I think this might be a fairly "trivial" question.
Given a block diagonal unitary $$\hat{U} = \sum_{\alpha=0}^N \hat{U}_\alpha$$ (where $\hat{U}_\alpha$ is a matrix of all zeros apart from a $k_\alpha\times k_\alpha$ unitary subspace),
we can say $\hat{U}_\alpha = \hat{U}_\alpha \Pi_\alpha$, where $\Pi_\alpha$ is a projection operator onto the $\alpha$-th subspace.
My question is this: if we evolve any density operator through $\hat{U}$ is it correct to assume that the final state will be always block diagonal?
My reasoning is quite simple: given a density operator $$\sigma^0 = \sum_{\eta=0}^dp_\eta |{\phi_\eta}\rangle \langle{\phi'_\eta}|$$ evolving it through $\hat{U}$ means $$\sigma^f = \hat{U}\sigma^0\hat{U}^\dagger = \sum_{\alpha=0}^N \hat{U}_\alpha \Pi_\alpha\sigma^0\Pi_\alpha\hat{U}_\alpha^\dagger = \sum_{\alpha=0}^N \hat{U}_\alpha \sigma_\alpha^0 \hat{U}_\alpha^\dagger $$
where $\sum_\alpha^N \sigma^0_\alpha$ is a block diagonal operator comprised of non-normalized density operators $\sigma_\alpha$. Henceforth $\sigma^f$ will be block diagonal.