A new study led by Princeton University shows that by introducing trace amounts of oxygen or fluorine coatings on the surface of specific two-dimensional materials, the controllability of the plasma etching process can be significantly improved, which is expected to promote the manufacturing of a new generation of smaller, faster, and more energy-efficient computer chips. This breakthrough achievement provides a key process means for the introduction of ultra-thin new materials based on traditional silicon processes.
Today's commercial chips have integrated billions of silicon transistors per square inch, but silicon materials are gradually approaching physical limits in terms of size reduction and performance improvement. In order to continue the evolution of Moore's Law, the scientific research community has turned its attention to a type of ultra-thin transition metal dichalcogenide (TMD), hoping that it can work in conjunction with silicon to build future transistor structures. Among these candidate materials, molybdenum disulfide (MoS₂) is of particular interest, as it is only three atomic layers thick: a layer of molybdenum atoms in the middle, and a layer of sulfur atoms above and below.
To effectively integrate this type of TMD material into a chip structure, the manufacturing process often requires "peeling off only one layer" - precisely removing the top layer of sulfur atoms on the surface while leaving the lower molybdenum layer and the bottom sulfur layer intact. The current method commonly used in the industry is a plasma-based etching process, which uses high-energy charged particles similar to the physical state of the sun and stars to bombard the surface of the material and knock out atoms one by one.
The difficulty is that there is a distribution of ion energy in the plasma, and the process window is extremely narrow to remove the sulfur atoms on the surface without damaging the molybdenum atoms immediately below. If the energy is slightly lower, the sulfur atoms cannot be removed completely; if the energy is slightly higher, the molybdenum layer may be damaged, causing the entire material to lose its value as a high-performance channel layer. It is this "slight difference" process control problem that has restricted the large-scale application of TMD materials in advanced manufacturing processes for many years.
This work carried out by a research team from Princeton and other institutions, through large-scale computer simulations, found a seemingly simple but very effective "chemical assist" solution: functionally coating the surface of molybdenum disulfide with oxygen or fluorine before plasma treatment. Simulation results show that this extra step significantly widens the safety process window, making it easier to remove only the top layer of sulfur atoms without damaging the underlying molybdenum layer.
Research shows that to remove one sulfur atom on the surface of untreated molybdenum disulfide, an incident energy of approximately 30 electron volts is required. Once pre-coated with fluorine, this energy threshold can be reduced to about 10 electron volts; with oxygen coating, it can be reduced to about 14 electron volts. In comparison, the energies corresponding to the two results of "removing sulfur" and "piercing the molybdenum layer" in the original situation are very close, making it difficult to avoid damage to the main body of the material during actual processing.
With oxygen or fluorine coatings, the energy required to detach sulfur atoms is significantly reduced, creating a greater distance from the "damage threshold." Under this wider operating window, even if there are certain fluctuations in the ion energy in the plasma, there is still a greater probability that only the selective removal of sulfur atoms on the surface will be triggered without damaging the molybdenum layer at the core of the structure. This difference is critical in the pursuit of atomic-level precision in semiconductor manufacturing.
The research team pointed out that the key to the new strategy is to "let chemical reactions help" rather than relying entirely on the physical impact of plasma particles. When high-speed ions hit the MoS₂ surface pre-covered with oxygen, two nearby oxygen atoms will tend to combine with a sulfur atom to generate a molecule of sulfur dioxide gas. This molecule is thermodynamically very stable and is easier to spontaneously detach from the surface of the material, which is equivalent to "taking away sulfur through a chemical reaction."
Similarly, if a fluorine coating is used, an intermediate compound containing sulfur-fluorine bonds will be generated, which is also easier to break than the original S-Mo bonds, thereby achieving gentle and selective surface etching. The first author of the paper, Yury Polyachenko, a graduate student in the Department of Chemistry at Princeton University and a 2025 summer member of the Princeton Plasma Physics Laboratory (PPPL), said that they did not directly break the strongest chemical bonds inside the material, but first generated "better" intermediate products through functionalization, and then removed them with lower energy.
This result was published in The Journal of Physical Chemistry Letters and discussed in detail the impact of different surface functionalization methods on energy barriers and damage risks. The current simulation work is mainly focused on answering the question "Will it be damaged?" In the next phase, the team plans to further quantify the specific defect types and densities produced under different process conditions, thereby providing more operational parameter guidance for the industry.
The researchers also plan to extend this idea to a wider range of material systems, such as replacing molybdenum with tungsten, replacing sulfur with selenium, etc., to see whether this combination of oxygen/fluorine functionalization and plasma selective etching is also applicable. If similar effects can be reproduced in a variety of TMD materials, it will open up more space for the selection of ultra-thin channel materials and the design of multi-material stack structures in the future.
The research was funded by the U.S. Department of Energy's Office of Science and was conducted under the framework of the Extreme Lithography & Materials Innovation Center microelectronics research project undertaken by the Princeton Plasma Physics Laboratory. Relevant large-scale numerical simulations are mainly completed on the National Energy Research Scientific Computing Center (NERSC) and the Stellar, Della and Tiger high-performance computing clusters of Princeton University.