Despite millennia of glass making, the physics of what happens when silica sand is melted and cooled is still something of a mystery to researchers. Now, in a groundbreaking window into what happens on a molecular level as liquid silica cools into glass, researchers at the University of Wisconsin-Madison have used sophisticated computational techniques to create a compelling model of how and why flow behaviors change during the glass-cooling process.
The study was led by Bu Wang, an assistant professor of civil and environmental engineering and materials science and engineering at UW-Madison, and Zheng Yu, a PhD student in materials science and engineering. It appears in the July 2022 issue of the journal Physical Review Letters.
When a glass-forming liquid cools, it becomes more viscous until it reaches the point that it is solid glass. For many liquids, the viscosity increases at an accelerating rate as the temperature decreases. However, in some glass-forming liquids, including liquid silica, the viscosity increase is more complex, with an initial large rise followed by a more gradual change. This type of transition is called the fragile-to-strong transition (FTS).
Researchers have previously characterized the FTS transition using thermodynamics (heat) and configurational entropy (particle arrangements), but those theories have not captured the process exactly. “What we wanted to do is just look at the patterns at the atomic level to see what is happening when the system goes through this transition,” says Wang.
In their computational simulation, supported by resources from the National Science Foundation-funded Materials Research Science and Engineering Center at UW-Madison, the team assigned each liquid silica atom a probability that it will change position, or “hop,” relative to neighboring atoms. “From there, we can predict the motion of a large number of atoms and connect those motions to the whole system’s behavior,” says Yu. “The key insight here is that on the system level, the diffusion or flow dynamics are actually rooted in the individual atom’s motion. That’s a part of what has not been shown before.”
Based on the atom-hopping probability, the researchers’ model estimated the diffusion coefficient of the liquid glass (describing how quickly atoms move) at any given temperature. The researchers found that, in liquid silica, there are two distinct types of atoms. One type of atom hops very easily, and their concentration decreases very fast as the temperature decreases, while the other type of atom is not affected by temperature as strongly. During cooling, the first type of atom rapidly disappears, leading to the FTS. If researchers think of this type of atom as defects in the glass, the finding would allow them to apply many theories used to describe dynamics in crystals to the much harder-to-characterize dynamics in glass.
Wang says that prior to the development of machine learning techniques, classifying atoms based on their dynamic behaviors and then using the information to predict glass behavior would have been too complicated. Now, armed with data and new understandings from the simulation, engineers may be able to have more precise control over the glassmaking process. For example, they might be able to tune glass to make it better for use as screens, coatings, sealants, electrolytes in batteries, and pharmaceuticals, among other uses.
Other UW-Madison authors on the paper include Dane Morgan, a professor of materials science and engineering, and Mark Ediger, a professor of chemistry. This research was primarily supported by NSF through the University of Wisconsin Materials Research Science and Engineering Center (DMR-1720415). This work used the Extreme Science and Engineering Discovery Environment (XSEDE), which is supported by National Science Foundation Grant No. ACI-1548562.