That depends on your definition of life. If you include any self-replicating automata, you probably do not need an atmosphere at all.
So the question should rather be: Which atmospheres do not support the spontaneous development of life. I would put my bet on anything that is either:
- too dense, i.e. of such a high pressure that no complex molecular structure can surive. In the very extreme, I do not expect life in the "atmosphere" of a neutron star.
- too empty, i.e. distances between atmons are so large that a merging into more complex molecules is extremely unlikely (think of the moon)
- too monomorphic, i.e. so pure that there simply are not enough different kind of atoms to form complex molecules. I have no example for that, though.
- too cold, i.e. there is simply not enough energy to create complex molecules from the existing matter
- too hot, i.e. there is so much energy that any complex molecule is instantaneously destroyed (e.g. by radiation) again
All these seem to be rather extreme conditions, but in any other case there is at least the possibility of some form of life.