It should be noted that the methodology implemented in Ansys Polyflow is different from a frequently used "two-fluid" approach, in which the region that does not contain the liquid being tracked is modeled as a low-viscosity fluid (e.g., "air") whose flow nevertheless needs to be computed. In Ansys Polyflow, empty regions are simply excluded from the calculation. This has major advantages:
The calculation time and required memory are significantly smaller at the beginning of the simulation when many nodes are still dry (and hence require no computations). Overall, this proves to be a significant saving.
When the air viscosity is set to an extremely low value in a two-fluid approach, it can yield a Reynolds number that is high or even very high, such that it is by far more difficult (at least with velocity/pressure techniques in use in Ansys Polyflow) to compute the flow of the otherwise absent material than the polymer. If you instead select a higher air viscosity, the flow of the air might interfere with the flow of the polymer by slowing it down or creating unrealistic "bubbles".
Another numerical consideration for the Ansys Polyflow VOF model concerns the interpolation method. It is common for the elements to be partially filled, as a result of some of their nodes being wet and others dry (remember that the control volume around a node extends into neighboring elements). Consequently, it is not possible to use an incompressible velocity-pressure interpolation (for example, the mini-element method) for the VOF model. Because it is necessary to accept any wet/dry node configuration within an element and a varying ratio between active velocity and pressure nodes, the LBB condition (see Controlling the Interpolation) cannot be satisfied. This is the reason why VOF simulations can only be performed using the stabilized linear velocity/constant pressure interpolation. The value of the stabilization coefficient is adjusted automatically.