Ferroelectricity discovered in ultrathin two-dimensional gallium oxide films

In a result that could reshape the future of miniature electronics, researchers have demonstrated ferroelectricity in a two-dimensional semiconductor just six ångströms thick—roughly the width of a single chemical bond stretched a few times over. The material, single-crystalline gallium oxide (Ga₂O₃), is a wide-bandgap semiconductor long prized for its ability to withstand high electric fields and temperatures. Now, a team led by Tong Jiang, Hanyan Chen and Yu Yuan and colleagues has shown that when this oxide is thinned to a half-unit-cell thickness, it spontaneously reorganizes into a ferroelectric layered structure capable of holding electric polarization with remarkable stability and switching it at voltages low enough to satisfy the most demanding requirements of modern chip manufacturing.

Ferroelectric materials are the working horses of a growing class of memory and logic devices. Their defining feature—a switchable, persistent electric polarization that survives without power—makes them ideal for non-volatile memories, energy-efficient logic and novel neuromorphic architectures. But as researchers have pushed these materials ever thinner to pack more devices onto a chip, a fundamental problem has emerged. Below a certain thickness, typically a few nanometres, most ferroelectrics lose their polarization entirely. The culprit is the depolarization field, an electrostatic penalty that grows punishingly strong as the ferroelectric layer shrinks, destabilizing the polar state and erasing the very property that makes the material useful. At the same time, the voltages needed to flip the polarization in conventional ferroelectric films often exceed what silicon-based complementary metal–oxide–semiconductor (CMOS) technology can tolerate, stalling efforts to integrate ferroelectrics directly onto advanced chips.

The new study, published in Nature Electronics, tackles both problems at once with an elegantly simple strategy: let strain do the work. The team began with epitaxial β-Ga₂O₃, a crystalline form of gallium oxide grown with precise atomic alignment on a supporting substrate. Using what the authors describe as a self-limiting exfoliation mechanism, they peeled the crystal apart layer by layer until only a half-unit-cell-thick sheet remained—just six ångströms, or 0.6 nanometres, thick. The exfoliation process is self-limiting in the sense that it naturally stops at this exact thickness rather than continuing to fragment the material, yielding large, uniform, single-crystalline films of consistent quality. This controllability is crucial; devices built from atomically thin materials are only as good as the uniformity of the layers from which they are made.

Something remarkable happened at this extreme thinness. The biaxial strain imposed during exfoliation—a compression or tension applied equally along two in-plane axes—drove the gallium oxide through a structural phase transition. Instead of collapsing into a non-polar configuration as most materials would, the ultrathin Ga₂O₃ adopted a layered, polar structure in which the arrangement of gallium and oxygen atoms creates a built-in electric dipole. In other words, the material became ferroelectric precisely because it was squeezed to almost nothing, turning the usual curse of extreme thinness into the very source of its functionality. This strain-induced route to ferroelectricity is distinct from the approaches used in other two-dimensional ferroelectrics and offers a clear design principle: engineer the strain, and the polar phase follows.

The performance figures are striking. The researchers measured polarization switching voltages as low as 0.8 volts. To appreciate why this number matters, consider the trajectory of the semiconductor industry. As transistors have shrunk, the operating voltages of CMOS circuits have fallen steadily, and today’s most advanced chips run at supply voltages of well under one volt. A ferroelectric material that requires two, three or five volts to switch simply cannot be driven by such circuits without additional voltage-boosting circuitry that consumes area, cost and energy. A switching voltage of 0.8 volts meets the voltage-scaling requirements of CMOS technology outright, meaning ferroelectric Ga₂O₃ devices could in principle be driven directly by standard chip transistors without any intermediate circuitry—a milestone that has eluded the ferroelectronics community for years.

Equally important is the material’s resilience. Theoretical calculations performed by the team reveal the microscopic origin of the switching process: it proceeds through the reconstruction of covalent bonds between gallium and oxygen atoms. As the applied electric field pushes the structure from one polar state to the other, existing bonds break and reform in a cooperative, atomically coordinated fashion. This bond-reconstruction mechanism is not merely a curiosity of the switching pathway; it actively counters the depolarization field that normally destroys ferroelectricity at the nanoscale. Because the polar state is stabilized by strong covalent bonding rather than by weaker ionic displacements alone, the ultrathin films retain their polarization over long times and elevated temperatures. The authors report both high retention—meaning the polarization persists for extended periods after being written—and robust thermal stability, two properties that together determine whether a ferroelectric can survive the rigors of real-world device operation, where memories must hold data for years and chips must endure the heat generated by their own operation.

The practical significance of the work extends beyond the material’s intrinsic properties to how it can be incorporated into existing technology. The team demonstrated that ferroelectric two-dimensional Ga₂O₃ can be integrated onto silicon using a low-temperature process compatible with the back end of line—the later stages of chip fabrication in which interconnects and passivation layers are built atop the already-completed transistors. Temperature budgets in back-end-of-line processing are notoriously tight, typically capped at around 400 degrees Celsius, because the aluminium and copper interconnects and the carefully tuned transistor structures beneath can be degraded by excessive heat. Many promising functional materials fail at precisely this hurdle: they require high-temperature synthesis or annealing steps that cannot be performed once the underlying silicon circuitry exists. A low-temperature, back-end-of-line-compatible integration route means ferroelectric Ga₂O₃ devices could be added as a final manufacturing step on top of finished CMOS chips, opening a path to hybrid circuits that combine the computational power of silicon with the non-volatile, low-energy switching of atomically thin ferroelectrics.

The implications for memory technology are particularly compelling. Conventional non-volatile memories each carry compromises: flash memory is slow to write and wears out with use; dynamic random-access memory is fast but loses its contents when power is removed; and emerging resistive and phase-change memories face their own scaling and endurance challenges. Ferroelectric memories offer a different bargain—fast, low-energy writes, non-volatile storage and theoretically near-unlimited read endurance—provided the ferroelectric layer can be made thin enough to allow dense stacking and low-voltage operation. A six-ångström ferroelectric is about as thin as such a layer can conceivably be. Combined with the demonstrated CMOS-compatible switching voltage, the material could enable memory cells packed at densities approaching the physical limits of atomic-scale engineering, in which a single polar sheet stores each bit of information.

Gallium oxide itself brings additional advantages to the table. As a wide-bandgap semiconductor—its bandgap of roughly 4.8 electronvolts is far larger than silicon’s—it inherently resists leakage currents, the parasitic flows that bleed charge and erode stored information in thin dielectric layers. This property complements the covalent-bond stabilization of the polar phase, giving the ultrathin films a two-fold defence against the loss of information: electrical leakage is suppressed by the wide gap, and depolarization is countered by the bond-reconstruction switching mechanism. The combination of a wide-bandgap semiconductor with intrinsic ferroelectricity in a single two-dimensional material is rare, and it suggests possibilities beyond memory, including ferroelectric field-effect transistors, gate dielectrics with tunable properties, and sensors that exploit the coupling between polarization and the material’s chemical or thermal environment.

Challenges undoubtedly remain before six-ångström ferroelectric memories reach production lines. Scaling exfoliation-based fabrication from laboratory flakes to wafer-scale films is a perennial hurdle for two-dimensional materials, and the long-term reliability of atomically thin devices under billions of switching cycles must be rigorously established. Endurance, fatigue and retention specifications for commercial memories are exacting, and the behaviour of the strain-induced polar phase over such lifetimes will need careful characterization. Nevertheless, the demonstration that ferroelectricity can survive—and even be created by—extreme dimensional confinement, while simultaneously meeting the voltage and thermal constraints of silicon technology, removes two of the most stubborn obstacles on the road to atomic-scale ferroelectric electronics.

The study also delivers a conceptual lesson to the broader materials community. For decades, thinness has been treated as the enemy of ferroelectricity, with a well-defined critical thickness below which polar order succumbs to the depolarization field. By showing that biaxial strain can invert this narrative—transforming a conventional wide-bandgap oxide into a stable, low-voltage ferroelectric at a thickness where ferroelectricity was thought impossible—the work expands the design space in which engineers can hunt for new functional materials. Covalent bond reconstruction, the mechanism underpinning the switching and stability, offers a template that may be found in other layered oxides and chalcogenides. If so, the six-ångström gallium oxide reported here may be remembered not only as the thinnest ferroelectric of its class, but as the first member of a family of strain-engineered polar semiconductors that finally allow memory and logic to shrink to the scale of atoms.