Solution (source code)

= Solution

With only $\alpha_3$ nonzero, $T=\operatorname{diag}(\alpha_3,-\alpha_3)$ and
$$
U=\operatorname{diag}(e^{i\alpha_3/2},e^{-i\alpha_3/2}).
$$
For $\alpha_3\notin2\pi\mathbb Z$, the <eigenvectors> are \b[the nonzero multiples of $(1,0)^T$ and $(0,1)^T$]. When $\alpha_3=2\pi k$, $U=(-1)^kI$ and \b[every nonzero vector is an eigenvector].