Laying The Groundwork For Atomic Layer Etching

Even if a housepainter takes extraordinary care when working on window frames, a bit of paint will still end up on the edges of the glass. But it’s no problem because the painter can use a sharp razor blade to scrape the unwanted paint without damaging the paint job or the frame.

Now imagine trying to do that scraping on the nanoscale. That’s the task the semiconductor industry faces today.

As transistors and other integrated circuit components shrink to make ever smaller and more powerful electronics, chip makers are searching for exceptionally fine methods for removing minuscule amounts of material. As Intel staff engineer Satyarth Suri explains, these etching methods are needed to precisely form critical features of circuit elements, which in recent years have shrunk to the low-nanometer size range and are becoming increasingly complex and three-dimensional. The margin of error is small: Just a tiny deviation in shape, or a speck of unwanted material left behind, can alter or ruin device performance.

For that reason, researchers are busy developing a technique known as atomic layer etching (ALE). This chemistry-driven layer-by-layer film-removal procedure can be thought of as the reverse of atomic layer deposition (ALD), which is a commercially advanced layer-by-layer technique to build up films. ALD is the subject of a story on page 54 of this issue of C&EN.

“The first research efforts [on ALE] date back 25 years, yet ALE is still very much in its infancy,” says W. M. M. (Erwin) Kessels, a thin-film specialist at Eindhoven University of Technology, in the Netherlands.

Now the ultrafine etching technique is poised to grow up quickly. “In just the past six months, there has been an explosion of interest in ALE,” says University of Colorado chemistry professor Steven M. George, a specialist in vapor deposition processes.

Like many of his colleagues, who for years focused on understanding the fundamentals of atomic-level film deposition, George has just recently begun investigating film removal. George, Suri, Kessels, and others in the burgeoning ALE community gathered in Portland, Ore., last month to present their findings at a first-of-its-kind workshop devoted to the technique.

“We were hoping to get maybe 50 people,” said Eric A. Joseph, a research manager at IBM who organized the workshop. The event actually drew closer to 300 scientists, an indication of the high level of interest in this developing field. The one-day ALE workshop and the multiday international ALD meeting it followed were sponsored by AVS, a science and technology organization that focuses on materials research and processing.

As with the deposition method, ALE is based on surface chemical reactions that affect one atomic layer of a material per reaction cycle. In both procedures, researchers expose the surface to a pulse of a reagent that bonds selectively to just one type of surface functional group. Because of this chemical selectivity, the adsorption step is described as “self-limiting.” The idea is that after reagent molecules pair with all available functional groups, no additional reagent molecules can react with the surface. To remove the chemically altered surface layer, researchers convert the molecules to a volatile form that desorbs from the surface. Some researchers carry out this desorption step by using a second reagent and heat. Others use mildly energetic species, such as those found in plasmas, to gently nudge the molecules in the altered layer to desorb from the surface.

George’s group at the University of Colorado, for example, developed thermal chemical procedures for ALE of oxides commonly used in microelectronics. In one study, the team showed that treating an alumina surface with hydrogen fluoride and tin(II) acetylacetonate (Sn(acac)₂) selectively removes layers of the oxide.

By using X-ray reflectivity and quartz crystal microbalance techniques to monitor the surface reactions, the team demonstrated that the HF-Sn(acac)₂ reactions are indeed self-limiting and that the number of etched alumina layers increases linearly with the number of reactant cycles (ACS Nano 2015, DOI: 10.1021/nn507277f). The team proposes that the HF-Sn(acac)₂ mechanism involves fluorination and ligand-exchange steps that lead to formation of Al(acac)₃, SnF(acac), and water, which are stable compounds that readily desorb from the surface.

The team also demonstrated recently that the same reagents can be used to drive ALE of hafnium oxide, another common microelectronics insulator. George noted that the thermodynamics of fluorination suggest that these types of ALE reactions are not limited to metal oxides. They should also be applicable to other materials including metal nitrides, metal phosphides, metal arsenides, and elemental metals. The hafnium oxide study was one of several ALE papers published earlier this year in an issue of ECS Journal of Solid State Science & Technology focused on the technique (2015, DOI: 10.1149/2.0041506jss).

Several other scientists at the Portland meeting reported on ALE methods that involved energetic species. For example, Dominik Metzler, a graduate student working at University of Maryland, College Park, with materials science professor Gottlieb S. Oehrlein, described such a method for etching silica. In a study conducted with IBM’s Joseph and others, the team found that exposing silica to an argon plasma and periodically injecting a controlled quantity of C₄F₈ modified the SiO₂ surface with an angstrom-thick layer of a reactive fluorocarbon. Sputtering the surface with very low-energy Ar⁺ ions selectively removed the modified SiO₂ surface layer but not the underlying unmodified one (J. Vac. Sci. Technol. 2014, DOI: 10.1116/1.4843575).

Yasushi Sonoda of Hitachi, in Kudamatsu, Japan, described a similar but faster ALE process for etching polycrystalline silicon that involved chlorine, an electron cyclotron resonance plasma, and Ar⁺ ions. Meanwhile, Seiji Samukawa of Tohoku University, in Japan, reported on plasma-driven ALE with a twist. He noted that ions and ultraviolet light emitted by traditional plasma sources can induce defects in semiconductor materials. To avoid that kind of collateral damage, his group is developing an ALE device that etches with energetic beams of neutral species such as Ar and O₂.

Twenty-five years after the first ALE experiments were conducted, researchers are finally starting to take notice of this method for etching materials with atomic-scale precision. If those scientists and others can learn to control the surface chemistry that underpins ALE, they are sure to help the semiconductor industry continue its march toward ever smaller, faster, and more powerful electronics.