Consider the theory of scalar QED with the Lagrangian $$\mathcal{L} = - \frac14 F^{\mu\nu} F_{\mu\nu} + (D^\mu \phi)^* (D_\mu \phi) - m^2 \phi^* \phi \tag{1}$$ where $\phi$ is a complex scalar field with mass $m$. Counting the degrees of freedom, we have
- two massless real degrees of freedom from $A_\mu$
- two massive real degrees of freedom from $\phi$
Now, even though there's no symmetry breaking going on here we can still choose to go to unitary gauge, i.e. fixing the gauge so that $\phi$ is real. We now have the gauge-fixed Lagrangian $$\mathcal{L} = - \frac14 F^{\mu\nu} F_{\mu\nu} + (D^\mu \varphi) (D_\mu \varphi) - \frac12 m^2 \varphi^2\tag{2}$$ where $\varphi$ is a canonically normalized real scalar field, and there is no gauge symmetry. Then we have
- three real degrees of freedom from $A_\mu$
- one massive real degree of freedom from $\phi$
where I know there are three real degrees of freedom in $A_\mu$, because gauge fixing always removes one and we have no gauge fixing here.
What confuses me is that there must be two massive degrees of freedom, just as there were in the original theory. So that somehow means that one of the degrees of freedom in $A_\mu$ is massive while the other two aren't -- but how can that be? There's no mass term for $A_\mu$ in sight.