Chemical experiments and predictions by computer

Solid-like structure (with short-range order in three dimensions) of n-dodecane narrowly confined between solid surfaces is induced by interfacial forces. This effect may explain the orders-of-magnitude higher viscosity observed in confined-fluid experiments compared with bulk fluid values. Click here for more photos.

The term “chemistry” conjures up images of people in white lab coats, pouring liquids from test tubes into a beaker. But an increasing number of chemists do most of their chemistry on computers, partly to save money and increase safety. Simulating chemistry experiments on the computer instead of doing them in the lab can produce chemical reactions and properties with a faster turnaround and better accuracy. Computational chemistry also can aid understanding of actual lab results.

David Bernholdt, who helped develop the NWChem computational chemistry software package while a postdoctoral scientist at DOE’s Pacific Northwest National Laboratory, leads the chemistry initiative at DOE’s Center for Computational Sciences at ORNL. Bernholdt knows the value of computational chemistry, so he is working to develop scaling techniques that will allow chemists to work with even larger molecules on ORNL’s new supercomputers.

“Computational chemistry allows you to predict which chemical compounds are more likely to give the desired property or result,” he says. “For instance, computational chemistry was used by Kodak’s color film developers to predict which chemicals would produce the right colors yet still hold up during chemical processing to develop the film. It saves you from having to synthesize lots of different chemicals and then screen them for the needed properties, such as what might be effective in a therapeutic drug.”

At ORNL several researchers have been using supercomputers to work on computational chemistry projects. Don Noid and Bobby Sumpter, both in ORNL’s Computer Science and Mathematics Division, were the first to computationally model the dynamics of fluid flow inside carbon nanotubes—tiny cylinders resembling rolled-up chicken wire. Their molecular dynamics simulations showed that argon slowed down more quickly than helium and that both fluids slowed down faster in a flexible nanotube than in a rigid one.

More recently, they found in their simulation studies with Clemson University researchers that hydrocarbons break up more slowly when heated in carbon nanotubes than in a furnace under vacuum. They believe the rate of breakup of these polymers (similar to crude oil cracking under pyrolysis) is altered by the chemistry of confined spaces.

Working with Mike Barnes of ORNL’s Chemical Sciences Division (CSD), Noid and Sumpter showed both computationally and experimentally that a nanosized polymer droplet can be formed from materials that don’t normally mix and that the droplet can be forced through a micron-sized orifice. The resulting particles exhibit unique properties that could make them useful for optical displays and industrial coatings. The researchers also helped develop a software tool to calculate how far and which way thousands of atoms move relative to their neighbors (vibrational modes), providing insight into the structure and behavior of various materials.

Hank Cochran, Peter Cummings, and Shengting Cui, all with both CSD and the University of Tennessee at Knoxville, are doing molecular simulations of microdispersions stabilized by surfactants in supercritical carbon dioxide (CO2). Surfactants act like detergent by reducing surface tension; they are being used in several new dry cleaning establishments where CO2 replaces traditionally used carcinogenic solvents. DuPont is building a big plant in North Carolina to produce Teflon and other fluoryl polymers using other new surfactants with supercritical CO2 instead of ozone-destroying chlorinated fluorocarbons.

The CSD group is also simulating the effects of velocity gradients (shear flow) on the arrangement and behavior of long-chain molecules in such situations as during the extrusion of filaments and the melting of polymers. In addition, they are looking at the behavior of lubricants in the narrowest nanoscale separations between motor vehicle components, which can be substantially different from their behavior in bulk.

Cochran and Cui are simulating the behavior of water that contains salts and DNA or proteins in channels thousands of times smaller than a hair in a “nanofluidic lab on a chip.” When developed, under the leadership of lab-on-a-chip inventor Mike Ramsey of CSD, such a device might be used in a doctor’s office for ultrafast DNA sequencing of blood drops from individual patients for rapid disease diagnosis. Liquids containing DNA or protein could be moved by the influence of electric fields through ultrasmall channels, which might enable the increase of sequencing and diagnostic speeds by a million to a billion times. In their simulations, the researchers take into account the electric-field and surface forces that extend through the liquid inside the nanoscale channel. These simulations help to guide and interpret Ramsey’s experiments.

As devices get smaller, computational chemistry is likely to play a bigger role in figuring out how to make them work.