PECM to Aid Researcher in Study of Complex Fluids

By Herbert Morgan, NCSA

The Performance Engineering and Computational Methods (PECM) group plans to assist University of Illinois researcher Erik Luijten and his colleagues in improving the performance of their novel simulation codes. As a result, the researchers expect their parallel codes to run more efficiently on NCSA machines.

Luijten, an assistant professor in the Department of Materials Science and Engineering at UIUC, and graduate student Jiwen Liu devised a new simulation algorithm that speeds up computer simulations of complex fluids. They applied this method to explain the original experiment that motivated their research.

Background

Figure 1. In a preliminary simulation, large particles (green spheres) cluster when too few small particles (blue dots) are added to the simulation box.

In 2001, University of Illinois researchers, led by Jennifer Lewis, discovered a fundamentally new approach for tailoring the stability of colloidal suspensions. Colloidal suspensions are complex fluids used in numerous applications ranging from advanced materials to drug delivery. The researchers dubbed this technique "nanoparticle haloing," a self-organizing process that imparts stability to naturally attractive colloidal microspheres, decorating their superficial areas with highly charged nanoparticles.

Nanoparticle haloing can be used to control the phase behavior and structure of materials assembled from colloidal systems. By tailoring interactions between particles, the desired degree of colloidal stability can be engineered.

As a result, researchers can create designer colloidal fluids, gels, and crystals. The ability to control colloidal forces and phase behavior depends on the charge and the size of the nanoparticles. Lewis and her colleagues continue to study the structure and flow behavior of colloidal fluids assembled from these microsphere-nanoparticle mixtures.

Simulating the Experiment

Luijten, who was familiar with the work of Lewis and her colleagues, thought he could simulate it. But this was a daunting challenge because many complex fluids contain particles of widely different sizes that also move at vastly different time scales. A simulation that faithfully captures both the motions of the faster small particles and the slower large particles would be extremely slow and impractical. This type of simulation, though, is precisely what was required because the mixture of two very different types of particles is what led to the new behavior.

Luijten and Liu developed a new simulation algorithm, which they used to crack this problem. Their algorithm generates a random pivot point, picks a particle for a point reflection, and pivots it 180 degrees to a new position. Any nearby particles, either within the field of influence of the pivoted particle's old or new position, are candidates for joining the cluster. These particles similarly pivot and contain their own field of influence. As long as particles are within the realm of influence of the growing cluster, the cluster will continue to grow. Otherwise, growth will cease, and the algorithm picks a new pivot point where new cluster building may begin.

Luijten and Liu's simulation method thus identifies natural groups of particles based on the elementary forces that act between them. Their method accelerates the simulation of complex fluids, and it becomes more advantageous as the disparity in particle size increases.

Achieving a Stable Fluid Phase

In an effort to find a stable fluid phase, Luijten and his team created a simulation box, which contained 1-micron-sized microspheres in a fluid. These spheres, left alone, clustered together in the fluid. The team then added small particles to the mix.

Figure 2. Stable fluid phase. Luijten's algorithm achieves this state by increasing the number of small particles in the simulation box, which counteracts the attractive force that causes the large particles to cluster.

They discovered that when they added just a few small particles, the large particles still clustered (Fig. 1), which is the unwanted behavior. As more small particles were added, the large particles no longer clustered (Fig. 2), and a stable fluid phase occurred. But, at some point, if even more small particles were added, the large particles clustered again.

Achieving a stable fluid phase is the goal because if large particles tend to cluster together they form irregular clumps. In experiments, these clumps then descend and form a messy mass at the bottom of the beaker. However, if the particles descend without clustering, they form into a regular crystal.

Each particle is a thousand times larger than an atom. Even though the final assembly looks like a crystal, it is constructed of much larger silica building blocks. From this, one can create structures that allow light manipulation because the diameter of these particles is of the same length scale as the wavelength of visible light. That, then, allows the creation of devices that use or transmit light in certain ways.

"In order to get this effect," says Luijten, "you want to know how small should the small particles be, how many should I add? If I add smaller ones, can I add less? It turns out the small particles carry an electrical charge. Does it matter what charge it is? If I increase the charge, can I add less?"

Luijten has been running his simulations on Mercury, NCSA's 512 1 GHz processor TeraGrid cluster, and has been appreciative of the support that he has received from NCSA staff. In the near future PECM will be assisting him with code optimization.

For further information, see http://www.mse.uiuc.edu/faculty/Luijten.html.