Towards the Computational Design of Materials Using First Principles Methods
Wednesday, October 31, 2012
9:00 am - 10:30 am
Fitzpatrick Center Schiciano Auditorium Side B
O. Anatole von Lilienfeld
Assistant Computational Scientist Argonne National Laboratory It is a timely goal in the physical and engineering sciences to use computers to design novel materials that exhibit sets of specific desired properties, and that are straightforward to synthesize. Some of the most important materials properties depend explicitly on atomistic details, rendering an atomistic resolution of any employed simulation model mandatory. For many of the more complex systems and properties, however, brute force screening of all possible compounds is beyond any capacity due to the combinatorial nature of materials compound space. Consequently, smart design algorithms must not only negotiate model accuracy and computational speed but must also aim for rapid convergence with respect to number of compounds sampled. I will first give a brief description of recent method developments based on variational (deductive) and correlational (inductive) approaches. Thereafter, I will discuss applications that illustrate the versatility of a first principles approach to materials design. Specifically, I will show how to gain control of (i) interaction and adsorption energies relevant to catalyst design, (ii) eutectics of heat-transfer-fluids for concentrated solar power facilities, and (iii) molecular electronic properties important for photovoltaics. Concluding remarks will address current and future research efforts.