Research scientists Ding Wang (left) and Ping Wang (right) investigate the growth behavior of their ferroelectric semiconductor, which is deposited using the molecular beam epitaxy system visible on the left. Photo: Robert Coelius.Ferroelectric semiconductors are contenders for bridging mainstream computing with next-generation architectures, and now a team of researchers at the University of Michigan (U-M) has made versions that are just 5nm thick—a span of just 50 or so atoms.
This paves the way for integrating ferroelectric technologies with conventional components used in computers and smartphones, expanding artificial intelligence (AI) and sensing capabilities. It could also lead to battery-less devices for the Internet of Things (IoT). The researchers report their work in a paper in Applied Physics Letters.
"This will allow the realization of ultra-efficient, ultra-low-power, fully integrated devices with mainstream semiconductors," said Zetian Mi, U-M professor of electrical and computer engineering and co-corresponding author of the paper. "This will be very important for future AI and IoT-related devices."
Ferroelectric semiconductors stand out from other semiconductors because they can sustain an electrical polarization, like the electric version of magnetism. But unlike a fridge magnet, ferroelectric semiconductors can switch which end is positive and which is negative. This ability can be utilized in many ways—including for sensing light and acoustic vibrations, as well as harvesting them for energy.
"These ferroelectric devices could be self-powered," Mi said. "They can harvest ambient energy, which is very exciting."
They also offer a different way of storing and processing both classical and quantum information, as the two electrical polarization states can serve as the one and zero of digital computing. What is more, they can emulate the connections between neurons, which is the basis for both memory storage and information processing in the brain. Known as neuromorphic computing, this kind of architecture is ideal for supporting AI algorithms that process information through neural networks.
Storing energy as electrical polarization requires less energy than the capacitors in conventional computer memory, which constantly draw power or else lose the data they store. This kind of memory could also be more densely packed, increasing capacity, as well as being more robust to harsh environments, including extreme temperatures, humidity and radiation.
Mi's team had previously demonstrated ferroelectric behavior in a semiconductor made of aluminum nitride spiked with scandium, a metal sometimes used to fortify aluminum in performance bicycles and fighter jets. However, to use it in modern computing devices, they needed to be able to fabricate it as films thinner than 10nm.
They achieved this with a technique called molecular beam epitaxy, which is the same approach used to make the semiconductor crystals that drive the lasers in CD and DVD players. In a machine with strong steampunk vibes, they were able to lay down a crystal just 5nm thick—the smallest scale ever achieved. They did this by precisely controlling every layer of atoms in the ferroelectric semiconductor, as well as by minimizing the losses of atoms from the surface.
"By reducing the thickness, we showed that there is a high possibility that we can reduce the operation voltage," said Ding Wang, a research scientist in electrical and computer engineering, and first author of the paper. "This means we can reduce the size of the devices and reduce the power consumption during operation."
In addition, the nanoscale manufacturing improves the researchers' ability to study the fundamental properties of the material. This means they can discover the limits of the its performance at small sizes, and possibly open the way to its use in quantum technologies due to its unusual optical and acoustic properties.
"With this thinness, we can really explore the miniscule physics interactions," said Ping Wang, U-M research scientist in electrical and computer engineering. "This will help us to develop future quantum systems and quantum devices."
This story is adapted from material from the University of Michigan, with editorial changes made by Materials Today. The views expressed in this article do not necessarily represent those of Elsevier. Link to original source.