April 23, 2024

The ability of manufacturers to fit more components in the same space on a silicon chip is the key to the rapid growth of computing power. However, this progress is approaching the limits of physics. New materials are being investigated as possible replacements for the silicon chips that have been the core of the computer industry.

Developing new materials could also open up new design possibilities for individual components and their overall design. The ferroelectric field effect transistor (FE-FET) is a long-promised breakthrough. These devices can switch states quickly enough to perform computations, but they also can store those states without power, making them useful for long-term storage. The FE-FET device would double as a RAM/ROM device, making chips more efficient and powerful.

Manufacturing has been a significant obstacle to practical FE-FET devices. The ferroelectric materials are too hot for mass production and need suitable properties for producing silicon components.

A team of researchers from the University of Pennsylvania School of Engineering and Applied Science has solved this problem. Recent studies have shown that scandium-doped aluminum nitride, a material recently found to exhibit ferroelectricity (AlScN), can be used to create FE-FET and diode-memristor-type memory devices with commercially feasible properties.

Deep Jariwala (assistant electrical and systems engineering professor) and Xiwen Luu (grad student in his laboratory) led the studies. They worked with Troy Olsson (an associate professor in ESE) and Eric Stach (professor in the Department of Materials Science Engineering and director of the Laboratory for Research on the Structure of Matter).

Their findings were published in Nano Letters and Applied Physics Letters.

Jariwala says that one of the critical ways chip designers try to overcome the limitations of processing large amounts of data with silicon is by finding materials that allow memory components to be built directly onto the processor. This would allow for two-in-one designs. We knew that AlScN could be deposited at low temperatures, so we were able to combine memory and logic transistors. It was just a matter of integrating it with the rest.

Jariwala and his team discovered a promising two-dimensional material called molybdenum disulfide (or MoS2), a promising solution. The team tested the AlScN-based FE/FET device’s switching speed and stability using a single-layer MoS2. These results were published in the Nano Letters paper.

Jariwala says engineers have been working on FEFET memory since the 1960s. These devices can operate at meager power. It was essential to make these materials compatible with processors and last longer. Our 2D materials are the answer. They are thin enough that once a memory bit has been written into them, that information could be stored in the form of charge for many years.

Jariwala’s next step was to shrink the dimensions of their memory devices. Their Applied Physics Letters paper demonstrated that AlScN could be produced as thin as 20 nanometers. This reduces the device’s overall size and the voltage required.

Olsson says it needed to be clarified before the study whether AlScN would maintain ferroelectric properties when reduced to this size.

Stach adds, “we also found that by removing the MoS2 from the device and using AlScN with a two-terminal geometry, it can function as a diode memristor-like storage device.” Because they are less complicated than FEFET devices, diode memristors can be integrated commercially with fewer steps and parts.

Jariwala will continue working with his colleagues to develop manufacturing techniques that would enable these devices to be mass-produced and integrated into consumer electronics.

Leave a Reply

Your email address will not be published. Required fields are marked *