SIEMENS

Today’s production materials are characterized by high performance. But as researchers seek to further improve them, the parameters that influence their performance become so varied that it becomes impossible to comprehend and experimentally detect their numbers and complex interrelationships," says Dr. Wolfgang Rossner in describing the challenges materials researchers face. Rossner works at Siemens Corporate Technology (CT), where he supervises the development of ceramic materials, which are used in everything from light emitting diodes to detectors in medical systems and coatings for gas turbines.
So where do you begin when optimizing materials? Should you start with their structures, chemical compositions, or the processes used to manufacture them? This immediately raises other questions. If, for example, you want to give a porous substance a different structure, you can change the size, shape, or distribution of its pores; but which measure will be most effective?
It’s as if researchers were standing at a control panel full of switches whose effects they were only vaguely familiar with. What’s more, changing one parameter may cause changes in others in unpredictable ways. Producing and testing every conceivable variant in a laboratory takes too much time and costs too much money.
But there is one tool in particular that can help to accelerate this process significantly: computer modeling (see Pictures of the Future, Spring 2006, Simulation). By developing a material in the virtual world, scientists can analyze the effects individual parameters will have on its behavior. They can thus identify relationships between material properties, specific microstructures, and chemical compositions, and then use this information to induce a desired behavior. In this way, they know which changes are likely to be successful even before lab testing begins. Virtual materials are also perfectly defined — free of the unknown side effects that occur in the lab, which can easily mask the presence of the effect being pursued. Models of virtual materials work under precisely defined conditions, so if a material alteration does not produce the desired effect in a simulation, you can assume with near certainty that it won’t work in the real world.
Researchers thus use models to determine the physical limitations of a material, while at the same time overcoming the physical limitations of an experiment by means of the virtual materials the model produces. Scientists can, for example, use a computer to simulate loads — such as those occurring under extremely high temperatures that cannot be recreated in a lab. The model allows them to assess which mechanisms determine the lifespan of a specific material, and thus to select the components it should be used in. "You don’t want to find out how a material will behave in an application after the component has been produced; you want to get that information while its still being developed," Rossner explains.
This is an important issue — and not just for Siemens, as new or improved materials provide competitive advantages and serve as engines of innovation. Germany’s Ministry of Education and Research therefore funds virtual material development to the tune of €20 million annually, and two of its current 14 joint projects are being coordinated by Siemens CT.

Analyses at the Atomic Level. Researchers want to precisely understand how the structure and composition of a material will affect its behavior in each application, since that’s the only way to find out how they can give a material the desired properties, or ensure that a component will not fail. Their work takes them deep inside a material. Around four years ago, for example, Rossner’s colleague Dr. Stefan Lampenscherf began experimenting with multiscale modeling, which combines various simulation methods to enable scientists to completely describe a material all the way from its atomic structure to its behavior in a component. The models address the most diverse dimensions: Microstructures such as pores, grains, and cracks typically measure only fractions of a millimeter, for example. These structures can be accurately described using the Finite Element Method (FEM), a procedure that subdivides a virtual material down to the smallest building block — all the way to the size of individual grains that typically measure anywhere from a few tenths to several hundredths of a millimeter.
To examine a material’s chemical composition, on the other hand, researchers require methods that can describe interactions on an even smaller scale, namely those involving atoms. Here, the most precise technique is provided by the density functional theory (DFT), which describes the quantum mechanical interactions between atoms. Extremely high computational requirements, however, which can amount to several days, severely limit the use of this method. As a result, DTF methods can address only around 1,000 atoms.
In contrast, simulating a structure measuring only one-thousandth of a millimeter at the atomic level necessitates calculating the interactions of billions of atoms. This is beyond the scope of available resources, of course. But one possibility for solving this problem is to construct less detailed molecular-dynamic models that describe in a simplified manner the movements of atoms in a force field known as the interatomic potential, whereby the quantum mechanical processes in the material are contained in the definition of this potential.
The skill of the materials scientists on Rossner’s team lies in their ability to select the right model for any question they pose. They do this by extensively analyzing materials in the lab and then continuously comparing their results with model calculations. Rossner believes this approach forms the basis for successful materials development. "The close interplay between our experiments and models is what makes our team so exceptional," he says.
Another major challenge lies in linking models — i.e. using the atomic model to derive properties for the elements of the FEM model. That’s one reason why the team is constantly in touch with universities and research institutes around the world, including the universities in Cambridge and Oxford in the UK and the Max-Planck-Institute for Iron Research in Düsseldorf, Germany.
Many atomic model experts work at these institutes, and their deep understanding of methodologies and quantum mechanical processes within materials allows them to help develop effective material models from the atomic to the macroscopic scale. Rossner believes that such models will become much more practical in the future. And, just as FEM is well established at many development departments today, it will become possible over the long run for materials experts to utilize atomic models without having a profound understanding of quantum mechanics.

Efficiency through Simulations. Rossner and colleague Dr. Philip Howell point out that heat insulation coatings for gas turbine blades provide a good example of how FEM and microstructure analysis affect materials development. These ceramic coatings, which are only a few tenths of a millimeter thick, protect turbine blades from temperatures in excess of 1,300 °C and can withstand temperature differences of several hundred degrees.
Ramesh Subramanian, an engineer at Siemens’ Fossil Power Generation Division, explains the benefits of using improved ceramic coatings: "A heat insulation coating that withstands higher temperatures and displays lower heat conductivity reduces the cost of cooling the turbine. For instance, a temperature increase of 100 °C can boost energy efficiency by around one percent, which can save a turbine operator about one to two million euros in fuel costs per year."
In their search for ceramic materials with better heat insulation, scientists initially focus on the influence of microstructures. Using computer tomography (CT), they measure a ceramic section and transfer its microstructure as precisely as possible to an FEM model. This virtual sample is compared with its real-life counterpart to determine whether its simulated heat conductivity corresponds to lab values, thereby confirming the validity of the model. Only at that point do researchers begin altering the structure and composition of the virtual material.
A case in point is zirconium oxide, a high-performance heat insulation ceramic material. Researchers want to test whether adding other atoms can reduce the material’s heat conductivity. To estimate the effect of this measure, Howell is developing atomic models for such ceramics with other researchers — for example, with colleagues at the Fritz Haber Institute in Berlin. He says the main challenge lies in making adjustments in line with pertinent questions. "The molecular-dynamic model we want to use to describe heat conductivity requires a suitable interatomic potential, in this case for zirconium oxide. Scientists engaged in pure research are mostly interested in basic questions, however — for example, whether a certain model is capable of describing the desired property at the atomic level, in our case heat conductivity. All that’s required for that is evidence gained with a model material for which a potential exists. We in industry, however, have to develop components for use in products, and we need to make definitive statements regarding a specific material. Potentials for specific substances therefore need to be developed." Howell is currently studying whether his molecular-dynamic approach is suitable for generating useful trend statements regarding the heat conductivity properties of zirconium oxide ceramics.

Physical Limits. Howell has already attained his goal in another area. After an Osram rival announced it had achieved a dramatic increase in lamp efficiency, inquiries revealed that a patent had been registered describing a metallic photonic crystal that would emit a higher proportion of visible radiation than the coiled filaments used in today’s lamps. Such filaments primarily emit heat. Photonic crystals behave like very precise light filters because their structure enables them to emit or absorb only light of a particular wavelength. Instead of trying to produce such a crystal in the lab, which would be a very complicated process, Osram engineers turned to Siemens CT to find out if the structure described in the patent would actually deliver the result claimed.
Rossner’s team initially attempted to create the described structure in the lab. But as was expected, even the first few steps proved to be extremely difficult to carry out. However, there was a chance that the rival had already come up with an effective manufacturing process. Howell therefore began working in parallel with experts whose speciality was the computer modeling of photonic crystals. And the specialists created perfect virtual structures that no lab had previously been able to produce. Then, because the simulation did not lead to the behavior described in the patent, the researchers became convinced that the claim regarding the revolutionary light source ultimately would not hold up. Their suspicion was later confirmed when the patent application was withdrawn.
Dr. Klaus Orth, Director of Thermal Sources Development at Osram Consumer Lighting, was satisfied with the outcome. "The idea of using metallic photonic crystals isn’t new, but we were certain they would not be able to withstand temperatures of up to 2,700 °C over a lamp’s entire life," he says. "The simulations also showed that the structures described in the patent could not be successful. Without positive simulation results, we wouldn’t have started to conduct that type of complex work in the lab."