3.II.19G

Let $U$ be a finite-dimensional real vector space endowed with a positive definite inner product. A linear map $\tau: U \rightarrow U$ is said to be an orthogonal projection if $\tau$ is self-adjoint and $\tau^{2}=\tau$.

(a) Prove that for every orthogonal projection $\tau$ there is an orthogonal decomposition

$U=\operatorname{ker}(\tau) \oplus \operatorname{im}(\tau)$

(b) Let $\phi: U \rightarrow U$ be a linear map. Show that if $\phi^{2}=\phi$ and $\phi \phi^{*}=\phi^{*} \phi$, where $\phi^{*}$ is the adjoint of $\phi$, then $\phi$ is an orthogonal projection. [You may find it useful to prove first that if $\phi \phi^{*}=\phi^{*} \phi$, then $\phi$ and $\phi^{*}$ have the same kernel.]

(c) Show that given a subspace $W$ of $U$ there exists a unique orthogonal projection $\tau$ such that $\operatorname{im}(\tau)=W$. If $W_{1}$ and $W_{2}$ are two subspaces with corresponding orthogonal projections $\tau_{1}$ and $\tau_{2}$, show that $\tau_{2} \circ \tau_{1}=0$ if and only if $W_{1}$ is orthogonal to $W_{2}$.

(d) Let $\phi: U \rightarrow U$ be a linear map satisfying $\phi^{2}=\phi$. Prove that one can define a positive definite inner product on $U$ such that $\phi$ becomes an orthogonal projection.

*Typos? Please submit corrections to this page on GitHub.*