Chinese Researchers Extend Memory Endurance 100-Fold in Semiconductor Advance

Chinese scientists have developed a technique that makes an emerging type of memory chip dramatically more durable, potentially removing a key reliability obstacle to its use in high-performance computing and future artificial intelligence systems, as the AI boom drives demand for more advanced semiconductors.

The team demonstrated more than 10 billion write cycles in wurtzite ferroelectrics — a class of materials that can switch between two electric states to store data. That result is roughly 100 times the endurance previously achieved with the same material and could help advance a promising next-generation memory technology, according to a report published Saturday by the local outlet Xian Daily.

The breakthrough brings ferroelectric memory closer to practical use in future computing hardware. The research, led by scientists at Xidian University in Xian in collaboration with City University of Hong Kong and Fudan University, was published Thursday in the journal Science.

In recent years, wurtzite ferroelectrics such as aluminium scandium nitride (AlScN) have drawn attention as promising next-generation memory materials because they offer rapid switching speeds and potentially low energy consumption. Crucially, AlScN is also compatible with existing semiconductor manufacturing processes, which could ease its integration into future memory devices.

However, the material has faced a significant hurdle: it deteriorates after repeated electrical switching. Existing AlScN devices have typically failed after roughly 100 million write cycles — far short of the billions required for commercial application.

The researchers identified the accumulation of nitrogen vacancies — spots where nitrogen atoms are missing — as a key cause of the relatively rapid failure rate, because these defects create paths for electricity to leak through the chip.

Wang Ruiqing, a doctoral researcher at Xidian University and one of the paper's authors, explained the problem in simple terms for Xian Daily.

"Think of ferroelectric material as a neatly planted cornfield, with nitrogen vacancies representing spots where seedlings are missing," Wang was quoted as saying.

The core issue, Wang explained, was not simply the number of these defects but their behaviour as the material was repeatedly switched. The nitrogen vacancies could move and cluster together, eventually forming pathways through the material that allowed electricity to leak and could lead to breakdown.

"In the past, researchers knew that the devices failed and could observe some of the symptoms, but no one had been able to explain at the atomic scale what was moving, how they were moving and how that movement ultimately caused failure," Wang said.

The team addressed the problem by designing a layered structure that confined the movement of the nitrogen vacancies. By restricting their movement and accumulation, the researchers were able to significantly slow the material's deterioration, allowing it to withstand more than 10 billion write cycles, they said.

The findings, though still at the laboratory stage, point to a path towards making these materials more durable for dense, low-power memory chips — an advance that comes as the global AI boom continues to fuel demand for faster and more reliable chips.