Researchers Successfully Grow Atom-Thick Sheets of Hexagonal Boron Nitride

Researchers are slowly identifying innovative ways to extend Moore’s Law. One new way shows a path toward built-in circuits that have two-dimensional (2D) transistors.

A researcher from Rice University and his colleagues in China and Taiwan have reported that atom-thick sheets of hexagonal boron nitride (hBN) have been grown as two-inch diameter crystals over a wafer. The study results were recently published in the Nature journal.

hBN is a wide bandgap semiconductor. The researchers had surprisingly achieved the long-sought aim of producing perfectly ordered hBN crystals by leveraging the disorder that exists among the random steps on a copper substrate. These meandering steps are responsible for keeping the hBN in line.

Wafer-scale hBN, integrated into chips as a dielectric between nanoscale transistor layers, are known to excel in damping electron scattering and trapping that restrict the efficiency of a built-in circuit. However, to date, none has been successful in making perfectly ordered hBN crystals that are sufficiently large—in this example, on a wafer—to be useful.

Boris Yakobson is a materials theorist at Brown School of Engineering and is also a co-lead researcher on the study performed with Lain-Jong (Lance) Li of the Taiwan Semiconductor Manufacturing Co. (TSMC) and his group.

Theoretical analysis, as well as first-principles calculations, was performed by Yakobson and Chih-Piao Chuu from TSMC and they subsequently revealed the mechanisms of what their co-authors observed in experiments.

Experimentalists at Taiwan’s National Chiao Tung University and TSMC developed a two-inch, 2D hBN film as a proof of concept for manufacturing. They then transferred this hBN film to silicon and added a layer of field-effect transistors that were patterned onto 2D molybdenum disulfide over the hBN film.

The main discovery in this work is that a monocrystal across a wafer can be achieved, and then they can move it. Then they can make devices.

Boris Yakobson, Materials Theorist, George R. Brown School of Engineering, Rice University

“There is no existing method that can produce hBN monolayer dielectrics with extremely high reproducibility on a wafer, which is necessary for the electronics industry,” added Li. “This paper reveals the scientific reasons why we can achieve this.”

Yakobson believes that the technique could also be applied extensively to other types of 2D materials, using few trial and error methods.

I think the underlying physics is pretty general. Boron nitride is a big-deal material for dielectrics, but many desirable 2D materials, like the 50 or so transition metal dichalcogenides, have the same issues with growth and transfer, and may benefit from what we discovered.

Boris Yakobson, Materials Theorist, George R. Brown School of Engineering, Rice University

Back in 1975, Intel’s Gordon Moore anticipated that the number of transistors present in a built-in circuit would increase by two-fold every couple of years. However, as the design of integrated circuits becomes smaller, with circuit lines measuring only a few nanometers, it has been rather difficult to maintain the rate of development.

The potential to stack 2D layers, each with an unlimited number of transistors, may resolve these limitations if these layers are separated from one another. Due to its wide bandgap, insulating hBN is the main candidate intended for the purpose.

In spite of having “hexagonal” in its name, hBN monolayers as shown in the above image appear as a superposition of a pair of distinct triangular lattices of nitrogen and boron atoms.

If the material has to perform up to the specification, then hBN crystals have to be perfect; in other words, the triangles have to be linked and all should point in the same direction. Grain boundaries, which exist in non-perfect crystals, degrade the electronic properties of the material.

For a perfect hBN, its atoms have to align accurately with the atoms on the substrate below. The scientists discovered that while copper in a (111) arrangement—the number refers to the orientation of the crystal surface—does the job, it does so only after it is annealed in the presence of hydrogen and at an extreme temperature on a sapphire substrate.

Annealing removes grain boundaries present in the copper and leaves behind a single crystal. But a perfect surface like this would be “way too smooth” to implement the hBN orientation, added Yakobson.

The previous year, Yakobson reported a study on how to grow pristine borophene on silver (111). He also reported a theoretical prediction that copper is capable of aligning hBN by virtue of the complementary steps that exist on its surface. The surface of the copper was vicinal—which means somewhat miscut to reveal the atomic steps between the expansive terraces. That study triggered an interest in Taiwan-based industrial researchers, who approached the professor after a talk there in the previous year.

“They said, ‘We read your paper,’” recalled Yakobson. “‘We see something strange in our experiments. Can we talk?’ That’s how it started.”

Informed by his previous experience, Yakobson proposed that copper (111) are able to retain step-like terraces over its surface due to thermal fluctuations, even when its own grain boundaries are removed.

The atoms in these random “steps” provide the perfect interfacial energies to bind and restrain hBN. This hBN subsequently grows in one direction, while it binds to the copper plane through the rather weak van der Waals force.

Every surface has steps, but in the prior work, the steps were on a hard-engineered vicinal surface, which means they all go down, or all up. But on copper (111), the steps are up and down, by just an atom or two randomly, offered by the fundamental thermodynamics.

Boris Yakobson, Materials Theorist, George R. Brown School of Engineering, Rice University

Due to the orientation of copper, the horizontal atomic planes are offset by a fraction of the lattice beneath. “The surface step-edges look the same, but they’re not exact mirror-twins,” explained Yakobson. “There’s a larger overlap with the layer below on one side than on the opposite.”

That phenomenon makes the binding energies on every side of the copper plateau different by a small 0.23 eV (for every quarter-nanometer of contact). This energy is sufficient to push the docking hBN nuclei to grow in the same direction, added Yakobson.

The experimental research team observed that the optimal thickness of copper was 500 nm, which is sufficient to prevent its evaporation at the time of the hBN growth through chemical vapor deposition of ammonia borane on a sapphire/copper (111) substrate.

Tse-An Chen of TSMC is the study’s co-lead author. Other co-authors are Chien-Chih Tseng, Chao-Kai Wen, Wei-Chen Chueh, and Wen-Hao Chang of Chiao Tung; H.-S. Philip Wong and Tsu-Ang Chao from TSMC; Shuangyuan Pan and Yanfeng Zhang from Peking University, China; Qiang Fu from the Chinese Academy of Sciences, Dalian, China; and Rongtan Li from the Chinese Academy of Sciences and the University of Chinese Academy of Sciences, Beijing.

Yakobson is also the Karl F. Hasselmann Professor of Materials Science and NanoEngineering and a professor of chemistry at Rice University. Chang is a professor at Chiao Tung and director of Rice University’s Center for Emergent Functional Matter Science. Li is Director, Corporate Research, Taiwan Semiconductor Manufacturing Co.

The study was supported by the Ministry of Science and Technology of Taiwan, the Ministry of Education of Taiwan, TSMC, the National Natural Science Foundation of China, the U.S. Department of Energy, and the Chinese Academy of Sciences.