Quantum dots rewrite memory options
June will see the kick-off of a European project aiming to establish III-V quantum dot based devices as a serious candidate to succeed flash memory.
Led by Germany s University of Duisberg-Essen, each of four academic partners in the three-year project is due to receive €250,000 to perform the research.
Germany will provide a second partner institution in the form of the Technical University of Berlin. The Netherlands will be represented by the Technical University of Eindhoven and the UK by Lancaster University.
The overall project is entitled QD2D, short for “Coupling of single quantum dots to two-dimensional systems”.
QD2D is founded on the ideas of TU Berlin professor Dieter Bimberg and his former student Martin Geller, who is now leading the project at Duisberg-Essen.
The partners have since discussed the technology s potential with memory manufacturers in the form of insolvent Qimonda and the recently established Intel-STMicroelectronics collaboration Numonyx.
Netherlands-headquartered NXP Semiconductors has also shown interest, while UK-based defense technology firm Qinetiq is partnering with Lancaster University on its efforts.
“The aim of this project is to study a single quantum dot in detail at room temperature to understand the ultimate limits of charge-based memory,” commented Lancaster's Manus Hayne.
“ Flash is not a good name,” Hayne told compoundsemiconductor.net. “The guys from NXP labelled its performance as mediocre .”
Despite its current popularity, writing data to flash memory is slow and eventually damages the silicon chips it uses.
The type of memory that QD2D is working toward will surround quantum dots made from a narrow band-gap material like GaSb, with a wider band-gap semiconductor like GaAs.
The quantum dots act as a potential well that stores charge, similarly to how silicon flash memory stores data as charge between silicon dioxide barriers.
The design means that the write speed could be up to 1000 times faster than flash. Hayne concedes that erasing the data will still be slow, but says that this can be “engineered out in chip design”.
The QD2D collaborators also propose an alternative to the other major form of memory in use today, DRAM, that would be ten times faster.
Hayne points out that flash memory producers have shrunk their lithographic feature sizes especially rapidly to increase data capacities. Consequently this strategy may hit the physical scaling barriers expected for silicon by 2014.
“This is very soon,” Hayne said. “The memory guys really need a solution and there are a great many ideas around.”
“Phase change memory is definitely the leading candidate, but has been around for a while, and problems still remain. The memory industry is certainly still interested in other options - it is not a one-horse race.”
