Development of a Wavelet-based Multiphysics/Multiscale Framework and its Application to Diffusion Problems with Chemically Reactive Boundary

Many energy production and manufacturing technologies that are critical to the Nation’s security involve closely coupled multi-physics processes operating over a wide range of length and time scales. The list of such technologies includes traditional large scale manufacturing processes such as catalytic cracking of petroleum and polymer production, highly specialized materials production such as the generation of nanofibers and self-assembled molecular complexes, and emerging energy sources like chemical looping combustion and fuel cells. In all these cases, atomistic features on solid surfaces interact strongly with mesoscopic and macroscopic heat, mass, and momentum transport to determine overall product quality and energy efficiency. Computer simulations provide a powerful tool for optimizing these processes, but simulations of the multi-scale interactions occurring over such large scales presents a very complex computational problem. The biggest challenge is to synchronize the solution of the governing equations and/or constitutive relationships at the different length and time scales so that the effects of their coupling is accurately captured.

The existing computational tools used in scientific simulation for the above processes are often ad hoc and ineffective when it comes to linking the impact of micro-scale features to macro-scale outcomes. This is due to the fact that the currently available macro-scale models can not resolve the physics at the micro-or atomistic-scale which often is, in fact, the controlling factor. On the other hand, direct utilization of the micro-scale models alone is computationally prohibitive if the objective is to produce predictions for the macro-scale over long time intervals. Some combination of the micro and macro-scale models is needed if one hopes to achieve a physically valid result, but it is also necessary to make this combination carefully in order to fully exploit the power of high-performance parallel computing.


During the last year, Oak Ridge National Laboratory has developed and implemented a novel multiscale/multiphysics computational framework that can consistently transfer relevant information across scales efficiently (both upscaling and downscaling) to simulate coupled process dynamics efficiently. Specifically this framework has been successfully applied to problems with chemically reactive boundary coupled with diffusion process as these constitute building blocks in a wide variety of chemical processes. The underlying multiscale transfer algorithm is referred to as the compound wavelet matrix method (CWM), shown above, and uses wavelets to separate the information from various scales. The various techniques developed as part of this framework enable effective fusion of this information from the different models and thus offer a methodology to efficiently exchange relevant information between various scales. The resulting framework can be implemented readily in parallel form which can utilize thousands of processors on leadership class computing facilities.


For more information, please contact:

Sreekanth Pannala