On the harmonic interpolation scheme for the diffusivity coefficient
Hi all,
browsing the forum recently I've found in an old post that in our beloved code (OpenFOAM) there is the possibility to use the harmonic interpolation scheme (rather than the classical linear scheme) for the evaluation of the diffusivity coefficient on the cell faces. Well, knowing from Patankar's book that when there could be abrupt changes in the diffusivity coefficient the harmonic interpolation scheme behaves much better than the linear one, I started to think about two questions: 1) why the harmonic option for the laplacian schemes is so rarely cited in this forum? 2) if any of you has some experience about it, does this choice give any significant improvements in turbulent flows calculations (where, in principle, the effective diffusivity coefficients should have significant changes from a point to the neighbour ones)? And if so, does it give any stability issues? Every opinion/contribution about this matter would be really appreciated. Thank you in advance V. 
Mmmmh...so it seems like I'm the only one who has read this passage of Patankar's CFD book? :confused:

Hi Vesselin,
did you have any more information about this topic? Did you get this to run? I'm thinking about using the harmonic scheme as well. Best kathrin 
I use harmonic interpolation mostly because I have sharp changes in the transport coefficient between regions in my model. In my model I was having big problems before I thought of using the harmonic scheme: I made some comparisons here (and in some posts above this one): http://www.cfdonline.com/Forums/ope...tml#post253611
I think most people don't use harmonic interpolation because they don't have sharp changes in the transport coefficient. The best thing to do is to try it to see if it changes your solution. 
Hi ben!
Thank you for your reply! I will definitly give it a try! Kathrin 
Hi Kathrin, I think that Ben's answer can be considered quite exhaustive. This kind of interpolation was indeed introduced in finite volume calculation to account for very sharp changes in the diffusive transport coefficient (think about heat conduction through anisotropic or in general composite materials). Concerning CFD, I can tell you that for HighRe incompressible turbulent flows (RANS/URANS modeling, without heat transfer) I've found no benefits in using harmonic interpolation instead of the linear one (in addition the harmonic scheme seems to be slightly more unstable). Of course, things may be different when heat transfer is involved, so you have definitely to try it for your case!
Regards V. 
All times are GMT 4. The time now is 20:22. 