![\"\"](\"http:\/\/hu2.me.gatech.edu\/Ko\/wp-content\/uploads\/2021\/03\/Screenshot-2021-03-25-183738.png\")
<\/a><\/p>\nDouble-diffusive convection<\/strong>\u00a0is a branch in the\u00a0<\/span>buoyancy-driven flow<\/strong> where there are two scalar fields(usually temperature and salt concentration) \u00a0that contribute to the density. The fact that the two scalars have different diffusivities and affect\u00a0density differently brings instability and in some cases,\u00a0<\/span>turbulence<\/strong>.<\/span><\/p>\n It is ubiquitous in nature as well as in engineering applications. For example, the non-uniformity of salinity and temperature distribution gives rise to the interesting\u00a0salt finger phenomenon and crystal growth.<\/strong>\u00a0Engineers designed\u00a0the Solar pond<\/strong>\u00a0to exploits the adverse temperature and concentration gradient layer in double-diffusive convection to prevent heat from going out while collecting radiation energy.<\/p>\n The numerical method we used to study double-diffusive convection is fairly complicated. This is also the major part of the project. To partially avoid the difficulties brought by the pressure\u00a0coupling in incompressible flow, we used the\u00a0vorticity-velocity formulation<\/strong>\u00a0where vorticity is treated as our primary variables to\u00a0integrate. We use\u00a0a high-order compact (HOC) scheme<\/strong>\u00a0to discretize the equations. The program\u00a0is written in\u00a0Fortran<\/strong>\u00a0and\u00a0parallel computing technique<\/strong>\u00a0was adopted as well to save computational time.<\/p>\n We used an analytical solution to test the\u00a0resolution<\/strong>, to ensure that we achieve\u00a04th-order accuracy<\/strong>, and to compute our\u00a0parallel efficiency<\/strong>. The program was also tested quantitatively and proved efficient and accurate, in simulating the benchmark cases such as\u00a03D driven-lid problem<\/strong>\u00a0and\u00a0natural convection<\/strong>.<\/p>\n Collaborators: Zhenfu Tian and Peixiang Yu<\/p>\n