Scientists increase the performance of hafnium-based memory devices by adding aluminum to ferroelectric materials

The research team has significantly increased the data storage capacity of ferroelectric memory devices. Using hafnia-based ferroelectric materials and an innovative device design, their findings were published June 7 in the journal progress of science, represent a significant advance in memory technology. The team was led by Professor Jang-Sik Lee from the Department of Materials Science and Engineering and the Department of Semiconductor Engineering at Pohang University of Science and Technology (POSTECH).

With the exponential growth in data production and processing resulting from advances in electronics and artificial intelligence (AI), the importance of data storage technology has increased. NAND flash memory, one of the most widespread mass storage technologies, can store more data in the same area by arranging cells in a three-dimensional rather than planar structure. However, this approach relies on charge traps to store data, resulting in higher operating voltages and lower speeds.

Recently, hafnium-based ferroelectric memory has emerged as a promising next-generation memory technology. Hafnia (hafnium oxide) enables ferroelectric memories to operate at low voltages and high speeds. However, a significant challenge was the limited memory window for multi-level data storage.

Professor Lee’s team at POSTECH addressed this problem by introducing new materials and a novel device design. They increased the performance of hafnium-based memory devices by doping aluminum ferroelectric materials to create high-performance ferroelectric thin films.

Additionally, they replaced the conventional metal-ferroelectric-semiconductor (MFS) structure, in which the metal and ferroelectric materials that make up the device are simply arranged, with an innovative metal-ferroelectric-metal-ferroelectric-semiconductor (MFMFS) structure.

The team successfully controlled the voltage on each layer by adjusting the capacitance of the ferroelectric layers, which included tuning factors such as the thickness and area ratio of the metal-to-metal and metal-to-channel ferroelectric layers. Efficient use of the applied voltage to switch the ferroelectric material improved device performance and reduced energy consumption.

Conventional hafnium-based ferroelectric devices typically have a memory window of about 2 volts (V). In turn, the research team’s device achieved a memory window exceeding 10 V, enabling the use of Quad-Level Cell (QLC) technology, which stores 16 levels of data (4 bits) per unit transistor. It also showed high stability after more than one million cycles and operated at 10V or lower, much lower than the 18V required for NAND flash memory. Moreover, the team’s storage device showed stable data storage properties.

NAND flash memory programs its states using pulse step programming (ISPP), which leads to long programming times and complex circuits. In contrast, the assembly’s device enables rapid programming through one-time programming by controlling ferroelectric polarization switching.

Professor Lee of POSTECH commented: “We have created the technological foundation to overcome the limitations of existing memory devices and have pioneered a new research direction for hafnia-based ferroelectric memory.” He added: “With further research, our goal is to develop low-power, high-speed, high-density memory devices, contributing to solving power issues in data centers and artificial intelligence applications.”