Machine learning simulations uncover how plastic transforms into hard carbon
Seventy-five years after Rosalind Franklin's important research on carbons that do not turn into graphite, chemists at the University of Oxford have worked out how hard carbon forms, atom by atom. This solves a mystery that has lasted for a decade.
The study was published in Chemical Science. The researchers used computer simulations powered by machine learning to model how polyvinylidene chloride (PVDC) breaks down when heated. Their results show how its structure stops hard carbon from forming the neat, ordered crystals found in graphite.
These findings give scientists a computer-based method for designing custom materials for the anodes, or negative electrodes, in next-generation sodium-ion batteries.
How atoms form hard carbon and stop it from turning into graphite
Hard carbon is an important material for the anodes in sodium-ion batteries. It is a low-cost option made from elements that are widely available on Earth, offering an alternative to conventional lithium-ion technology. Graphite is made of flat layers of hexagonal carbon rings stacked on top of one another. Hard carbon has a curved, irregular structure that stays disordered even when exposed to extreme heat.
The Oxford team combined highly accurate quantum-mechanical calculations with fast machine-learning algorithms to map out how the starting PVDC polymer breaks down, step by step:
- Removing chlorine and hydrogen:
- Heat removes hydrogen and chlorine atoms from the PVDC chains. This exposes carbon atoms that can react easily with other atoms.
- Building connected networks:
- These reactive sites join nearby pieces of the polymer, creating a tightly connected, three-dimensional network of carbon atoms.
- Creating shapes that cause curves:
- Instead of forming only six-sided carbon rings, which are the basic building blocks of flat graphite, the network also forms carbon rings with five and seven sides.
- Changing the structure's shape:
- Just as five-sided shapes can help a pattern of hexagons curve into a round football, these rings with different numbers of sides bend and twist the carbon structure. This stops the layers from lining up in the flat, orderly way seen in graphite.
The research team and computer methods
The study was led by DPhil researcher Litong Wu, working with former students Zitong Wu and Dr Zakariya El-Machachi. Professor Volker Deringer supervised the team at the University of Oxford's Department of Chemistry.
Solving the limits of computer calculations
Traditional quantum-mechanical methods require too much computing power to model high-temperature chemical reactions involving thousands of disordered atoms over long periods. The researchers overcame this problem by using machine-learned models of the forces between atoms. These models predict atomic forces with quantum-level accuracy while making the simulations much faster.
By simulating how the starting material breaks down at the level of individual atoms, researchers can directly connect its original chemical makeup to the final electrochemical performance of battery anodes.