Chapter 26 — Key Takeaways

The one idea

A real matrix with complex eigenvalues is rotating, and the complex eigenvalue $\lambda = re^{i\theta}$ reports the rotation angle ($\theta$) and the scale factor ($r$) of the spin inside an invariant 2D plane. This is the chapter's threshold concept: once you read $re^{i\theta}$ as "turn by $\theta$, scale by $r$," complex eigenvalues stop being an algebraic accident you tolerate and become a geometric signal you can read — the unmistakable fingerprint of oscillation, spiraling, and cyclic dynamics.

The big ideas, in order

  1. Why complex at all (geometry first). A pure rotation turns every direction, so it has no real invariant direction — hence no real eigenvector and no real eigenvalue. The missing real eigenvectors are the fingerprint of spinning, and the eigenvalues flee into $\mathbb{C}$ to encode it: the rotation by $\theta$ has eigenvalues $e^{\pm i\theta}$.
  2. Conjugate-pair theorem (state the condition!). For a real matrix, complex eigenvalues come in conjugate pairs $a \pm bi$, with conjugate eigenvectors $\mathbf{v}, \bar{\mathbf{v}}$. Proof: conjugate the eigen-equation; the real $A$ is unchanged, yielding $A\bar{\mathbf{v}} = \bar\lambda\bar{\mathbf{v}}$. Consequence: real-matrix eigenvalues are symmetric across the real axis, and an odd-dimensional real matrix has at least one real eigenvalue.
  3. Rotation-plus-scaling interpretation. $\lambda = re^{i\theta}$ with $r = |\lambda| = \sqrt{a^2+b^2}$ and $\theta = \arg(\lambda) = \operatorname{atan2}(b,a)$: the argument is the rotation, the modulus is the stretch. Multiplying by a complex number is rotate-and-scale, and on the invariant plane $A$ acts exactly as multiplication by $\lambda$.
  4. The real canonical (block) form. A real $2\times 2$ matrix with eigenvalues $a \pm bi$ is similar over the reals to $C = \begin{psmallmatrix}a & -b\\ b & a\end{psmallmatrix} = r\begin{psmallmatrix}\cos\theta & -\sin\theta\\ \sin\theta & \cos\theta\end{psmallmatrix}$, via $A = PCP^{-1}$ with $P = [\operatorname{Re}\mathbf{v}\mid-\operatorname{Im}\mathbf{v}]$. This trades a complex diagonal matrix for a real rotation-scaling block — the everyday working form.
  5. Modulus controls the dynamics (stability preview). Under iteration $\mathbf{x}_{n+1} = A\mathbf{x}_n$: $r > 1$ spirals out (unstable), $r < 1$ spirals in (stable, damped oscillation), $r = 1$ holds a pure orbit (marginal). Powers are easy: $A^n = PC^nP^{-1}$ with $C^n = r^n\begin{psmallmatrix}\cos n\theta & -\sin n\theta\\ \sin n\theta & \cos n\theta\end{psmallmatrix}$ — De Moivre in matrix form. Continuous-time version (real part, not modulus) comes in Chapter 37.
  6. Quick $2\times 2$ test. Complex eigenvalues $\iff \operatorname{tr}(A)^2 - 4\det(A) < 0$. Then the real part is $\operatorname{tr}(A)/2$ and $r = \sqrt{\det(A)}$ (so $\det(A) = r^2$, the area-scaling identity).
  7. The unitary bridge. A real orthogonal matrix (rotation/reflection) has all eigenvalues of modulus $1$ — on the unit circle. The complex generalization is a unitary matrix ($U^{*}U = I$), the length-preserving map of $\mathbb{C}^n$, also with unit-circle eigenvalues — the mathematics of quantum gates (Chapters 27, 34).

Skills you gained

  • Explain geometrically why a rotation has no real eigenvectors, and predict complex eigenvalues from the discriminant.
  • Apply the conjugate-pair theorem, and reason about how many real eigenvalues a real matrix of given size must have.
  • Convert a complex eigenvalue to polar form and read off the rotation angle and scale factor.
  • Build the real canonical block $\begin{psmallmatrix}a & -b\\ b & a\end{psmallmatrix}$ from a complex eigenpair, and verify $A = PCP^{-1}$.
  • Use $|\lambda|$ to classify a discrete system as spiral-in (stable), orbit (marginal), or spiral-out (unstable), and compute matrix powers via the block form.
  • Compute and interpret complex eigenvalues and eigenvectors with np.linalg.eig, mindful that outputs are unsorted, complex-normalized, and phase-ambiguous.

Terms to know

complex eigenvalue · complex conjugate pair · complex eigenvector · modulus $|\lambda| = r$ · argument $\arg(\lambda) = \theta$ · polar form $re^{i\theta}$ · rotation-scaling · real canonical (block) form $\begin{psmallmatrix}a & -b\\ b & a\end{psmallmatrix}$ · invariant plane · spiral (in / out / orbit) · discriminant $\operatorname{tr}^2 - 4\det$ · De Moivre's theorem · unitary matrix · complex norm $\sqrt{\mathbf{v}^{*}\mathbf{v}}$

How this connects to the book's themes

  • Eigenvalues reveal what a matrix really does. This chapter is the dramatic case: a matrix with complicated-looking entries is, after a real change of basis, nothing but a rotation and a scaling. Its true nature — spin and stretch — was hidden by the coordinate system, and the complex eigenvalue exposes it.
  • Geometry and algebra are two views of one object. "A rotation has no invariant line" (geometry) and "the characteristic polynomial has complex roots" (algebra) are the same fact; $re^{i\theta}$ is simultaneously a root of a polynomial, a rotation angle, a scale factor, and the growth rate of a spiral.
  • Linear algebra is the most applied branch of pure mathematics. The identical complex-eigenvalue analysis describes a ringing circuit, a damped vibration, a predator–prey cycle, a business cycle, and a quantum gate — physics, engineering, ecology, economics, and computing, all spinning to the same algebra.

Where this is going

  • Chapter 27 (The Spectral Theorem) is the complementary extreme: symmetric real matrices have no complex eigenvalues at all — they are pure stretching, diagonalized by an orthogonal matrix. Complex eigenvalues are exactly what symmetry rules out.
  • Chapter 34 (Inner Product Spaces) formalizes the complex norm $\sqrt{\mathbf{v}^{*}\mathbf{v}}$ glimpsed here and builds the unitary matrices that preserve it — the proper home of the complex eigenvectors you computed.
  • Chapter 37 (Matrix Exponential and ODEs) turns the discrete spiral into continuous time: $\mathbf{x}' = A\mathbf{x}$ has solutions $e^{\sigma t}\cos(\omega t)$, with the complex eigenvalue $\sigma \pm i\omega$ splitting into decay rate and frequency — the continuous mirror of modulus and argument, and the full theory of stability and phase portraits.
  • Chapter 36 (Jordan Normal Form) generalizes the real block form to the defective and higher-dimensional cases, where conjugate pairs give $2\times 2$ rotation-scaling blocks inside a real block-diagonal decomposition.