Designing electronic devices using supercomputers

Density functional theory calculation of the molecular structure of three benzenedithiols sandwiched between two gold surfaces. Click here for more photos.

A large supercomputer at ORNL is being used to learn more about the best ways to design electronic device components on a very small scale.

Designing Nanocircuits

The successor to the silicon chip may be a nanoscale device—a self-assembled monolayer of organic molecules of benzene (a ring of six carbon atoms bonded with four hydrogen atoms) attached to sulfur atoms at each end. Because sulfur (thiol) has an affinity for gold, a single layer of these benzenedithiol molecules can be sandwiched between thin gold contacts. Scientists at Rice and Yale universities have induced self-assembly of such a device by dipping a gold surface into a beaker of benzenedithiol molecules. In large numbers these molecules attached themselves to the gold surface.

The scientists added nitrogen-containing (nitro) groups to the molecule’s center benzene ring. The resulting perturbed electron cloud made the asymmetric molecule twist when an electric field was set up by applying a voltage between the gold contacts. When the molecule twisted, current flow through the “molecular wire” was blocked. When the voltage was removed, the molecule adopted its original shape, allowing current to flow again.

Such a device, if fabricated on a large scale, could be used as an ultrafast on-off switch, a key to creating ultrasmall, highly dense computer circuits required to make computers fast and powerful enough to mimic the human brain. Or the device could be used to make superior computer memory elements. A charge can be stored on the nitro group to prevent electrical conduction (binary 0), or the group can have no charge, allowing conduction (binary 1). Such a molecular memory cell retains a stored bit for nearly 10 minutes. By comparison, today’s silicon-dynamic, random-access memories must be refreshed by an electrical current every 20 milliseconds. The new type of memory would save energy, allowing laptop computer batteries to last 100 times longer.

Such a concept is being modeled computationally on ORNL’s IBM supercom-puter, dubbed Eagle, in a Laboratory Directed Research and Development project. ORNL’s David Dean, Bill Butler (now at the University of Alabama), Peter Cummings (an ORNL-UT Distinguished Scientist), Predrag Krstic, David Schultz, Mike Strayer, Jack Wells, and Xiaoguang Zhang are running the calculations using a modified version of NWChem, a computational chemistry code.

“Using ab initio methods, we modeled the self-assembly and electrical conductivity of five benzenethiol (BT) and benzenedithiol (BDT) molecules on a gold surface,” says Dean. In the example shown in the illustration, two gold lattices are shown on the top and bottom. Three BDT molecules are seen in the middle area. This particular configuration has 70 atoms and 590 active electrons. A single calculation of this type requires 46.67 hours on 80 nodes of Eagle, or 14,930 processor hours. The single-particle wave functions resulting from this calculation will be used in a conductance calculation to determine the current-voltage characteristics of this molecular device.

The researchers have also performed preliminary molecular simulations of self-assembled monolayers composed of BT molecules on the [111] surface of gold. They included state-of-the-art force fields generated through electronic structure calculations. Both molecular dynamics (MD) and Monte Carlo (MC) simulations are being used. Gibbs ensemble MC simulation is being used to establish the equilibrium between adsorbed monolayers of BT and BDT with a low-density solution. MD is then being used to equilibrate the structures thus found. “The structure of the adsorbed monolayers appears to be consistent with available experimental results,” Dean says.

To produce a useful device, self-assembly must be combined with fabrication methods such as photolithography. The ORNL scientists will model how best to assemble these molecules and align them with the gold contacts to optimize electrical conductivity.

Device Design and Performance

Marco Buongiorno Nardelli, who holds a joint position with ORNL and North Carolina State University (NCSU, one of UT-Battelle’s core universities), has been exploring the feasibility of using carbon nanotubes in nanoscale electronic devices. He is currently using Eagle to run his own suite of codes simulating electron transport in carbon nanotubes in contact with other materials.

In one project in which he provided computer modeling, experiments at the University of North Carolina (UNC) at Chapel Hill have shown that it is possible to build a nano-rheostat, similar to a dimmer light switch. In such a device, a carbon nanotube—a cylinder resembling rolled-up chicken wire because its carbon atoms are arranged in a hexagonal configuration—is placed on a sheet of graphite whose carbon atoms also have a hexagonal arrangement.

“If you place the carbon cylinder on the graphite sheet so that the carbon atoms of both are aligned, a current will flow at the interface,” Buongiorno Nardelli says. “As you rotate the carbon cylinder on the graphite sheet, changing the angle between the atoms in the system, you get increased electrical resistance and reduced current flow. As the atoms become aligned, you get low resistance and high current flow.”

Computational simulations by Buongiorno Nardelli verified that the interface between a carbon nanotube and graphite gives tunable resistance (as in a dimmer switch). His theoretical predictions on the feasibility of a nano-rheostat agreed with the UNC experimental results. The work was published in Science magazine in 2000.

If carbon nano-tubes are to be used as nanowires or other components in nanoscale devices, electrons must flow between these nanotubes and metal contacts in the device. In some experimental devices, high resistance at the tube-contact interface can make the mechanism of electron transfer quite inefficient. Buongiorno Nardelli and his NCSU colleagues have used computer modeling to address the question of why some nanodevices have better performance than others.

“In some devices,” Buongiorno Nardelli says, “electrons in the carbon nanotube stay in the tube and electrons in the aluminum stay in the metal. Our simulations suggest that contacts can be improved by mechanical deformations. For example, if a carbon nanotube sandwiched between two aluminum contacts is squeezed and deformed, new bonds form between the carbon and aluminum atoms, increasing electron flow at the tube-contact interface.”

The strength of carbon nanotubes is also of interest to Buongiorno Nardelli. Of all materials, carbon nanotubes have the highest tensile strength. They are 100 times stronger than steel but have one-sixth its weight. Scientists propose using carbon nanotubes as fibers in a polymer composite to form stronger structural materials for aircraft, spacecraft, and suspension bridges.

Computational simulations by Buongiorno Nardelli and his colleagues have shown that the geometry—the arrangement of the carbon hexagons along the nanotube—influences tube strength. “Our simulations,” he says, “predicted that whether a nanotube is brittle or ductile depends on the temperature at which it was deformed, the orientation of the hexagons with respect to the tube’s axis, and the amount it is stretched—that is, strain.”

Carbon nanotubes are very small, but simulations of their behavior in nanoscale electronic devices require a large amount of computer capacity.