@User We don't see the symmetry plane in your image, but does your gas enter through the injector, create a small downward bump of penetration and then thin out along the top? It looks like your air is sticking to the top because of buoyancy and also because it is VoF, so it will avoid dispersion into bubbles unless you have a crazy resolution. It doesn't help that there is symmetry condition which will further remove chances of asymmetric stuff happening. The straight edges of the way air boundary is spreading along the top looks purely artificial. Basically, plenty of scope that this simulation will be away from reality due to the way this domain and interface method is setup. In turn, that means scope for artificial spikes, instabilities appearing at critical points of your sim. I say this because I don't see anything catastrophic in the two dicts you shared... Did you try eulerian model with dispersed model for gas..?

Hi, yes that is correct, air enters in the injector. The symmetryplane is essentially at +Z in the alpha.water image. I understand your concerns, but the way the air spreads matches very well with experimental data that i am trying to match. Both in terms of how thick the "hump" is, the length of the air layer, and the width of the "strand" where air escapes at the -Z wall. Regarding dispersing into bubbles, that is not expected at this air flow rate, according to the experiments, but is something that should occur with decreasing flow rate (or increasing water velocity).

Okay, so it seems that this is how things should be, more or less... so then, the major concern is instability when the air film reaches the outlet.. If I were you I would try to describe the 'instability' in some detail, the affected variables, which region shows problematic features in variables., or which change in feature correlates to this onset... etc

Sure, so several things. I notice that the air phase deforms unphysically, i.e. large chunks of the air just disappears, the convergence of the p_rgh residual decreases significantly (from ~10 iterations to ~100), the timestep decreases by a factor of 10-100 (large, seemingly random, velocity spikes close to the outlet). My suspicion is that it is related to how backflow is treated, but that is just a hunch.

yes sounds like a fair assessment about the backflow.. I assume it is zeroGradient on alphas at the oulet?

correct, i tried inletOutlet with both alpha.water 0 and 1 as the backflow

Does the spiking cells, or pressure values close to symmetry/interface cells/bottom patches show any pattern (you'll have to save some rapid timesteps to gauge this...)...?

Good question, i will have to look into that. I hoped that i had just missed something obvious regarding the boundary conditions... 😅

On that note, I am not familiar with the fixedMean condition .. maybe it allows for oscillatory stuff as long as the "mean" is okay... not sure... same with the far-field setup you have at the bottom - zeroGradient on velocity + flux zero for pressure - seems awfully like a wall... hence, my point about 'where' do the instabilities emerge most