'Memristor' electronics could be future for integrated devices

Engineers at UCL created a phase-changing memresitor device entirely in silicon that can operate at ambient conditions.

‘We’re reaching the limits of what the current generation of flash technology can do both in limits of scaling and in reliability — as you make them smaller you start having problems trying to dissipate heat from the system,’ said Dr Tony Kenyon, UCL Electronic and Electrical Engineering.

Although a number of novel competing technologies have been proposed, devices based on memory resistors have a emerged as an early leader.

In these ‘memristor’ devices, resistance changes depending on how much current has passed through it. They need just a thousandth of the energy and are around a hundred times faster than standard flash memory chips.

However, previous memristor devices have used expensive materials that require a vacuum to operate effectively.

‘We can control the switching in ambient conditions and it’s all purely in silicon,’ said Kenyon. ‘There are some people who work on resistive switching by diffusing metals in silicon. We don’t do any of that, it’s inherent in the material itself.’

The technology was discovered by accident while engineers were using silicon oxide material to produce silicon-based LEDs. At first they noticed that their devices appeared to be unstable, but later realised they flipped between various conducting and non-conducting states very predictably.

They are now working on making a quartz device with a view to developing transparent electronics.

‘There is an awful lot of resistance in the semiconductor industry to any change or introduction of any material that is not what they generally work with,’ said Kenyon. ‘But I’m much more optimistic because this doesn’t require any change in any of the processing steps that people already perform to create silicon chips.’

Instead of just switching between on-off states, the device could be engineered to deliver a continuous change in the resistance, so the more current flows through it the lower the resistance becomes. This could provide the basis for computing power in a way similar to biological neurons.

‘These devices can be used to both store and process, so that means you don’t have to keep moving data back and forwards between processor and memory — it can all be done in situ,’ said Kenyon.