By controlling the Jekyll and Hyde properties of semiconductor replacements, which move from electrically strong insulators to current-conducting metals, Nebraska's Xia Hong and colleagues are developing smaller, more efficient semiconductors. It may have opened up a new path to digital devices.
Semiconductors' ability to conduct electricity in the Goldilocks zone (weaker than metals, better than insulators) makes it ideal for engineers looking to build transistors, tiny on-off switches that encode binary ones and zeros. It has been positioned as an option. . Applying a voltage to the control knob, known as the gate insulator, causes current to flow through the semiconductor channel (1). If you remove it, the flow will stop (0).
Millions of these nanoscale semiconductor-based transistors coat modern microchips, turning on and off to collectively process or store data. But while transistors are already very small, consumer and competitive demands are forcing electrical engineers to make transistors even smaller, either to pack in more functionality or to shrink the devices that house them. I keep getting asked to do it. Unfortunately, those engineers now face practical, even fundamental, limits to how small semiconductors can be made.
Researchers are starting to look at semiconductors as a whole, not just the industry's favorite silicon. More than 20 years ago, some people started using a type of material called Mott insulators. If semiconductors are the happy medium that has driven decades of clockwork progress, Mott insulators are rather two-faced, whose ambivalence is the source of both their fascination and frustration. Similar to a wildcard.
According to long-standing conductivity theory, materials with the electronic properties of Mott insulators should generally be classified as metals. However, unlike electrons in metals and semiconductors, electrons in Mott insulators do not behave like independent particles. Instead, they are confined to local sites and interact in ways that prevent them from moving freely within the material. Still, certain conditions (high temperature, introduction of more electrons) can overwhelm these forces and eventually free up the electrons, essentially turning the Mott insulator into a conducting metal.
“So (traditionally) you have either itinerant electrons or localized electrons,” says Hong, a professor of physics at the University of Nebraska-Lincoln. “It’s very clearly defined.
“However, in the case of Mott insulators, electronic interactions cannot be ignored. Such correlations make it difficult to simply define them as either metals or insulators. It can be an insulator or it can be an insulator.”
By placing a gate insulator made of a so-called ferroelectric material on top of the Mott insulator and using voltage to reverse the polarization, or arrangement of positive and negative charges, of the latter, the researchers were able to create the Mott transition. I realized that I could induce it from an insulator to a metal and vice versa. In this way, the pairing behavior and most promising features were modeled after that of semiconductors.
But the metallic phase of Mott insulators offers a decisive advantage over long-reigning insulators: they carry charge numbers and densities that far exceed what semiconductors have been capable of to date. . The higher the density, the less space the charged electrons need to block the electric field, which can switch the polarization of the ferroelectric and prevent the transistor from remaining in the “off” state. And the shorter the shielding length required by charged particles, the smaller the transistor can be made, potentially smaller than previous semiconductor transistors.
problem? The same downsizing density further increases the difficulty of transitioning the Mott channel from an insulator to a metal and vice versa through the overlying ferroelectric material. Engineers often measure the technical feasibility of a transistor in terms of its on-off ratio, the amount of current that flows when a voltage is applied and the amount of current that flows when a voltage is pulled (ideally close to zero). Measure. The higher the ratio, the greater the margin for error when processing and storing data. Minimizing the current in the “off” state also saves energy, while maximizing the current in the “on” state increases processing speed.
In 2018, a year after Hong, doctoral advisor Yifei Hao, and postdoc Xuegang Chen first tackled the problem, another research team determined the on-off ratio in a Mott ferroelectric pair at room temperature. reported the highest value ever of 11. Husker's team did some experiments of their own, and they ended up pushing him up to 17. This has improved, but is still too low.
Eventually, Hong and his colleagues decided to try adding another layer beneath the Mott channel. For the third, lower layer, the team chose a material that cannot carry as much charge density as the Mott material above it, but importantly allows the charge to move downward from the Mott. Sparse areas.
In effect, the team kept Jekyll and tamed Hyde. The space-saving benefits of the high-density region remained, but because the overall density was reduced (thanks to the additional bottom layer), the team also maintained greater control over the insulator-to-metal transitions. That advantage came in the form of the highest on/off ratio on record, at a whopping 385, more than 20 times higher than previously reported. This number may exceed the upper limit of what can be achieved with the Mott ferroelectric approach, Hong said.
Is it also advantageous? Ferroelectrics are nonvolatile. In other words, it can hold 1's and 0's without constant power supply. And the fact that only a small voltage is required to reverse polarization makes Mott ferroelectric pairs more energy efficient than similar nonvolatile but magnetic-based memories such as MRAM.
“For me, in terms of technology development, this is a big deal because it shows that it's possible,” Hong said. “By retaining many of the manufacturing processes of traditional semiconductors and overcoming their fundamental limitations, we can achieve extremely high-performance devices.”
Hong believes Mott-based transistors could be used in these devices sooner rather than later.
“I think I’m ready,” she said of the concept. “It's very competitive with other non-volatile memory technologies. I think anyone with the right mindset can understand the concept and implement it.”
The research team reported their results in the journal Nature Communications. Hong, Hao, and Chen are Le Zhang of Nebraska, Yimei Zhu, Myung-Geun Han, and Wei Wang of Brookhaven National Laboratory, Yue-Wen Fang of the University of the Basque Country, and New York University in Shanghai. co-authored the study with Hanghui Chen of The researchers were primarily supported by the National Science Foundation, which awarded Nebraska a $20 million grant in 2021, as well as a semiconductor research company.
