Stanford discovery could pave the way for super-fast, energy-efficient computing

Scientists have spent decades researching faster, more energy-efficient memory technologies for everything from large data centers to mobile sensors and other flexible electronic devices. Among the most promising data storage technologies is phase change memory, which is thousands of times faster than conventional hard drives but consumes a lot of electricity.

Today, engineers at Stanford University have overcome a key hurdle that has limited the widespread adoption of phase change memory. The results are published in a September 10 study in Science.

“People have long expected phase-change memory to replace much of the memory in our phones and laptops,” said Eric Pop, professor of electrical engineering and lead author of the study. “One of the reasons it was not adopted is that it requires more power to operate than competing memory technologies. In our study, we have shown that phase change memory may be at risk. both fast and energy efficient. “

Electrical resistance

Unlike conventional memory chips built with transistors and other hardware, a typical phase change memory device consists of a compound of three chemical elements – germanium, antimony, and tellurium (GST) – sandwiched between two metal electrodes.

Conventional devices, such as USB sticks, store data by turning the flow of electrons on and off, a process symbolized by 1s and 0s. In phase change memory, the 1s and 0s represent measurements of the electrical resistance in the GST material – how well it resists the flow of electricity.

“A typical phase change memory device can store two resistance states: a high resistance state 0 and a low resistance state 1,” said doctoral student Asir Intisar Khan, co-lead author of the study. “We can go from 1 to 0 and back in nanoseconds by using the heat of the electrical impulses generated by the electrodes.”

Heating to around 300 degrees Fahrenheit (150 degrees Celsius) turns the GST compound into a crystalline state with low electrical resistance. At around 1100 F (600 C), the crystalline atoms become disordered, turning some of the compound into an amorphous state with a much higher resistance. The large difference in resistance between amorphous and crystalline states is used to program memory and store data.

“This large change in resistance is reversible and can be induced by turning electrical pulses on and off,” Khan said.

“You can go back years later and read memory just by reading the resistance of each bit,” Pop said. “Also, once the memory is defined, it no longer uses energy, like a USB flash drive.”

‘Secret sauce’

But switching between states usually requires a lot of power, which could reduce battery life in mobile electronics.

To meet this challenge, the Stanford team set out to design a phase change memory cell that operates at low power and can be embedded on flexible plastic substrates commonly used in foldable smartphones, wearable body sensors, and the like. battery operated mobile electronic devices.

“These devices require low cost and low power consumption for the system to operate efficiently,” said co-lead author Alwin Daus, postdoctoral researcher. “But many flexible substrates lose their shape or even melt at around 390 F (200 C) and above.”

In the study, Daus and his colleagues found that a plastic substrate with low thermal conductivity can help reduce current flow in the memory cell, allowing it to operate efficiently.

“Our new device reduced the programming current density by a factor of 10 on flexible substrate and by a factor of 100 on rigid silicon,” said Pop. “Three ingredients went into our secret sauce: a superlattice made up of nanoscale layers of memory material, a pore cell – a nanoscale hole in which we stuffed the superlattice layers – and a thermally insulating flexible substrate. Together, they have dramatically improved fuel efficiency. “

Ultra-fast and flexible computing

The ability to install fast and energy efficient memory on mobile and flexible devices could enable a wide range of new technologies, such as real-time sensors for smart homes and biomedical monitors.

“The sensors have a high strain on battery life, and collecting raw data to send to the cloud is very energy inefficient,” Daus said. “If you can process the data locally, which requires memory, it would be very useful for the IoT implementation.”

Phase change memory could also usher in a new generation of ultra-fast computing.

“Computers today have separate chips for computing and memory,” Khan said. “They calculate the data in one place and store it in another. The data has to go back and forth, which is very inefficient from an energy perspective.”

Phase change memory could enable in-memory computing, which bridges the gap between computing and memory. In-memory computing would require a phase change device with multiple resistance states, each capable of storing memory.

“Typical phase change memory has two resistant states, high and low,” Khan said. “We programmed four stable resistance states, not just two, an important first step towards flexible in-memory computing.”

Phase change memory could also be used in large data centers, where data storage accounts for around 15% of electricity consumption.

“The big appeal of phase change memory is speed, but energy efficiency in electronics matters too,” Pop said. “It’s not just an afterthought. Anything we can do to make electronics energy efficient and extend battery life will have a huge impact.”