Scientists at Stanford University have created an innovative phase-change memory, offering the potential to enhance the speed and efficiency of computers in handling substantial amounts of data.

Our computers are handling more and more data to speed up finding new drugs, predict weather and climate, train artificial intelligence, and more. To meet this demand, we need computer memory that is faster and uses less energy than ever before.

 

Scientists at Stanford have shown that a new material could make phase-change memory, which switches between high and low resistance states to create computer data, a better option for future AI and data-centric systems. Their technology, explained in Nature Communications, is fast, uses little power, is stable, durable, and can be made at temperatures suitable for commercial manufacturing.

 

"We're not just improving one thing; we're improving many things at the same time," said Eric Pop, a professor at Stanford. "This is a big step towards a universal memory."

 

A Quicker, Long-lasting Memory

 

Current computers store and process data separately in volatile and nonvolatile memory. Volatile memory is fast but loses data when the computer turns off, while nonvolatile memory is slower but can store information without constant power. Shifting data between these two locations can cause delays, and the processor has to wait for large amounts of data.

 

"It takes a lot of energy to move data back and forth, especially with today's computing workloads," said Xiangjin Wu, a doctoral candidate at Stanford. "With this memory, we hope to bring processing and memory closer together into one device, using less energy and time."

 

Creating a universal memory that can store data for a long time and process it quickly with low power has challenges. However, the new phase-change memory developed at Stanford is a significant step forward. The researchers hope it will encourage further development and adoption as a universal memory.

 

The memory relies on a material called GST467, made of four parts germanium, six parts antimony, and seven parts tellurium. The researchers found ways to layer this material with other thin materials in a superlattice structure, achieving good nonvolatile memory results.

 

"The unique composition of GST467 gives it a fast switching speed," said Asir Intisar Khan, co-lead author on the paper. "Integrating it within the superlattice structure in nanoscale devices enables low switching energy, good endurance, stability, and nonvolatility."

 

Setting New Standards

 

The GST467 superlattice meets several important benchmarks. It is highly stable over time, operates at below 1 volt for low power, and is significantly faster than a typical solid-state drive.

 

"A few other types of memory can be a bit faster, but they operate at higher voltage or higher power," said Eric Pop. "The fact that we're switching at a few tens of nanoseconds while operating below one volt is a big deal."

 

The superlattice also packs a good amount of memory cells into a small space. The researchers have shrunk the memory cells to 40 nanometers in diameter, less than half the size of a coronavirus. Although not as dense as it could be, the researchers are exploring ways to stack the memory in vertical layers, thanks to the superlattice's low fabrication temperature.

 

"The fabrication temperature is well below what you need," Pop said. "This type of memory can enable future 3D layering."