Ferroelectric thin films are intriguing candidates for nanoscale electronics, including memory systems in which information is stored as polarization states equivalent to the 1s and 0s of binary systems. Researchers at the University of New South Wales (UNSW) and Monash University in Australia have recently overcome a key limitation of this technology—the decay of polarization states over short time scales. As they described in Nature Communications, they utilized designer defects in a thin film of bismuth ferrite (BiFeO₃, BFO) to set a new record for polarization retention.

In ferroelectric random-access memory (FeRAM) systems, information is stored in the polarization states of nanometer-scale ferroelectric domains. Different states are separated by thin boundary regions known as domain walls. Typical FeRAM systems retain distinct polarization states for just days or weeks before they begin to decay and information is compromised.

Research has shown that BFO films grown on lanthanum aluminate substrates (LaAlO₃, LAO) experience strain due to the lattice mismatch between BFO and LAO. At thicknesses of 30 nm or more, this leads to a mixed-phase state, which includes BFO in a tetragonal-like phase and a rhombohedral-like phase.

In this new study, a research team led by UNSW’s Jan Seidel fabricated a thin film of BFO in the tetragonal-like phase on a LAO substrate, using pulsed laser deposition. The researchers introduced defects in the BFO during the growth process by precisely controlling the temperature, gas pressure, pulse energy, pulse rate, and target composition. High-resolution images of the 60-nm-thick BFO film revealed a high-quality surface with uniformly distributed defects. The defects were larger than a single atom, with an average width of ∼5 nm and height of ∼2 nm. “Their structure is complex and has not been investigated in detail yet,” says Seidel.

To create a memory storage system, the team inserted a 3-nm-thick electrode between the BFO and substrate. Then, using the conducting tip of an atomic force microscope probe, they applied a voltage pulse across the film. The pulse created a stable nucleus of polarization in the film directly below the tip from which the domain spread outward. This spreading can be described by the motion of the domain walls. By systematically varying pulse duration and tip voltage while scanning the film, the team created domains of different sizes.

A statistical analysis of domain diameter as a function of voltage and duration revealed that activating the motion of a domain wall in the film required an electric field 3–6 times larger than in conventional BFO systems. This suggested that the defects exerted local strain that effectively pinned the domain walls in place.

To study the extent of the pinning, the researchers formed domains of various sizes using a tip voltage of –9 V and a pulse duration ranging from 5 ms to 200 ms. They imaged the film with high-resolution piezoresponse force microscopy, which simultaneously captured topography and ferroelectric domains, during a span of 8904 hours. Even after more than one year, the diameters of the domains remained essentially the same.

The normalized polarization retention of this system was at least one to two orders of magnitude better than other ferroelectric systems, the researchers reported. Furthermore, the stability persisted across domains of all sizes. This was surprising because small domains normally decay faster than large domains, and indicates the system’s potential for high-density memory applications that utilize small domains. With an optimally sharp tip, the researchers estimate they may be able to achieve a storage density up to 1300 Gbit/in².

“[This] work shows a promising path forward to producing superior high-density nonvolatile memories based on ferroelectric materials,” according to Matthew Dawber, an expert on ferroelectric materials at Stony Brook University, The State University of NewYork, who was not associated with this project. “[The researchers] show that introduced defects can help stabilize tiny domains for very long times. This is a win-win, normally it’s hard work to get rid of defects, and conversely, it’s not too hard to introduce them,” Dawber says.

The team focused on one kind of defect in this research, but Seidel says that there are many options for pinning domain walls and further improvement may be possible. “Another interesting aspect is the intrinsic properties of domain walls themselves, which can be exploited for nanoelectronics,” he says. “[Domain walls have] been known for a long time, but insight into their intrinsic properties and functionality has been investigated in more detail only recently.”

Originally published in the April 2020 issue of MRS Bulletin.