General aspects of coupling CFD and MD

The ability to influence macroscopic flow behaviour by manipulating interfaces holds promise for significant impact in future engineering technologies. For example, surface features of micron-size resolution and hydrodynamic coatings are sought for turbulent drag reduction, in much the same way as secretions of tono-filaments and lipid droplets reduce drag on a dolphin skin; morphing surface textures are envisaged for ``fly- by-feel'' technologies that mimic the ability of insects to optimise lift; and use of electric fields is explored for manipulating the interfacial behaviour of droplets in electro-sprays.

The key to achieving these challenging goals is to recognise that the dependency between the micro- and the macro-scale is bi-lateral: while the microscopic interface region can significantly alter the macroscopic flow, the outer state co-equally affects the microscale dynamics, thus establishing a feedback loop. A faithful representation of the dynamics must therefore span 7-8 orders of magnitude in spatial extent and 9-10 orders of magnitude in time. As a result, this level of multi-scale modelling has remained beyond the most advanced engineering simulation capabilities.

In this dCSE project we have developed the software infrastructure for coupled continuum-molecular systems. The Molecular Dynamics (MD) simulations track the motion of individual discrete atoms based on their mutual interactions inside a confined domain. On the other hand, the continuum software uses Direct Numerical Simulations (DNS) to solve the non-linear, partial differential equations for the fluid velocity at all points within a large computational domain.

Coupling between discrete and continuous simulations is a very active research field, dating back to a seminal paper by O'Connell and Thompson [1]. There have been a number of key papers over the last eighteen years, including, [2,3,4,5,6,7,8,9]. As the number of researchers utilising coupled techniques continues to grow, so does the need for an efficient, scalable and robust ways to implement the coupled methodology.

To this end, the coupling of an MD algorithm ( $ \mathcal{S}tream$$ MD$) and a DNS solver ( $ \mathcal{T}rans$ $ \mathcal{F}low$) within a single, coupled framework has been achieved in the first phase of the dCSE proposal. This will form the basis of a multi- purpose computational coupling utility which will be made available to the HECToR community.

Lucian Anton 2012-05-31