Ferroelectric materials belong to a class of crystals whose low symmetry gives rise to a spontaneous polarization along a crystal axis. Those polarized, thermodynamically stable states can be switched from one to the other by applying what’s known as the coercive electric field Ec. That switchability forms the basis for the nonvolatile RAMs that are common in computing. Between 2019 and 2021, researchers found that solid films of alloyed aluminum nitride turn out to be ferroelectric. The reports were a surprise to most scientists: The films are well-known pyroelectric and piezoelectric crystals, but few believed they could be ferroelectric because their coercive field is just too perilously close to the material’s dielectric breakdown field. Apply a high enough electric field to switch the polarization, and you risk destroying the material.
Computer engineers have been making ferroelectric logic circuits since the 1960s, but integrating them with mainstream silicon-based semiconductors has proven difficult. What’s more, conventional ferroelectrics, such as perovskites, are difficult to scale down to atomic dimensions. The new ferroelectrics, which crystallize in the so-called wurtzite structure, are free of both problems. Boron-doped AlN, in particular, consists exclusively of elements common in CMOS electronics.
Pennsylvania State University researchers led by Jon-Paul Maria and Susan Trolier-McKinstry—ceramic scientists who had demonstrated ferroelectricity in boron-doped AlN—have now teamed up with a Carnegie Mellon University group led by Elizabeth Dickey to analyze the film’s structure using transmission electron microscopy (TEM). The researchers realized that if they could infer the mechanism by which the polarization switches at the atomic scale, they might be able to manipulate it—perhaps by straining the film, growing it thinner, or altering its doping concentration. Such tricks may allow them to dramatically lower the coercive field from roughly a few thousand volts per centimeter to levels on the order of 1 V/cm. That transformation would render the films practical and durable enough for energy-harvesting and electro-optical devices, among other applications.
As deposited on a tungsten electrode, the experimental Al1−xBxN film forms with a polarization whose orientation points downward (into the substrate), a configuration shown in the left-hand panel of the figure and commonly referred to as N-polar growth. The gray atoms in the inset represent aluminum and boron, and the blue ones represent nitrogen. (The image is an on-edge view of the film—a mere 6 nm thick—that shows the projections of nitrogen and aluminum or boron atoms through the lattice.) Focused onto a small segment of the film, the electron beam interferes with the electrostatic potential of the crystal and images the separate sublattices of the larger aluminum and smaller nitrogen atoms at the same time.
The TEM beam also triggers the polarization switch: As it scatters electrons throughout the film, they ionize local Al, B, and N atoms. Auger and secondary electrons are ejected from the film, which positively charges the film’s surface. Once the local electric field exceeds Ec, the film’s polarization switches upward to an Al-polar orientation. In that orientation, the Al–N dipolar bonds now all point with upward polarization (out of the film’s plane), as shown in the right-hand panel.
The research group had assumed the doped wurtzite ferroelectric flips from down to up polarization by passing through an intermediate, nonpolar stage in which the metal atoms and N atoms adopt a flattened configuration. But when they ran a first-principles calculation on the film, they found that a more complicated atomic structure emerges. That intermediate structure is indeed nonpolar but turns out to include substantial local bonding and structural distortions. Both N-polar and Al-polar unit cells appear in the atomic configuration, combining in a doubled unit cell both up and down locally polarized states that cancel each other. What’s more, the energy barrier required to attain that intermediate state is small—just 0.2 eV. Reassuringly, the collaboration's experimental image of the boron-doped film as it goes through the polarization flip bears out the calculation’s prediction.
Why should the atomic motions pass through a much lower energy barrier found by the simulation when the coercive field required to trigger the polarization switch, roughly 5 MV/cm, is so high? The researchers don't yet know, but they expect it may have to do with the material's extreme stability. The covalent bonds in AlN are among the strongest in nature, and the material doesn’t melt until it reaches 2200 °C. (S. Calderon V et al., Science 380, 1034, 2023.)