Wurtzite ferroelectric memory is an experimental nonvolatile-memory technology that stores information in the direction of electrical polarization inside a wurtzite-structured material such as aluminum scandium nitride (AlScN). It matters because researchers at Xidian University have now reported more than 10¹⁰ write cycles in an AlScN/AlN superlattice while preserving a complete-switching criterion.
The study, published on September 10, 2026, targets one of the field’s nastiest reliability problems: defects called nitrogen vacancies can cluster and migrate through the material, increasing leakage current and eventually contributing to dielectric breakdown. Xidian’s approach combines defect confinement with dynamic recovery. That is a meaningful device-level advance—but it is still a laboratory result, not a memory module ready for a computer.
What wurtzite ferroelectric memory actually is
First, wurtzite is a crystal structure, not a chemical element or a single material. It is a polar hexagonal arrangement of atoms. AlScN is one material that can adopt this structure and exhibit ferroelectric behavior: its polarization can be switched electrically and retained after power is removed.
That gives engineers a physical way to represent information without continuously supplying power. Depending on the device design and operating method, the polarization can support binary states, partial states or multiple levels. Proposed architectures include ferroelectric capacitors, FeFETs, ferroelectric tunnel junctions, diodes and in-memory-computing structures.
The attraction is easy to understand. A large, persistent polarization signal can make a memory state easier to distinguish electrically. The catch is that the material must survive repeated high-field switching without developing leakage paths or breaking down. Memory is not impressive because it works once; it is impressive because it keeps working after an absurd number of writes.
What Xidian changed
The Xidian device uses an AlScN/AlN superlattice: alternating layers that create a designed material structure rather than a single uniform AlScN film. In the reported architecture, the superlattice confines the evolution of nitrogen vacancies, while a dynamic recovery protocol helps stabilize the defects during cycling.
Nitrogen vacancies are missing nitrogen atoms in the crystal lattice. The study links their clustering and long-range, percolative migration with growing leakage current, dielectric breakdown and ferroelectric degradation. In plain English, defects that begin as local problems can organize into pathways that make the device increasingly leaky. The superlattice is intended to keep that evolution contained instead of allowing it to spread through the active region.
Under the study’s complete-switching criterion, the device exceeded 10¹⁰ cycles while retaining a remnant polarization of at least 100 μC/cm². That criterion matters: it requires the device to preserve a substantial polarization signal after the switching operation, rather than merely registering a small electrical response.
Why the earlier 10¹⁰-cycle result is different
A January 9, 2026 study from the University of Pennsylvania reported another AlScN device exceeding 10¹⁰ cycles. It used a 45-nm-thick Al0.64Sc0.36N capacitor and controlled partial polarization switching. Under complete reversal, the same study reported approximately 10⁸ cycles, with a 2Pr value of about 200 μC/cm².
Those results describe different architectures and switching conditions. The headline number is therefore not an apples-to-apples ranking. One study reports more than 10¹⁰ cycles for complete switching in an AlScN/AlN superlattice; the other reports more than 10¹⁰ cycles for controlled partial switching in Al0.64Sc0.36N, while its complete-reversal endurance is about 10⁸ cycles.
| Study | Architecture or material | Switching criterion | Reported endurance | Quantitative condition |
| Xidian University, published September 10, 2026 | AlScN/AlN superlattice | Complete switching | >10¹⁰ cycles | Remnant polarization ≥100 μC/cm² |
| University of Pennsylvania, published January 9, 2026 | 45-nm Al0.64Sc0.36N capacitor | Controlled partial switching | >10¹⁰ cycles | 2Pr >30 μC/cm² |
| University of Pennsylvania, published January 9, 2026 | 45-nm Al0.64Sc0.36N capacitor | Complete reversal | ≈10⁸ cycles | 2Pr ≈200 μC/cm² |
The distinction is not a footnote. Partial switching can reduce the stress applied to a ferroelectric device, while complete reversal demands a larger polarization change. Endurance figures only become meaningful when the switching operation and the retained signal travel with the number.
Why AlScN is interesting beyond endurance
AlScN combines ferroelectric polarization with thin-film processing possibilities. The January study reports deposition at 350 °C and describes sub-400 °C deposition as compatible with back-end-of-line-oriented semiconductor processing. Earlier work cited in that study has also reported switching in films as thin as 5 nm, although thinness alone does not settle questions about yield, density or system integration.
The reported polarization is another attraction. Wurtzite ferroelectrics can exceed 100 μC/cm² in remnant polarization, while the cited comparison describes HfO₂-based ferroelectrics as typically offering about 10–40 μC/cm². A larger signal could be useful in dense memory or computing-in-memory designs, but it does not automatically translate into better overall performance. Coercive fields, switching energy, defect control, scaling and manufacturing repeatability still matter.
An optimized 10-μm-diameter device in the January study approached a breakdown field of 10 MV/cm. That measurement is a property of the reported research devices and conditions, not a guarantee for every AlScN geometry or future array.
What the result means for real computers
The immediate significance is reliability engineering. Xidian’s architecture addresses a specific failure pathway by controlling nitrogen-vacancy evolution while the device is electrically cycled. That makes long endurance more credible as a materials-and-device objective, rather than treating defects as an afterthought.
The next practical questions are larger than a single capacitor: how the material behaves in arrays, how much energy switching requires, how densely cells can be packed, how consistent fabrication is across a wafer, and how the devices perform inside a complete memory or compute system. The research describes possible uses in nonvolatile memory and in-memory computing, but those system properties are separate engineering problems.
For now, the strongest claim is also the most precise one: wurtzite ferroelectric memory has gained a new laboratory endurance result in which an AlScN/AlN superlattice exceeded 10¹⁰ complete-switching cycles while retaining the study’s stated polarization threshold.
