Let $V$ be a finitely-generated inner product space and let $\alpha $ be a normal endomorphism of $V$ . If the minimal polynomial of $\alpha$ is completely reducible, then it does not have multiple roots.
This is a theorem from this book http://www.springer.com/us/book/9789400726352. I have a problem in a part of the proof. In this book $\sigma_0$ is being used to denote the $0$-map and $\sigma_1$ is being used to denote the identity map.
Proof: Let $p(X)$ be the minimal polynomial of $\alpha$, which we assume is completely reducible. Assume that there exists a scalar $c$ and a polynomial $q(X)$ such that $p(X) = (X − c)^2q(X)$. Since $p(\alpha) = \sigma_0$, we have $(\alpha − c\sigma_1)^2q(\alpha) = \sigma_0$ and so $ker((α − c\sigma_1)^2q(\alpha)) = V $. Note that $\beta = \alpha − c\sigma_1$ is a normal endomorphism of $V$ . Let $v \in V$ and let $w = q(\alpha)(v)$. Then $\beta^2(w) = 0_V$ and so $\beta(w) \in Im(\beta) \cap Ker(\beta) = {0_V }$. Thus we see that $\beta q(\alpha)(v) = 0_V$ for all $v \in V$ and hence $α$ annihilates the polynomial $(X − c)q(X)$, contradicting the minimality of $p(X)$.
$\beta=\alpha-c\sigma_1$, I know why it's normal since $\beta\beta^*=\beta^*\beta$. What I don't understand is that why it says $Im(\beta) \cap Ker(\beta) = {0_V }$.