I recently tested a simple free surface problem in Comet - rise of water between two plates by capillary force and the code gave an error of about 10%. (The plate distance was 1 mm and the contact angle 45º.) Does anyone have any suggestions as to why this may have occured? So far tech support hasn't been able to solve the problem.

On what is the error based with this simple test case? 10% is the error on what? Could you be more explicite?

With a 45º contact angle and a surface tension of 0.07275 N/m water should rise 10.5 mm up between the plates. The best I am getting with Comet is a rise of 9.625 mm and the interface line between the air and water is very diffuse, over 40 cells between 0 and 1 concentrations. (I had to use a cell density of 32 cells/mm to get an answer even this accurate.) Note that using the same conditions with StarCD I get the correct answer of 10.5 mm with a tight air/water interface.

It sounds to me that you are running a free-surface flow without using the HRIC scheme. What kind of discretisation scheme do you use and with which blending factor? Does it happen that you are running a too big a time-step? Remember that STAR-CD does subcycles if the time step is too large. Finally, 32 cells in the gap is a minimum, if the cells are equidistant. You have to capture the high gradients of concentration at the wall.

I tried cd, ud and hric. Blending was kept at 1.0. I am using a time step of 1e-4 sec with 50 inner iterations. If >32 cells per mm are required then Comet is unusable for my real model as it is a complex 3D geometry of about 11,000 mm3.