A Spin on Semiconductors Paves the Way for Brain-Inspired Computing

Billy Hurley, Digital Editorial Manager

“Our brain is a fantastic computer,” says Professor Tamalika Banerjee from the University of Groningen in the northern Netherlands.

The brain, after all, has the ability to process vast amounts of information with an energy efficiency far superior to that of today’s computers.

By integrating storage, memory, and processing into one unit, however, Banerjee and fellow physicists at the University of Groningen  hope their semiconductor device someday supports a parallel computing architecture that rivals the workings of the brain.

Banerjee’s research group studies spintronics, an area of semiconductor electronics that relies upon electron spins to increase data storage and transfer.

The team explores new mechanisms to create and manipulate spin transport across material interfaces, exploiting a phenomenon known as spin-orbit coupling. The interaction of a particle’s spin and orbital angular momentum leads to varying transition energy levels.

Banerjee’s latest device combines a niobium doped strontium titanate (SrTiO₃) semiconductor with ferromagnetic cobalt. This grouping allows an electric field to change the SrTiO₃ semiconductor from low to high resistance.

“n-doped SrTiO₃ is our workhorse,” Banerjee told Tech Briefs.

The semiconductor has unconventional charge transport capabilities compared to traditional options such as silicon or gallium arsenide. The heavy SrTiO₃ atoms cause spin orbit coupling at the interface, at room temperature.

Unlike silicon, the material also features a large, non-linear response in temperature, electric field, and frequency.

By additionally choosing a ferromagnetic material like cobalt (rather than the commonly used platinum or gold), the team demonstrated a spintronics phenomenon called tunneling anisotropic magnetoresistance, or TAMR.

The result: a “spin-memristor” with storage abilities.

“This means we can store additional information in a non-volatile way in the memristor, thus creating a very simple and elegant integrated spin-memristor device that operates at room temperature,” said Banerjee, the university’s professor of Spintronics of Functional Materials.

When a magnetic field was applied across the interface between cobalt and the semiconductor, the researchers demonstrated a 1.2 mV tunablity of the TAMR spin voltage.

Because the devices interface common transition ferromagnets onto the niobium doped strontium titanate, the fabrication process is a simpler one, says Banerjee.

By using the electric field to tune spin lifetime in semiconductors, the University of Groningen-developed technology opens up new approaches in device design and control.

With a combination of spin-based storage memory and computing, the devices could someday be used in a brain-like computer architecture.

“We are actively looking into this,” Banerjee told Tech Briefs. “They can be used as synapses in a neuromorphic computing architecture but also in new spintronic devices where we can control the magnetoresponse using electric fields.”

Banerjee's spin-memristor is relatively simple. Scaling up the electronics component to a full computing architecture is the next big step.

The researchers are currently learning more about what exactly happens at the interface of cobalt and the strontium semiconductor – a complicated interaction.

'Once we understand it better, we will be able to improve the performance of the system,” said Banerjee.

The professor is also considering how to build a more complex system with spin-memristors, testing actual algorithms for specific cognition capabilities of the human brain.

What do you think? Can computing surpass the capabilities of the human brain? Share your comments below.

---