Why to take Normal to a surface...
 

Hi,
I have written a Finite volume natural convection code for a square cavity using a structured cartesian grid using "Versteeg Malalasekara" as reference book. Now i am trying to do the same using an unstructured grid using "Blazek" as reference book. Blazek explained to take normal vector for each face. Why to do this. I never encountered this in "Versteeg". Can any one plz explain me this?..

You need the normal vector to compute the fluxes through the faces - that's how the fluxes are defined in the governing equations. If your previous experience is in structured Cartesian grids, the normal was hidden by the simplicity of the geometry.

The flux is a vector. For the entire quantity that goes through that surface, the normal component is needed.

Absolutely right. And one more thing. When you are using structured mesh, the connecting line of the geometry centers of two neighbor cells are orthogonal to the face shared by these two cells. This fact leads to ZERO cross-diffusion term anywhere in discretized momentum euqations. But when you are using unstructured mesh, the connecting line of the geometry centers of two neighbor cells are not necessarily orthogonal to the face shared by these two cells, which pretty much means the cross-diffusion term is not ZERO. So you have to calculate the cross-diffusion term, and you will need the normal vector of the face in the calculation.

Thanks to all for explanations. I clearly understood the meaning from jntan's explanation. I have to calculate (interpolate) the flux at a face by using the flow variables stored at two cells which share that same face and then multiple it by normal vector to get the normal component. is that rite?...