The virtual lab

Materials World magazine
1 Feb 2015

A deadly crack in an aeroplane wing can begin life as a single broken chemical bond. Now researchers have found a way to simulate material properties from the atomic level up. Simon Frost reports. 

Researchers at UCL’s Centre for Computational Science have developed an advanced multiscale simulation system to predict the properties of polymer-clay nanocomposites (PCN) based on their molecular structures. The methods could have applications in modelling a wide range of materials.

Lead researcher Peter Coveney explains, ‘Clays are structured like a pack of cards – sheets of aluminosilicates with lateral dimensions of around a micron, stacked with a nanometre of separation.’ Impregnating these spaces with different polymers creates composites with useful properties that, until now, were achieved through trial and error. Toyota brought PCN to prominence in 1985, when it discovered that clays could be dispersed into a polymer matrix – in this case, nylon – to create light-yet-strong composite materials that are desirable in car manufacture.

Using supercomputers housed at UCL, the Daresbury Laboratory, Cheshire, and the new national supercomputer, ARCHER, at the University of Edinburgh, Coveney and his colleagues, James Suter and Derek Groen, devised a system that characterises the properties of composites from the quantum mechanical level up through the atomic level to a more coarse-grained particulate description.

’It’s all about linking the smallest to the largest scale, carrying through the substantial signature of the material’s chemistry. It’s fundamental to match progressive levels of description. You start at the quantum mechanical level, where you’re dealing with electron densities and charges, and then move up to the next level of the hierarchy – molecular dynamics, solving Newton’s equations of motion, and eventually we aim to reach the full-blown engineering scale that’s familiar to modellers in materials engineering using finite element techniques,’ Coveney explains.

The virtual lab also incorporates processing conditions. ‘Do you shear them? Stir them? Extrude them?’ asks Coveney. ‘These conditions are important in materials science, so having a simulation technique that can include them is crucial.’

With today’s supercomputers, separate machines carry out each part of the process. ‘The algorithms differ quite substantially from one code to the next, so a specific machine will typically run one of these codes but not all of them,’ he says. The current crop of supercomputers is known as petaFLOPS and can run 1015 floating point operations per second, but the next generation, exaFLOPS, will run 1018, allowing a greater range of functions. ‘In the next five to 10 years it will become increasingly easy to do these things. The exaFLOPS machines will be able to carry out all of the calculations on the same machine, which will make it much more user-friendly.’

Understanding how clays interact with other contaminants is also beneficial to the oil and gas industry – Coveney’s longstanding interest in clays began when he worked for the oilfield service giant Schlumberger. ‘60–70% of all formations you drill through in the search for oil and gas are shales, a form of compacted clay,’ he says, ‘so it’s very important to understand their properties.’ Oil wells are often treated with organic compounds to prevent water ingress and damage to the well bore, but this can make the clay unstable.

It’s not only clays that this system could benefit, though. ‘Aluminosilicates are only about five atoms thick – not as thin as graphene, but still quasi-2D, so understanding how polymers cause the next stage of aggregation or dispersion and exfoliation to occur in clays is something that people want to know about graphene, too.’ Coveney believes that the methods could help to solve the problem of exfoliating sheets of 2D graphene from graphite.

Other layered materials, such as layered double hydroxides (LDH) are similarly structured. ‘Aluminosilicates are negatively charged, with cations in the interlayer space. With LDH it’s the opposite – they’re positively charged, with anions in between layers, so they can be used for different applications. Working with these kinds of material is the most obvious direct extension of this research.’ He even harbours an interest in applying the methods in biological and medical applications. ‘A heart attack, for example, is initially triggered by molecular events. In a broader sense, the linking of the small scale to the large is important in many areas of science.’

View the paper, Chemically specific multiscale modeling of clay-polymer nanocomposites reveals intercalation dynamics, tactoid self-assembly and emergent materials properties, at