Argonne National Laboratory

Tightly Coupled Atomistic-Continum Simulations of Brain Blood Flow on Petaflop Supercomputers

TitleTightly Coupled Atomistic-Continum Simulations of Brain Blood Flow on Petaflop Supercomputers
Publication TypeJournal Article
Year of Publication2012
AuthorsGrinberg, L, Insley, JA, Fedosov, DA, Morozov, V, Papka, ME, Karniadakis, G
JournalComputer Science Engineering
Date Published10/2012
Other NumbersANL/MCS-P3041-1012

Several advances on the mathematical, computational, and visualization fronts have led to the rst truly multiscale simulation and visualization of a realistic biological system.

Understanding physical phenomena at diverse scales has always been a great challenge. Experimental setups require sophisticated equipment to collect and process data at scales ranging from nanometers to kilometers in space, and from nanoseconds to days or even months in time. In the computational domain, designing mathematical models that can accurately predict physics in a multidimensional and parametrically uncertain environment isn\'t a trivial task. Furthermore, the extremely large number of degrees of freedom in the analyzed system requires the development of robust computational algorithms that are capable of both extracting the main feature and of quantifying the possible errors in the predicted results. These algorithms also must scale to hundreds of thousands of computer processors, and might even promote the development of specialized hardware. Similar to experimental data, computational simulation results must be collected and processed with adequate accuracy. Parallel multiscale visualization tools are required to interactively explore scale interactions.