Research

Atomic-level Computer Simulations in Materials Science

by Sara Latta, Science Writer

What do improved automobile catalytic converters, more efficient solar cells, stronger machine tools, and better adhesives have in common? They are all among the target applications for the pioneering atomic level computer simulations carried out on NCSA's CRAY Y-MP system by Jim Adams, UIUC assistant professor in materials science engineering.

The unifying thread running through the work in Adams' group of six graduate students and three postdoctoral fellows is the fact that little atomic-level simulation has been done on any of his project areas.

From diamonds to polymers

One of Adams' projects focuses on diamond film growth--the deposition of a very thin film of carbon atoms in the characteristic crystalline structure of diamond through a process called chemical vapor deposition, or CVD. There are a couple of possible applications for diamond films, according to Adams. ``Diamond films might be a very good way to make much harder coatings for materials. Machine tools coated with a diamond film would last a lot longer,'' says Adams. ``And, because diamond is also the world's best heat conductor, diamond films could have important applications in microelectronics.'' The arrangement of atoms on the surface of diamond films has been impossible to determine experimentally. According to Adams, ``We've done the first realistic calculations of the structure of the diamond surface.''

Amorphous silicon thin films-also manufactured by a CVD process-- are the primary component of those solar cells that power calculators and a host of other devices. Although amorphous silicon (so-called because of the random, rather than crystalline, arrangement of the silicon atoms) is quite inexpensive, it is only about 10% efficient in converting sunlight into electricity. Unfortunately, according to Adams, that efficiency drops to 8% when the material is exposed to sunlight for 24 hours--effectively a 20% decrease in efficiency.

``We're really the first group to determine the atomic defects in amorphous silicon that cause this degradation in efficiency,'' says Adams. ``We'd like to prevent that. Solar cells produce electricity for only slightly more than the cost of present-day nuclear power reactor,'' he continues, ``and the price keeps going down. It's possible that we could cover substantial areas of the desert with solar cells, for example, and supply most of the electricity for the U.S.''

Adams is also using atomic-level simulations to investigate the adhesion of polymers to different types of surfaces, including metals and metal oxides. ``Polymers, in general, don't stick very well to other materials,'' says Adams. By simulating the addition of different types of side groups to the polymers, Adams determines the modifications that help polymers form stronger bonds with the substrate--without otherwise affecting the properties of the polymer. ``We hope that these studies can contribute to everything from stronger adhesives to better paints and coatings,'' says Adams. Perhaps not surprisingly, Adams' group is the first to study the atomic-level bonding of functional groups in polymers to metals, metal oxides, and semiconductors.

Better paints, adhesives, coatings, and automobile engines plus more efficient production of electricity--for the home and industry--is available via materials science.

Catalytic properties

Automobile catalytic converters contain small clusters--on the order of several hundred atoms--of metals such as platinum, palladium, and rhodium. These clusters catalyze the conversion of deadly gases such as carbon monoxide and nitrous oxide to carbon dioxide, nitrogen, and oxygen.

``Rhodium is the element of choice,'' says Adams, ``except for the fact that the whole world's supply of rhodium comes from two mines in South Africa. It's incredibly expensive. We're working to determine the structure of the metal clusters and how the structure affects their catalytic properties. Eventually, we hope to use the results to find less expensive, or more effective, replacements.''

Using quantum methods

Experimental data in these projects are scarce, so Adams uses quantum mechanical methods to simulate these systems. While the methods are extremely accurate, they are also slow: he can only include 100 to 200 atoms in his simulations, for very short periods on the CRAY Y-MP system. ``There is no way we could do these calculations without NCSAÕs CRAY Y-MP or their IBM RS/6000 workstations,'' says Adams. In the future, Adams' IBM allocations will be transferred to the Cornell Theory Center, which will be the new national metacenter's IBM RISC-cluster hub.

Metallic glasses and interfaces

In yet another potential industrial application, Adams is studying the thermodynamics of metallic glasses. ``Metallic glasses are just metals that have random (rather than crystalline) structures--but they have wonderful magnetic properties that make them ideal for use in the core of power transformers,'' Adams says. They are made by cooling a metal extremely rapidly--``you cool a molten red hot metal to something you can hold in your hands, in about a hundredth of a second,'' according to Adams. His group is the first to undertake a study of the thermodynamics and stability of metallic glasses. They would like to understand why some elements, and not others, will form glasses under this process.

Lastly, Adams' group studies metallic interfaces--also called grain boundaries. Metals are composed of crystals, all of which may have a slightly different orientation with relation to the other crystals. This difference in orientation results in a defect, which greatly influences the strength or weakness of the metal as well as speed with which impurities can diffuse through the metal. ``We've been carrying out the first computer simulation of diffusion in these grain boundaries,'' says Adams. He points out with matter-of-fact pride that his group is one of the world leaders in understanding surface diffusion on metals.

Adams' work with metallic glasses and metallic interfaces is novel because he combines quantum mechanical first principles calculations with experimental data on bond energies, bond lengths, and the response of atoms to applied forces. ``We're the first group to try this method,'' says Adams, ``and it's working surprisingly well. We've found that this method works much better than other empirical models, especially in modeling alloys.''

UIUC student benefits

Noting that Adams' work includes a lot of ``firsts,'' he acknowledges that his success is a mixture of having access to NCSA's computational power and his use of novel models for carrying out the simulations.

UIUC students are among the beneficiaries of Adams' work on atomic-level computer simulation. He is teaching a graduate-level engineering course this semester on atomistic computer simulation, in which students use the codes developed on NCSA's RS/6000s to perform their own computer simulations on UIUC's campus- sponsored RS/6000s.

__________________________________________________________________

access / Spring 1993 / NCSA / pubs@ncsa.uiuc.edu