Development and application of a machine learning potential for diamond surfaces and their sliding interfaces with silicon nitride and silica: effect of passivation, orientation and load
This study develops a reactive machine learning potential to simulate diamond sliding against silicon nitride and silica at scales beyond first-principles methods, revealing how surface passivation, orientation, and load govern interfacial bonding and friction through distinct shear regimes.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Diamond is famous for being the hardest material known to science, a quality that makes it invaluable for cutting tools, precision bearings, and the delicate machinery that moves silicon wafers in computer factories. Yet, even this unyielding substance wears down when it rubs against other materials, and the speed of that wear depends entirely on what is happening at the very surface. If the air is dry or the space is a vacuum, the exposed carbon atoms on the diamond can reach out and grab onto the material it is sliding against, forming strong chemical bonds that cause friction and eventually tear atoms away. However, in humid air, water molecules break apart and coat the diamond with a protective layer of hydrogen and oxygen groups, preventing those bonds from forming and allowing the diamond to slide smoothly. Understanding exactly how this protection works, and how it holds up under the immense pressure of real-world machinery, has been a difficult puzzle for scientists because the events happen too fast and involve too many atoms for traditional computer models to track over useful periods of time.
To solve this, researchers at the University of Bologna created a new type of computer program that acts like a highly accurate, ultra-fast simulator for how atoms interact. Instead of using the slow, heavy calculations of standard physics, they trained a machine learning system on thousands of detailed snapshots of how carbon, silicon, nitrogen, oxygen, and hydrogen behave together. This system learned the rules of the game from high-level physics data, allowing it to predict how these atoms would move and bond in a fraction of the time. The team then used this new tool to watch what happens when a diamond surface slides against a piece of silicon nitride, a hard ceramic material, under a wide range of conditions. They tested four different diamond crystal faces, some clean and dry, others covered in the protective water-derived coating, and pushed them together with forces ranging from a gentle touch to pressures as high as 23 gigapascals, which is roughly the pressure found where a hard diamond tip digs into a softer surface.
The simulations revealed that the behavior of the sliding surfaces falls into two distinct categories, determined by whether the diamond is protected and how hard it is being pressed. When the diamond surface is covered with the protective layer of hydrogen and oxygen groups, or when it is a specific type of diamond face that is naturally less reactive, the two materials remain chemically separated at low to moderate pressures. In this state, the diamond and the ceramic slide past each other like two rigid blocks on a perfectly smooth plane, with almost no chemical bonds forming between them. The friction is low, and the wear is minimal because the atoms never get a chance to grab hold. However, as the pressure increases, the force becomes strong enough to break through this protective shield. At very high loads, the hydrogen and oxygen groups are pushed aside or transferred, allowing the carbon atoms of the diamond to form strong chemical bridges with the silicon and nitrogen of the ceramic.
Once these chemical bridges form, the nature of the sliding changes completely. Instead of the two surfaces sliding past one another, the bond between them becomes stronger than the ceramic material itself. The friction then stops happening at the interface and moves into the softer ceramic layer, which begins to flow and deform like a thick paste under the stress. The researchers found that for the clean, unprotected diamond faces, this strong bonding happens almost immediately, even at low pressures, causing the friction to jump to a high, constant level determined by how easily the ceramic can flow. For the protected surfaces, the friction remains low until the pressure reaches a critical point where the protection fails, at which moment the friction spikes to match the high levels of the unprotected surfaces.
The study also showed that not all diamond faces are equally vulnerable. One specific arrangement of atoms, known as the Pandey-reconstructed face, is naturally more resistant to bonding than the others, staying protected up to higher pressures even without the water-derived coating. In contrast, another face, the C(110) orientation, is highly reactive and forms bonds easily, leading to more wear. The team discovered a direct link between the number of chemical bonds formed at the interface and the amount of friction generated: as long as the bonds are sparse, the friction rises steadily with the number of connections, but once the interface is fully bonded, the friction levels off because the weakness shifts to the ceramic material itself. This work provides a clear, quantitative picture of how environmental conditions and surface chemistry control the wear of diamond, explaining why these materials perform so differently in humid air compared to dry vacuum, and offering a powerful new way to simulate and predict wear in complex industrial applications without the limitations of older methods.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.