Huazhong University: AlScN ferroelectric transistors with IGO channel show 20.6 V memory window in lab devices
Ferroelectric transistors on wurtzite AlScN with an ultrathin IGO channel reached a 20.6 V memory window, 10^6 endurance and 64 states. The MFMIS structure lowered interface trap density.
A team from Huazhong University of Science and Technology (Wuhan) and The Hong Kong Polytechnic University has reported ferroelectric field-effect transistors (FeFETs) built on wurtzite aluminum scandium nitride (AlScN) with an ultrathin indium gallium oxide (IGO) channel. The paper, which describes annealing-free devices for neuromorphic computing, appeared in Nature Communications on 3 November 2026. According to the authors, the best of three tested structures reached a memory window of up to 20.6 V and an endurance of 10^6. doi.org
Nature Communicationsdoi.orgOpen source ↗
Why AlScN and what limits it
Wurtzite AlScN ferroelectrics are considered candidates for non-volatile memory (NVM). The reasons given are a high remanent polarization (Pr above 100 μC·cm−2) and compatibility with back-end-of-line (BEOL) processes. doi.org
Nature Communicationsdoi.orgOpen source ↗
At the same time, the authors note that building high-performance wurtzite FeFETs remains a significant challenge. The obstacles are a high interface trap density (Dit) and a large variation of threshold voltage (VTH). doi.org
Nature Communicationsdoi.orgOpen source ↗
To examine this, the group made AlScN FeFETs with an ultrathin IGO channel in three different structures. The authors report that polarization switching in these structures follows distinct mechanisms. doi.org
Nature Communicationsdoi.orgOpen source ↗
Results for the MFMIS structure
FeFETs with a metal–ferroelectric–metal–insulator–semiconductor (MFMIS) structure showed high interface quality with low Dit and a stable VTH, according to the authors. In this structure the memory window reached 20.6 V, and the endurance was 10^6. doi.org
Nature Communicationsdoi.orgOpen source ↗
The devices also operated stably in a 6-bit mode with 64 distinct states. The authors present this as evidence that the devices are feasible for multilevel storage. doi.org
Nature Communicationsdoi.orgOpen source ↗
The team further demonstrated the use of the FeFET in neuromorphic networks, with a recognition accuracy of 93.02%. doi.org
Nature Communicationsdoi.orgOpen source ↗
The work is a laboratory study of individual devices. Device fabrication used a magnetron sputtering system at the Wuhan National Laboratory for Optoelectronics. The corresponding authors are Yang Chai and Xuefei Li.
