I found this question here but it does not fully answer my question. The answer there was that "composite bosons can occupy the same state when the state is spatially delocalized on a scale larger than the scale of the wavefunction of the fermions inside".
Let's say we do a BEC with bosonic atoms (for example in a harmonic trap). The BEC means that a huge number of atoms will occupy the same energy level. This cannot be exactly true because the atoms are made out of fermions. So I guess that "the" energy level is actually a collection of many different energy levels that originate somehow from the internal structure of the atoms. This effectively creates a degeneracy of "the" energy level. I think this is what he meant by "spatially delocalized on a larger scale than the scale of the wavefunction of the fermions inside".
I have a few questions regarding this:
Is this correct?
Where does these extra energy levels come from (There must be a huge amount of them)?
If there is a huge amount of internal energy states it should give a great enhancement of the density of states. Since many thermodynamic quantities depend on the density of states (for instance the particle number) this should change the thermodynamics of a gas (not only at small temperatures but also at higher ones)?
EDIT: This edit is about Chiral Anomaly's answer. I would like to do this a bit more quantitatively. Consider a sodium atom. Its Hamiltonian (like for the H-atom) can be composed in a rest frame part (which will become the spatial wavefunction of the atom later) and a internal part.
The internal part has a hydrogen-like spectrum. The quantum numbers of these states are what you called $n$. If the electrons have $k$ accessible states then there are $k$ over 11 possibilities to arrange the 11 electrons. For 20 Million atoms (as in here) you need about 34 internal states (This are all states up to $n \leq 4$). For Rubidium you need all states up to $n \leq 5$.
I'm not fully convinced of your argument because of several reasons:
This would imply that all of the atoms in a BEC are excited.
You need a specific electronic configuration for cooling and (even more important) trapping the atoms (i.e. you need one electron in a specific state). So all of those excited configurations where this state is not occupied would simply fall out of the trap.
One observes the BEC by shining light with some transition frequency on them. If all internal states are occupied there cannot be a transition.
EDIT 2:
Let's assume for a moment an idealized world. The nucleus and the electrons create a atom where the wavefunction splits into an internal part $\psi_i$ (with $k$ discrete states) and an external wavefunction $\psi(x)$. We put those atoms in a harmonic potential. Now assume that the internal structure is not affected by the potential and that there is no residual interaction between the atoms. So we can write the total Hamiltonian as $H = H_{ext} + H_{in}$ where $H_{ext} = p^2/2m + V(x) = \hbar \omega (n+\frac{1}{2})$ and $H_{in}$ is just the (independent) internal Hamiltonian.
Let's choose the groundstate of the harmonic trap to create a BEC. If the atoms were fundamental bosons this degeneracy of this energy level is 1 (which is no problem here). But now we have composite bosons so for the fermions this state has a degeneracy of $1 \times k$. So we can put at most $k$ atoms into this state. (I think we both agree on this).
Now turn on interactions. There are many different things changing.
The internal structure is affected by the potential (this is fine since it does not change the the number of states).
The atoms interact with each other. This will lift the $k$-fold degeneracy of the ground state (i.e. different atoms will have a different $e^{-iEt}$ time dependence). If the interaction is small the splitting will be small, therefore the time dependence of the atoms will be nearly equal. If we run our experiment only for a small time it will look like all the atoms have the same time dependence (BEC). If the interactions are not neglectable the level splitting will be of order $\hbar \omega$. So it will not look like all atoms occupying the groundstate but rather the two lowest states (no BEC). However now we can put $2k$ atoms into our gas because we are treating two (unperturbed) states as equal. But I doubt that this will solve the problem because as I said there won't be a BEC anymore.
Now comes the complicated part. The internal and external wavefunctions (even of different atoms) can mix. This is hard to analyze. But we know two things: 1. The overall numbers of states does not change. 2. The resulting gas must be able to form a BEC (i.e. you need enough states which have (nearly) the same time dependence). If you just crazy mix some high energy states into low energy states the nice time dependence will get lost. Also in this case all the BEC analysis would be completely wrong (since it does not account for such mixing). So I think this must be neglectable.
All in all when turning on interactions will not create extra states. Therefore if you see a BEC you have at maximum $k$ atoms in it.