N-bit ferroelectric tunnel junctions via all‑van‑der‑Waals stacking

In the relentless drive to pack more information into smaller spaces, scientists have long sought to move beyond the simple on-off switch that powers most of our digital world. For decades, computer memory has relied on binary logic, where a single unit of data is either a zero or a one. This works well, but it hits a physical wall when we try to make devices smaller and faster. To break through this limit, researchers have turned to a class of materials called ferroelectrics. These are special crystals that possess a permanent electrical polarization, meaning their internal charges are permanently shifted in one direction or another. By sandwiching a thin layer of such a material between two metal contacts, scientists can create a tunnel junction where the electrical resistance changes depending on which way the polarization points. This allows the device to act as a memory cell, storing data without needing constant power. However, traditional designs are limited to just two states, and stacking multiple layers to store more data has proven difficult because the layers tend to interfere with each other, forcing them to switch all at once rather than independently.

A team of researchers at Nanjing University has now proposed a way to overcome these limitations by stacking multiple layers of a different kind of material known as van der Waals ferroelectrics. Unlike the rigid crystals used in older devices, these materials are held together by very weak forces, similar to how a stack of sticky notes might sit on a table without being glued together. This weak connection is the key to the new design. In a study published recently, the researchers demonstrated a theoretical model for a memory device that can store far more than just a single bit of information. By stacking five distinct layers of these weakly bonded materials, they showed that each layer could be switched on or off independently, creating a single device capable of holding thirty-two different resistance states. This represents a massive leap in information density, where adding just one more layer doubles the amount of data the device can hold.

The researchers built their model using a specific combination of five different ferroelectric materials, each chosen to have a unique strength required to flip its electrical direction. They placed this stack between two identical metal electrodes made of titanium diselenide. In conventional memory devices, achieving a strong signal difference between the "on" and "off" states usually requires using two different metals for the top and bottom contacts to create an imbalance. Here, the team found that the imbalance comes naturally from the stack itself. Because the five layers are made of different chemicals and respond to electricity in different ways, the electrical voltage drops unevenly across the device. This creates an internal asymmetry that is strong enough to generate a giant difference in resistance between the various states, even though the outer metal contacts are exactly the same.

To verify that this idea works, the team used powerful computer simulations to model the behavior of every possible combination of the five layers. They found that all thirty-two possible arrangements of the layers remained stable, with no unwanted interactions causing the layers to flip together. Each of these thirty-two states produced a distinct electrical current, meaning a device could read out exactly which state it was in. The simulations showed that the difference between the most conductive state and the least conductive state was enormous, exceeding a factor of one hundred thousand. This high contrast is essential for reliable memory, as it ensures that the device can clearly distinguish between a zero and a one, or in this case, between any of the thirty-two unique levels.

A critical part of the discovery involves how the device is controlled. Because each layer requires a different amount of electrical force to switch, the researchers showed that the layers could be addressed in a specific order. By gradually increasing the voltage, the layer that needs the least force would flip first, followed by the next, and so on. This creates a predictable path through all the possible states, much like climbing a ladder where each rung requires a specific effort to reach. The team mapped these transitions using a geometric concept where every possible state is a point in a multi-dimensional space, and the allowed moves between states form a clear, connected network. This structure ensures that any state can be reached from any other state through a series of controlled steps, making the device programmable and reliable.

The study also addressed a major hurdle in building such devices: the difficulty of fitting different materials together. In traditional electronics, layers must be grown on top of each other with perfect alignment, or the device fails. However, because these van der Waals materials are held together by weak forces, they can be stacked even if their internal atomic patterns do not match perfectly. The researchers noted that while their specific computer model used materials with closely matched sizes to make the calculations feasible, the physical principle would work with a much wider range of materials. In a real-world experiment, these layers could be mechanically peeled and stacked like sheets of paper, allowing engineers to mix and match different ferroelectrics without worrying about the strict lattice matching required by older technologies.

This work suggests a new path forward for memory technology, one that scales exponentially rather than linearly. Instead of trying to make a single layer do more work, the design simply adds more layers, each acting as an independent bit of information. The researchers emphasized that while their results are currently based on computer simulations, the underlying physics is robust and the materials involved are either known or derived from known compounds. The next step would be to physically assemble such a stack and test whether the layers switch independently and whether the thirty-two distinct states can be read out in a real device. If successful, this approach could lead to memory chips that are not only incredibly dense but also capable of performing calculations directly within the memory itself, a concept known as in-memory computing.

The implications extend beyond just storing more data. By proving that a symmetric device with identical top and bottom contacts can still produce a massive signal difference, the study challenges the long-held belief that complex, asymmetric electrode designs are necessary for high-performance memory. This simplifies the manufacturing process and opens the door to using a wider variety of materials. The team's findings provide a universal framework for designing these multi-bit devices, showing that as long as the layers have different switching strengths, the same rules for controlling them will apply. This universality means that the design is not tied to a single set of materials but can be adapted as new ferroelectric compounds are discovered.

Ultimately, this research offers a blueprint for the next generation of ultracompact memory. By leveraging the weak bonds between atomic layers, scientists have found a way to preserve the independence of each layer while stacking them into a single, functional unit. The result is a device that can hold thirty-two times more information in the same footprint as a standard binary cell, with the potential to scale even further as more layers are added. The study stands as a significant theoretical demonstration that the exponential scaling of information density is physically possible, setting the stage for future experiments to bring this multi-bit memory from the realm of simulation into the physical world.