled back into the upper layer, gradually blocking current flow, though not exactly retracing the path by which current was switched on. The current flow through the memristor therefore depends on the voltage applied across it in the past.
’Memristor effects get stronger as you build smaller devices, because the dopants don’t have as far to flow,’ explains Williams. Chip-makers hope to reduce the size of transistors down to a width or ’half-pitch’ of 20nm. But Williams says his team have already built memristors at 15nm half-pitch - and believe they can take that down to 4nm. The team have also built hybrid circuits with arrays of transistors driving current through memristors.
’The memristor might provide a new path onwards and downwards to ever greater processor density,’ comments chemist James Tour, of Rice University, Texas. He says the current-voltage behaviour of Williams’ memristor explains a host of puzzling effects researchers have seen in nanoscale electronics, in which changing internal structure significantly alters the resistance of thin films. ’It turns out that people have been seeing and reporting these effects for over 50 years, but didn’t understand them,’ Williams adds.
The memristor’s existence was first proposed in 1971 by the electrical engineer Leon Chua, who noticed that the three traditional circuit elements which receive or store electricity (the resistor, capacitor and inductor) should logically be joined by a fourth, which he christened the memristor, in order to fully connect the relationships between the basic variables describing a circuit (charge, voltage, magnetic flux, and current). But no physical description of Chua’s suggestion had been obtained until now.
’This is a real breakthrough by the Williams group, which will make a huge economic as well as scientific impact,’ Chua told Chemistry World. ’Everyone agrees that there will be no more room for improving transistor size within 10 years, so this is almost like a salvation that could extend Moore’s law,’ he adds, referring to the oft-quoted prediction of Intel co-founder Gordon Moore, that the number of transistors crammed on to a chip would double roughly every two years.
One application of the memristor, Williams hopes, could be to create a new type of ultradense non-volatile random access memory (RAM); where bits are stored as the ’on’ and ’off’ states of a series of memristors. ’This means your computer wouldn’t have to boot up when you turn it on,’ he says. Combining memristors in an architecture known as a cross-bar latch can also emulate the workings of a transistor, though using different underlying physics. ’We’re also hoping to create some innovative logic applications, such as learning networks requiring a synapse-like function,’ Williams adds - though final applications will have to wait until a large-scale array of densely packed memristors is proven to work.
Richard Van Noorden
Enjoy this story? Spread the word using the ’tools’ menu on the left.
D B Strukov et al, Nature, 2008, 453, 80 (DOI: 10.1038/nature06932)