A study by a team of international researchers could lead to shatter-proof glass and spell the end of cracked phone screens.
The study, which has been published in Scientific Reports, was led by the Australian National University and the Institut de Physique du Globe de Paris in France.
Lead researcher Dr Charles Le Losq from the ANU Research School of Earth Sciences said that while we are surrounded by glass every day, we don’t actually know much about the structure of this material, with the research looking at links between the chemical composition and the atomic structure and properties of glass.
In particular, the research has looked at a type of glass called alumino-silicate, which is used in phone screens and screens for other mobile devices.
The team analysed glass that is mostly composed of aluminium and silicon oxides, and can also contain various elements such as sodium, potassium, calcium or magnesium. While it was thought that glass was structured randomly, Le Losq said at the microscopic level of a few atoms, it was found to actually be quite orderly.
Part of the research has used computer modelling – molecular dynamic simulations – that simulate the behaviour of the atoms in the glass in order to build a broad view of the glass structure and how it is influenced by chemical modifications.
While computer modelling has been used many times to study glass structure, Le Losq said present computing power means being able to simulate more and more complex glass. But as computer modelling in this kind of research is a specialist domain, it wasn’t without its challenges.
“It’s really demanding in terms of computing power. That’s also why it took us quite a long time to get the scenarios, because I started first experimenting in 2010 as a PhD in France and I’m now concluding this research as an ANU staff member,” he said.
While the computer modelling was carried out by his colleagues W Chen, Z Zhou and GN Greaves, Le Losq has a keen interest in solving volcanic problems with machine learning, which he was first introduced to when he was a post-doctoral student at the Carnegie Institution for Science in Washington D.C.
For example, he aims to use it to extract more information from spectroscopic measurements and glass properties.
This information could be used to calculate the relationship between the chemical composition of lava and its fluidity, and solve volcanic problems.
“It’s kind of a new area of research. It’s exciting because people are mostly using traditional techniques, but machine learning can bring us even further understanding of the topic,” he said.
These traditional techniques rely on knowing the physical equations that are related to observations, which isn’t always easy.
The idea behind using machine learning is that researchers can use it to solve problems without being required to know the mathematic equation. But using machine learning in the area of geochemistry can have its challenges, such as having only a small amount of data available.
“The simplest [way to overcome that] is to expand the size of the dataset by doing more experiments and more measurements of the properties of melt, glass and lava for various applications,” Le Losq said.
“If you start to have a piece of understanding about how the system behaves, instead of trying to model the entire system with a machine learning approach, you can model part of the system with physical equations and the parts you don’t know, you can model them with machine learning.”
While machine learning wasn’t used for the recent glass research because the dataset was too small and it wasn’t required, Le Losq says the technology could speed up that type of research in the future.
For example, it could predict which composition should be measured next to get the information that is needed in order to have a better understanding of the problem.
“That’s a way that machine learning is also very interesting for us – to use it in order to guide our experimental pathway,” Le Losq said.
Moving to complex glass
The first stage of the glass research involved trying to understand the glass structure, with the next step to be an expansion of this kind of research to a broader set of composition with other elements, such as calcium, magnesium and iron.
Future work will also involve trying to use the research’s concepts in an industry setting to see how glass properties can be influenced for specific applications like mobile phones. But Le Losq said how far away commercialisation is depends on how the glass industry responds to the research’s findings and the amount of manpower that is put behind it.
“We will continue for the next 12 months to work on the fundamental parts of those links between the glass structure, its properties and its chemical composition,” Le Losq said.