Appendix B — Homework 1
Exercise B.1 Let \(P: V \to V\) be a projection.
Show that \(\mathbb{I}-P: V \to V\) is also a projection, that \(\ker P = \operatorname{im}(\mathbb{I}-P)\), and that \(V = \ker P \oplus \operatorname{im}P\).
Show that an eigenvalue of a projection \(P\) can only be \(0\) or \(1\). Conclude that \(\operatorname{Tr}(P) = \dim(\operatorname{im}P)\), i.e., the trace computes the dimension of the subspace projected onto by \(P\).
Exercise B.2 If \(V\) is an inner product space and \(\|x_0\| = 1\), show \(P(x) \coloneqq \langle x_0, x \rangle x_0\) is an orthogonal projection (cf. Remark 2.6) onto \(\operatorname{Span}\{x_0\}\).
Exercise B.3 Show that a pairwise orthonormal set \(\mathscr{A}\) of an inner product space \(V\) is necessarily a linearly independent set.
Exercise B.4 Verify that the set \(S = \left\{ \frac{1}{\sqrt{2\pi}} e^{int}\ \middle|\ n \in \mathbb{Z}\right\}\) is pairwise orthonormal in the inner product space \(C([-\pi,\pi],\mathbb{C})\) with respect to \[ \langle f, g \rangle \coloneqq \int_{-\pi}^{\pi} \overline{f(t)} g(t) \mathop{}\!\mathrm{d}{t}. \]
Exercise B.5 Let \(Q \in \operatorname{Mat}_n(\mathbb{C})\). Prove that the following are equivalent:
- \(Q \in \operatorname{U}(n)\) (see Definition 2.29).
- The columns of \(Q\) are orthonormal with respect to the Hermitian dot product.
- The rows of \(Q\) are orthonormal with respect to the Hermitian dot product.
Exercise B.6 Let \(A = \begin{psmallmatrix} -10 & 18 & 18 \\ 3 & -7 & -6 \\ -9 & 18 & 17 \\ \end{psmallmatrix}\). Compute \(A^{10}\).
Remark: You can easily do this by brute force with a computer algebra system like SageMath, but that isn’t the point—we can do this manually! The hint here is: if \(S^{-1} A S\) is diagonal, computing its powers is easy. Does that help with powers of \(A\)?
Exercise B.7 Let \(V\) be an inner product space with \(L: V \to V\) a linear transformation.
Show that \(L\) can be decomposed as \(L = A+Bi\), where \(A\) and \(B\) are self-adjoint operators.
Show that \([L^*,L] = 2i[A,B],\) i.e., \(L\) is normal if and only if its real and imaginary part commute.
(Bonus) Suppose we have proven the spectral theorem (Theorem 2.19) for Hermitian matrices: Every \(M\) satisfying \(M = M^*\) admits some \(Q \in \operatorname{U}(n)\) such that \(Q^* M Q\) is diagonal. Using this, prove the spectral theorem for normal matrices.
Hint: Use (b) and modify the proof of Theorem 2.15 to show commuting normal matrices admit a simultaneous unitary diagonalization.
Show that the sum and product of commuting normal matrices is normal.
Remark: You may use the content and results of part (c) even if you did not prove them: Theorem 2.19 and the fact that, when two normal matrices commute, they simultaneously unitarily diagonalize.