Three linked grids reveal unsuspected property of clay

The finding emerged from some of the largest computer simulations ever attempted which required the pooled resources of supercomputers on three grids across two continents. The research is expected to provide insights into the properties of an important class of new materials, clay polymer nanocomposites, which are under investigation for many applications, for example as materials for use in car bodies.

Supercomputers on the UK National Grid Service, the US TeraGrid and DEISA (EU Distributed European Infrastructure for Supercomputing Applications), linked by dedicated high speed optical networks including UKLight, were pressed into service. Professor Peter Coveney and colleagues from University College London (UCL) used these resources to produce simulations of five computer models of the platelets that lock together to form clay sheets, the difference between each model being its size and complexity. Each model simulated accurately the motion and interactions between all the atoms in two sheets of clay separated by a layer of water and sodium ions. In the largest model, the motions of nearly 10 million atoms were taken into account. The simulations were run over timescales of up to 2 nanoseconds (a nanosecond is a billionth of a second).

By using distributed high performance computers linked by grids, it was possible to perform the many and vast simulations concurrently. Without such a facility, the time taken to perform the simulations on one supercomputer alone would have been too long to make the study practicable. The team was able to access these resources across multiple domains using grid middleware, called the Application Hosting Environment, which was originally developed under RealityGrid, an EPSRC funded e-Science project.

Data from the simulations were returned to computers back at UCL for visualisation. “Optical networks enabled us to link these grids together. The amount of data we produced is very large and UKLight is very valuable for getting the data back to us here,” says Professor Coveney.

The visualisations revealed the undulations. “As we moved from smaller to larger models we began to see collective undulations – the clay platelet sheets fluctuate up and down,” says Professor Coveney. This property, which was not known before in clay materials, is on too small a scale to be easily verified by experiment. But it has implications for the properties of clay on an ordinary scale which can be computed and then compared with experiment. For example, the team has used the response of the clay sheets to the undulations to calculate their elasticity (or Young’s modulus).

As a next step, the group plans to simulate clay platelets embedded in a polymer matrix. Such clay-polymer nanocomposites are under investigation for a number of applications ranging from car bodies and other automotive uses, through oilfield technology to drinks packaging. Compared with polymers alone, they have far greater mechanical strength, improved fire retardant properties and they make better barriers to the diffusion of gas. “These simulations will give us a better understanding of the properties of these new and important materials,” says Professor Coveney.

The work is published in the Journal of Physical Chemistry vol.111 pp8248-8259 2007.