Scientists close in on computers that work like the human brain

Scientists have been working since 2008 to develop technology based on memristors (short for memory resistors), which promise computers that need never boot up and function more akin to the human brain – like neurons, they can retain information and perform logic operations. Now scientists at Northwestern University have made a new breakthrough that may make possible brain-like computing capabilities.

Memristors are considered exciting for more than their potential to create brain-like computers. Unlike flash memory, they're fast. Unlike random access memory (RAM), they remember their state – whatever information they held – when they lose power. They also require less energy to operate, rarely crash, and are immune to radiation. The trouble is that they are two-terminal electronic devices, which results in them being tunable only through changes in the voltage applied externally.

The team at Northwestern transformed memristors from two-terminal to three-terminal electronic devices, thereby allowing their use in more complex electronic circuits and systems. The normal memristor setup as two-terminal devices allows only limited control over how electrical current flows through the system, but the third electrode used by the Northwestern researchers can act as a gate, finely controlling the resistance.

They achieved this by using a nanomaterial semiconductor called molybdenum disulfide, which has its "grains" of atoms arranged in a different direction to the memristors. A grain boundary sits between the molybdenum disulfide sheet and the metal electrode, acting as a kind of interface for the atoms. "These grain boundaries influence the flow of current, so they can serve as a means of tuning resistance," co-author Mark Hersam said.

The grain boundary moves when a large electric field is applied on the memristor, which causes a change in resistance. And that, Hersam noted, makes possible a new level of function and complexity that could lead to brain-like computing. "We are now actively exploring this possibility in the laboratory," he said.

A paper describing the research was published in the journal Nature Nanotechnology.