Researchers have been able to make structures far smaller than those possible with current computer chip fabrication methods – the problem is they can’t yet make complex, working devices out of them.
Now a team of IBM researchers has found a way to use existing mass-production methods to create controllers for groups of tiny wires, an advance they hope will lead to memory chips four times denser than current ones. The advance should also mean significant cost savings, says Kailash Gopalakrishnan, an IBM researcher who presented the findings this week at the IEEE’s International Electron Devices Meeting in Washington, DC.
Although memory density has been improving steadily, this advance may be a way to “jump ahead of the curve,” Gopalakrishnan said.
Connecting microfabricated circuits to nanoscale structures has proven difficult. Making complex devices out of nano structures has been so elusive because each wire needed its own, relatively bulky connection. Now just one connection, with three elements, can control multiple wires, allowing the wires to be packed together much more tightly.
“It’s a very elegant idea,” says Mark Reed, professor of electrical engineering and applied physics at Yale University. “This interface problem has been there for awhile, and I think this is a wonderful way to get around it. This is a core idea that will have some important implications for nanostructures.”
In the IBM research, standard methods are used to pattern a three-element controller. One element connects to the end of a set of parallel wires and supplies electrons. The other two sit on either side of the wires and emit electric fields that together can be used to shut down the current passing through all but any one of the wires. So far, the researchers have built connections involving four parallel wires, but their data suggests the same system can control eight wires.
Being able to select a particular wire means signals can be addressed – a key element in random access memory. For example, any cell in a memory grid can be selected, to read or write to, by activating two perpendicular lines, like tracing rows and columns on a map.
So far, the IBM interface method has not been used to make a working memory cell. But this could happen within the year, says Gopalakrishnan. And, if all goes well, he says, more complex memory devices will follow in years to come. Other applications may also be possible, such as computer processing. “Memory is just part of it,” said Gopalakrishnan. “It’s a very broad concept.”
This new data poisoning tool lets artists fight back against generative AI
The tool, called Nightshade, messes up training data in ways that could cause serious damage to image-generating AI models.
Rogue superintelligence and merging with machines: Inside the mind of OpenAI’s chief scientist
An exclusive conversation with Ilya Sutskever on his fears for the future of AI and why they’ve made him change the focus of his life’s work.
The Biggest Questions: What is death?
New neuroscience is challenging our understanding of the dying process—bringing opportunities for the living.
Driving companywide efficiencies with AI
Advanced AI and ML capabilities revolutionize how administrative and operations tasks are done.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.