Do you mean: polygonal booleans !?because I substracted the whole ice cubes
I see the logic with inverting the bubbles normals but that's telling wrong things to Indigo if they have an air material. The normal direction is important to determine wether a ray is entering or exiting the medium. In that case, the ice volume would be defined as ice from outside light, and as air for light exiting the bubble.
It's either:
a) the bubbles have inverted normals AND the ice material onto them. It's fine in order to hollow a medium (no bools !!!), but in your case they are likely to be filled with hibiscus or water wich has an higher precedence than Indigo's own air medium (always and by default).
b) as Zom-B says, as dakiru did I bet. Create a flat specular with ior 1.0 and name it air or CO2 or whatever. Assign it to the bubbles, but then keep normals untouched to signify that the light is entering your air definition (thus exiting ice).
I hope it helps clarifying from the confusion. Sometimes I'm confusing by giving to much details, just pick what fits for you
