Engineers puzzling over which materials should form the basis of the next generation of flash memory chips appear to have had the answer close to hand. It has been lurking in their computer鈥檚 CD drawer all along.
At least, that鈥檚 the view of and his colleagues at Numonyx, a flash memory chip maker in Santa Clara, California.
The company believes that the chemicals that allow CDs and DVDs to be rewritten by laser beams are the perfect material for a future generation of ultra-dense flash memory chips. And they have built a chip to prove it works too.
Advertisement
Flash memory is at the heart of most of today鈥檚 popular gadgets, like cellphones, music players and digital cameras.
Limited lifespan
Unlike the random access memory (RAM) in a computer, flash is 鈥渘onvolatile鈥� 鈥� it keeps hold of the data it stores even when the power is turned off, so the phone鈥檚 software and photos, texts and music don鈥檛 get lost. To do this, flash memory stores binary data in the form of electric charge in a capacitor/transistor combination.
Today, most flash memory transistors are about 65 nanometres wide, with the next generation set for 45 nm. 鈥淏ut serious issues will begin to arise when dimensions get below 20 nanometres due to the retention of electric charge in the flash transistor,鈥� warns Atwood.
One of those issues is the usable lifetime of flash transistors, which is already limited by the build up of 鈥減arasitic鈥� charge on the transistor every time new data is written to it.
This limits their use to around 10,000 write-and-erase cycles. But with 20 nanometre transistors that effect worsens considerably, limiting lifetime even further.
Switched states
So if future flash is to have a long lifetime and store even more data for, say, solid state (hard-drive free) laptops and smarter smartphones, a new nonvolatile storage mechanism is needed.
In a CD or DVD, heat from a laser beam is used to make a glassy coating of GST (an alloy of tellurium, antimony and germanium) switch between disordered and crystalline states.
Each state, or phase, is set by the rate at which the laser allows the substance to cool. Importantly, each state has a different reflectivity, allowing it to store digital 0s and 1s in an easily readable way.
Usefully, changes to the material鈥檚 state and reflectivity can be created through electrical heating as well as laser light. And GST can be read using electricity too, because each state has unique resistance.
Scaling possibilities
Making use of these properties, Numonyx has devised a novel microchip memory cell in which a blob of GST sits atop a simple resistor. Different current pulses applied to the resistor heat the blob and change its state, and a second, much lower (non-heating) current can sense the blob鈥檚 resistance.
This Phase Change Memory (PCM) appears to work reliably. Numonyx 鈥� a spinoff from Intel 鈥� has built a 128-megabit flash memory and is currently allowing unnamed gadget makers to assess its potential.
But Atwood is confident it is the answer to making ever-denser flash: 鈥淲e believe PCM offers new scaling possibilities. You can change and stably hold the phase states in GST cell sizes down to 5 nanometres and below,鈥� he says.
And, without the parasitic charge problems of today鈥檚 flash, it can be written to at least 1 million times. And, unlike flash, individual bits in PCM can be changed at will: flash has to rewrite chunks of data each time, slowing the writing process.
Cost factor
Atwood concedes that gaining credibility for a new memory technology will be tough: there has hasn鈥檛 been a new type that achieved any mass take-up since 1972, when Intel devised a UV-erasable memory format called EPROM. The electrically erasable version of that became flash memory.
Malcolm Penn, an analyst with , a UK-based microchip market researcher, says Numonyx will have its work cut out proving PCM鈥檚 viability.
鈥淲e don鈥檛 see the classic field-effect transistor running out of steam for at least 10 years. While it鈥檚 true that some materials are misbehaving, new materials are always coming along to fix things,鈥� he says. 鈥淚t鈥檚 going to be cost, not technology, that will be the driving factor. That鈥檚 the bottom line Numonyx has to address.鈥�
Journal reference:
