Pictures of the Future    Spring 2003

High-performance computers and robots are accelerating the search for new industrial materials. Researchers can create any combination of chemical elements in the lab and quickly analyze their properties. In addition, simulations can be used to demonstrate the behavior of materials from the atomic level all the way to finished components.

The search for new materials is usually limited to modifying and improving existing substances. If, however, a particularly industrious researcher wanted to test as many combinations of elements as possible to create a new catalyst, he or she would face a Herculean task. After all, elements from the entire periodic table could theoretically be selected.

But thanks to combinatorial chemistry—a technology that allows researchers to generate extensive libraries of chemical compounds quickly—there is now a way out of this dilemma. The process involves pipette robots or other machines that mix predefined amounts of basic materials into tiny depressions on solid substrates, thereby creating up to ten thousand different combinations. State-of-the-art high-throughput screening methods are used to test these combinations for specific properties, while computers and mathematical processes help evaluate the results. Although the origins of combinatorial chemistry are to be found in medical research, material researchers are increasingly using this process as well, particularly when searching for new catalysts for chemical reactions.

"For the first time ever, we now have a tool that supports the process of making discoveries," says catalyst researcher Wilhelm Maier, a professor at the University of Saarbrücken, Germany. Maier is particularly interested in developing efficient strategies for discovering and optimizing materials. In a single experiment, he can test and characterize more than 200 materials, including very exotic ones that would not normally be analyzed. Using an optimized process, U.S.A.-based Symyx Technologies is already creating 10,000 combinations a month. And when it comes to checking initial "hits" under real-life conditions, some 3,000 new tests are conducted each month. Thus, combinatorial chemistry dramatically helps to accelerate the search for new materials. In fact, according to Maier, it makes the search process move ten to 100 times faster compared with conventional methods.

Dividing and mixing. The company hte AG (hte stands for "high throughput experimentation") specializes in catalyst research. "We create metal oxide mixtures, which we then test for their suitability as catalysts for chemical reactions," explains Dirk Demuth, managing director of the Heidelberg-based company. One of the methods used by the company's scientists in their research is the so-called split-and-pool technique. This involves dividing a group of 5,000 aluminum oxide pellets, each of which has a diameter of 1 mm, into five more or less equal portions. The scientists rinse each portion in a different metal salt solution—for example cobalt, molybdenum, iron, manganese or vanadium. They then bring all of the pellets back together and divide and rinse them again. This process is repeated five times.

Once the solutions have transferred various mixtures of metals to the oxide pellets, the pellets are distributed on plates containing 384 holes and checked to see if they exhibit any catalytic activity. The researchers only analyze the metallic composition of "promising" pellets. "In another project, we discovered a catalyst for a particular reaction," says Demuth. "Like other catalysts, it had a yield of over 90 %. However, it was up to ten times more durable and far less expensive in terms of the material used."

In addition to conducting research commissioned by customers such as BASF, Degussa, and ChevronTexaco, hte develops catalysts for hydrogen carriers and extracting nitrogen oxides from automobile emissions. As such, it is creating an important collection of data that can be exploited in future simulation-based materials research. Among other things, scientists at Siemens Corporate Technology (CT) are conducting research on ferromagnetic materials, new dielectrics for extremely high frequency transceiver modules in cell phones, piezoelectric materials and LED phosphors. For example, the researchers apply mixtures of the oxides of elements such as strontium, barium, titanium and niobium to a substrate in specific ratios, creating over 100 mixtures. The mixtures are then analyzed by an automated system. Working closely with Symyx, the Siemens researchers have tested about 150,000 combinations over the last two years.

"In the past, this would probably have taken two decades," says Dr. Wolfgang Rossner. "We discovered complex combinations of phosphors that generate a white light when applied to a blue LED as a thin film. That wouldn't have been possible without combinatorial chemistry." The colors produced by the new phosphors in LEDs are more "natural" than those generated by other products. The phosphors are therefore being prepared for market launch.

Unlimited Parameters. For Rossner, the combinatorial method is much more than just a way of combining materials as desired. "The fact that we can now analyze the technical parameters in much greater detail is probably even more important," he says. That's because the properties of a phosphor are not only dependent on the material's chemical composition, but also on the way it has been processed. It is crucial to know, for instance, during which production stages the material was heated. Another important factor influencing the outcome is the surrounding atmosphere. For example, oxygen has an oxidizing effect, which creates a red form of a phosphor from the metal europium. On the other hand, the reducing effect of hydrogen creates a blue form. "There are an almost unlimited number of influencing parameters and resulting combinations," says Rossner, "and combinatorial chemistry helps us to discover better materials more quickly and cheaply."

Dr. Randolf Mock has decided to take another path to create better materials. Mock is a computer simulation specialist in the Sensor and Actuator System Group at Siemens Corporate Technology in Munich. His work has made it possible to develop a production version of a piezoactuator for diesel direct injection (see article Intelligent Materials). The cylindrical, piezoceramic component, which controls the injection of fuel into the combustion chamber by means of voltage pulses, is supported by a Bourdon spring. Stressed with a force of 85 kg, the spring prevents the buildup of tensile forces on the sensitive ceramics, thereby protecting the component. "No manufacturer worldwide was able to offer springs that would have lasted longer than a hundred hours," says Mock. This was far too short for a crucial component meant to last the life of an automobile.

"To remedy this situation, we developed our own spring on a computer in just six weeks," says Mock. The team's solution was to use the metal case surrounding the component as the spring itself. "We put slits into the virtual tube and even tried out some very bizarre shapes," explains Mock. To conduct the simulation, Mock used the finite element method. "It's like using a bunch of Lego blocks to create the component on a computer. The blocks are kept as small as possible so that we don't need an excessive amount of computing," he says. In total, the virtual spring consists of 10,000 individual elements. A computer program calculates how strong the forces are within the spring when it is subjected to various loads. The service life of the spring can then be estimated pretty accurately on the basis of the mechanical stress. "In this case the results were so accurate that the real-life values only deviated from them by 1 %," says Mock. Because the product's production engineers required very specific properties, Mock only improved the shape of the spring, not the material it was made of.

Following a different approach, a group of Siemens researchers headed by Dr. Wolfgang von Gentzkow in Erlangen, Germany used computer simulations to improve a composite material. The group's goal was to create iron that could be processed like plastic, but that would have the lowest possible plastic content. To this end, they simulated iron particles of various sizes as well as the particles' distribution in plastic. In addition, the researchers calculated the strength of the adhesive bonds between the plastic molecules and the surface of the particles so that they could optimize the material's mechanical stability. A great advantage of these simulations is that they allowed the researchers quickly to predict the results of changes to the materials without having to analyze a single particle in the lab.

Fast Cracks. Researchers specializing in development of computer-assisted materials hope to be able to fully simulate materials and components all the way down to their atomic structures and their quantum mechanical effects. A pioneer in this area is Prof. Huajian Gao, director at the Max Planck Institute of Metals Research in Stuttgart. His team is studying how cracks and crystalline dislocations originate and spread. Although these phenomena are of crucial importance for understanding materials, the physics behind them still cannot be fully explained. Cracks spread at an incredible rate of several kilometers per second. In order to make this process visible, Gao's team simulated a one-micrometer cube that contained around a billion atoms and had the ideal lattice structure of a metal. A crack was formed in the cube by virtually removing several atoms. The researchers then generated tensile forces that simulated strain-hardening—a process that causes flexible metals such as copper to become as brittle and fragile as glass.

Monster Calculations. Materials research can be extraordinarily data intensive. Consider a program that calculates the position of each atom in a material over a period of several billionths of a second. In fact, it took ten days for the ASCI-White IBM computer in California to make such a calculation, even though it is one of the fastest super computers in the world and is capable of handling up to ten teraflops (ten trillion calculations per second). While the computer simulated the spread of cracks, it was noticed that secondary cracks, which spread even more quickly than the original crack, begin to form at the ends. In practice, this finding could be used to develop materials that steer fatigue cracks into non-critical areas of a component.

It still takes an extremely long time to make such calculations. Nevertheless, methods that involve the simulation of materials and components on computers are becoming increasingly important as materials research in the nanotechnology and biotechnology sectors becomes more and more expensive. What's more, researchers expect growing computung power to result in a new era in materials development. They will then be able to not only forecast a material's exact mechanical properties, but also its optical, magnetic and electrical ones. In doing so, they will be using simulations that smoothly combine quantum mechanical, atomistic and finite element methods. But according to Gao, it will take another five to ten years before this kind of multiscale modeling finds its way into industry. It is almost certain, however, that super computer atomism will eventually play a key role in the development of bio-nanotechnologies. "It could become one of the most important engineering tools of the 21st century," says Gao.
                                                                                              Norbert Aschenbrenner