Ulrich Aschauer
Despite many years of intensive research there still remain many unresolved questions in powder and ceramic technology. A majority of these issues are linked to interfacial phenomena of atomic scale origin, which makes their experimental investigation very difficult due to limitations in the spatial resolution of the available analysis techniques. Computer simulations at the atomic scale provide with the advent of more and more advanced methods and increasing computer power an ever more powerful tool for the investigation of these phenomena. The understanding of experimental systems gained at a fundamental theoretical level will help target key experiments in a knowledge based fashion for the rapid advance in top-performance materials development, saving time and resources with simulations showing the most promising routes and key parameters to be explored in experiments. The present thesis investigates several key steps in the production cycle of a ceramic material using atomic scale computer simulation techniques, leading to advances in the understanding of the underlying atomic scale phenomena in each case. Instead of treating a single material throughout the ceramic production process, a series of different systems of experimental interest are looked at in order to show the generic nature of the approach. Very often ceramic powders are produced by precipitation from solution, where many powders properties can be modified, amongst them the particle size, morphology and state of agglomeration, all of which have an important effect on the quality of the final ceramic. Growth can be modified by the reactive environment (temperature, pH) or extrinsic species such as ions or molecular additives. Growth modification of hematite (α-Fe2O3) by phosphonic acid molecules was simulated by energy minimization and experimentally validated, making the method useful as a predictive tool for other phosphonic acid molecules. Significant changes in morphology were predicted and experimentally observed, the main reason for preferential adsorption being the surface geometry and topology, major distortions of the molecule leading to unfavorably high adsorption energies on some faces. Since experiments showed calcite (CaCO3) particles grown in presence of polyaspartic acid (p-ASP) to have a higher specific surface area than with polyacrylic acid (PAA) despite the same functional groups on both additives, molecular dynamics simulations were used to investigate this difference. The presence of charged surface defects at steps was found to be the key requirement for adsorption, the molecules' approach being hindered by the highly coordinated water at the surface and the additive even desorbing without the attractive electrostatic force. PAA was found to form more complexes with counterions in solution and a more negative enthalpy of adsorption was found for p-ASP, both of which will lead to more marked growth suppression by binding of p-ASP to steps, which are expected to be the main growth sites. Further the adsorption conformation of p-ASP results in a better colloidal stabilization than PAA, preventing particle agglomeration. These three aspects can explain the higher specific surface area for powders precipitated in presence of p-ASP. These computational approaches can thus explain subtle differences between additives as well the same additive on different surfaces helping in the targeted use of additives and the design of new additives. Dopants may be added to ceramics during powder synthesis or before sintering either as sintering agents or to add specific properties to the final material. Often dopant ions are oversized compared those replaced in the crystal lattice leading to segregation towards defects where their incorporation is facilitated. Interface segregation was studied for industrially relevant dopant ions in alumina and zinc oxide using energy minimization and microstructural models to help compare with experimental findings. The results show a strong tendency for surface and a slightly lesser tendency for grain boundary segregation in all cases, however with important differences between different interfaces. As a key result one possible reason for the microstructural homogenization effect of magnesium was found to be its capacity to narrow the grain boundary energy distribution resulting in more equiaxed grains due to isotropic grain growth. The segregation of luminescent neodymium ions in YAG ceramics for laser applications was investigated and dopants were shown to accumulate at grain boundaries where their high concentration will make them inactive for lasing and result in light scattering due to slight local variations of the refractive index. Using microstructural models the proportion of inactive dopant ions was found to increase with decreasing grain size, making nanoceramic lasers less powerful. Optical models based on atomistic segregation results on the other hand show that light scattering is more severe for larger grains, nanoceramics thus being more transparent. There is therefore an optimal grain size where laser power and transparency will result in the best possible laser performance. Different microstructural models could be linked to the sintering procedure, predicting that the best laser performance should be obtained by slow conventional sintering rather than by novel rapid methods such as spark-plasma-sintering (SPS). These results are key in understanding the role of interfaces in this novel class of laser materials and will help improve their fabrication methods. Experiment and simulation of oxygen self-diffusion in alumina so far showed large discrepancies, the simulations predicting diffusive jump activation energies by a factor five smaller than experiments. In this thesis a novel simulation approach based on Metadynamics allowed to determine activation free energies of individual jumps, then using these in kinetic Monte Carlo simulations to predict the macroscopic diffusion coefficient. The key result is that the activation energy for the diffusion dominating jumps is of the order of the experimentally determined one, the so far reported simulated jump energies not forming a continuous diffusion network. The calculated diffusion coefficient correlates relatively well with experimental results. The developed approach is generic and can be applied to investigate a variety of diffusive phenomena in solids providing answers to many more of the remaining questions. While atomistic simulation approaches give very interesting results on local atomistic phenomena and allow understanding and elucidating which processes occur spontaneously in a system, for predictions on experimental length and time scales the use of larger scale models is immensely important to extrapolate atomistic results to these scales. The combination of techniques from the electronic structure through the atomistic and mesoscale to the continuum level in a multi-scale modeling approach seems very promising for material science applications, where the goal is to understand experimental observations and to target the next key experiment in a knowledge-based fashion.
EPFL2008
