Thank you for investigating this. It seems to be related to this recent discussion: Radiative heat flux
In Abaqus, I also obtained slightly different results (around 645 K for surface 2) for a mesh with similar density. Maybe it needs some refinement on the inner surfaces, this was also done in COMSOL. I made the mesh quite dense but it’s possible that it still needs some refinement.
As I’ve mentioned in the first post, based on this quote from the documentation:
Sometimes, it is useful to specify that the radiation is closed. This is done by specifying a value of the environment temperature which is negative if expressed on the absolute scale (Kelvin). Then, the viewfactors are scaled to one exactly.
I tried with a sink temperature of -300 K but the analysis didn’t converge.
In Abaqus, I didn’t have to specify the ambient temperature because I selected the closed cavity type. In CalculiX specifying negative temperature for the cavity should be equivalent.