This sheet introduces the fundamental concepts of linear transformations, eigenvalues, eigenvectors, characteristic polynomials, and minimal polynomials—the essential machinery for understanding linear transformations and their canonical forms.

Commentary. A linear transformation T:V→VT: V \to V is completely determined by its matrix representation. But a single matrix contains n2n^2 entries—far too many to understand at a glance. Eigenvalues and eigenvectors reduce this complexity: if vv is an eigenvector, then TT acts on vv by simple scalar multiplication. The minimal polynomial captures the essential algebraic behavior of TT with a polynomial of degree at most nn, distilling all the structural information into a concise form. These concepts are the foundation for the rational and Jordan canonical forms developed in Sheet 11.

Prerequisites. This sheet uses the following results from earlier sheets:

  • Vector spaces, linear independence, basis, dimension (Sheet 12 — yes, we are jumping ahead slightly; Sheet 12 covers field extensions but also includes vector space fundamentals. Alternatively, this material is standard from linear algebra.)
  • Determinants (standard linear algebra)
  • Polynomial rings over a field (Sheet 9)
  • PIDs and UFDs (Sheet 9)

Linear Transformations and Matrices#

Definition 10.1. Let VV and WW be vector spaces over a field FF. A function T:V→WT: V \to W is a linear transformation (or linear map) if for all u,v∈Vu, v \in V and c∈Fc \in F:T(u+v)=T(u)+T(v),T(cv)=cT(v).T(u + v) = T(u) + T(v), \qquad T(cv) = cT(v).

Definition 10.2. Let VV be a finite-dimensional vector space over FF. The set of all linear transformations T:V→VT: V \to V is denoted End⁡F(V)\operatorname{End}_F(V).

Theorem 10.3. Let VV be an nn-dimensional vector space over FF. For any basis B\mathcal{B} of VV, the mapΦ:End⁡F(V)→Mn(F),Φ(T)=[T]B\Phi: \operatorname{End}_F(V) \to M_n(F), \qquad \Phi(T) = [T]_{\mathcal{B}}is a ring isomorphism. In particular, End⁡F(V)≅Mn(F)\operatorname{End}_F(V) \cong M_n(F).

Corollary 10.4. Eigenvalues, characteristic polynomials, and minimal polynomials are well-defined for linear transformations (independent of the choice of basis).

Corollary 10.5. Two matrices are similar iff they represent the same linear transformation in different bases.

Exercise 10.6. Let T:R2→R2T: \mathbb{R}^2 \to \mathbb{R}^2 be defined by T(x,y)=(y,0)T(x, y) = (y, 0). Find the matrix of TT with respect to the standard basis.

Exercise 10.7. Prove that composition of linear transformations corresponds to matrix multiplication.


Eigenvalues and Eigenvectors#

Definition 10.8. Let VV be a finite-dimensional vector space over a field FF, and let T:V→VT: V \to V be a linear transformation. A scalar λ∈F\lambda \in F is an eigenvalue of TT if there exists a nonzero vector v∈Vv \in V such that:T(v)=λv.T(v) = \lambda v.Such a vector vv is an eigenvector of TT corresponding to λ\lambda.

Definition 10.9. For an n×nn \times n matrix A∈Mn(F)A \in M_n(F), a scalar λ∈F\lambda \in F is an eigenvalue of AA if there exists a nonzero vector v∈Fnv \in F^n such that:Av=λv.A v = \lambda v.

Definition 10.10. The eigenspace of AA corresponding to λ\lambda is:Eλ={v∈Fn∣Av=λv}=ker⁡(A−λI).E_\lambda = \{v \in F^n \mid A v = \lambda v\} = \ker(A - \lambda I).

Theorem 10.11. λ∈F\lambda \in F is an eigenvalue of AA iff det⁡(A−λI)=0\det(A - \lambda I) = 0.

Theorem 10.12 (Distinct Eigenvectors are Independent). If λ1,…,λk\lambda_1, \ldots, \lambda_k are distinct eigenvalues of AA, and viv_i is an eigenvector for λi\lambda_i, then {v1,…,vk}\{v_1, \ldots, v_k\} is linearly independent.

Example 10.13. For A=(2112)A = \begin{pmatrix} 2 & 1 \\ 1 & 2 \end{pmatrix}:det⁡(A−λI)=det⁡(2−λ112−λ)=(2−λ)2−1=λ2−4λ+3=(λ−1)(λ−3).\det(A - \lambda I) = \det\begin{pmatrix} 2-\lambda & 1 \\ 1 & 2-\lambda \end{pmatrix} = (2-\lambda)^2 - 1 = \lambda^2 - 4\lambda + 3 = (\lambda - 1)(\lambda - 3).Eigenvalues: λ=1,3\lambda = 1, 3. Eigenvector for λ=1\lambda = 1: (1,−1)(1, -1). Eigenvector for λ=3\lambda = 3: (1,1)(1, 1).

Exercise 10.14. Find the eigenvalues and eigenvectors of A=(1221)A = \begin{pmatrix} 1 & 2 \\ 2 & 1 \end{pmatrix}.

Exercise 10.15. Show that λ=0\lambda = 0 is an eigenvalue of AA iff AA is singular.

Exercise 10.16. Prove Theorem 10.12.


The Characteristic Polynomial#

Definition 10.17. The characteristic polynomial of A∈Mn(F)A \in M_n(F) is:χA(t)=det⁡(tI−A)=tn+cn−1tn−1+⋯+c0.\chi_A(t) = \det(tI - A) = t^n + c_{n-1}t^{n-1} + \cdots + c_0.

Theorem 10.18. Similar matrices have the same characteristic polynomial. That is, if B=P−1APB = P^{-1}AP, then χB(t)=χA(t)\chi_B(t) = \chi_A(t).

Theorem 10.19. The constant term of χA(t)\chi_A(t) is χA(0)=(−1)ndet⁡A\chi_A(0) = (-1)^n \det A.

Theorem 10.20. The coefficient of tn−1t^{n-1} in χA(t)\chi_A(t) is −tr⁡(A)-\operatorname{tr}(A), where tr⁡(A)\operatorname{tr}(A) is the sum of the diagonal entries.

Example 10.21. For A=(abcd)A = \begin{pmatrix} a & b \\ c & d \end{pmatrix}, χA(t)=t2−(a+d)t+(ad−bc)\chi_A(t) = t^2 - (a+d)t + (ad - bc).

Exercise 10.22. Compute the characteristic polynomial of A=(100020003)A = \begin{pmatrix} 1 & 0 & 0 \\ 0 & 2 & 0 \\ 0 & 0 & 3 \end{pmatrix}.

Exercise 10.23. Show that similar matrices have the same trace and determinant.


The Cayley-Hamilton Theorem#

Theorem 10.24 (Cayley-Hamilton). Every square matrix AA satisfies its own characteristic polynomial:χA(A)=0.\chi_A(A) = 0.

Example 10.25. For A=(2112)A = \begin{pmatrix} 2 & 1 \\ 1 & 2 \end{pmatrix}, χA(t)=t2−4t+3\chi_A(t) = t^2 - 4t + 3. Indeed:A2−4A+3I=0.A^2 - 4A + 3I = 0.

Theorem 10.26. The Cayley-Hamilton theorem provides an algorithm for computing A−1A^{-1}:A−1=−1c0(An−1+cn−1An−2+⋯+c1I)A^{-1} = -\frac{1}{c_0}(A^{n-1} + c_{n-1}A^{n-2} + \cdots + c_1 I)provided det⁡A≠0\det A \neq 0.

Example 10.27. For A=(1234)A = \begin{pmatrix} 1 & 2 \\ 3 & 4 \end{pmatrix}, χA(t)=t2−5t−2\chi_A(t) = t^2 - 5t - 2. Thus A2−5A−2I=0A^2 - 5A - 2I = 0, so A−1=12(A−5I)A^{-1} = \frac{1}{2}(A - 5I).

Exercise 10.28. Verify Cayley-Hamilton for A=(1234)A = \begin{pmatrix} 1 & 2 \\ 3 & 4 \end{pmatrix}.

Exercise 10.29. Use the Cayley-Hamilton theorem to find A−1A^{-1} for A=(2112)A = \begin{pmatrix} 2 & 1 \\ 1 & 2 \end{pmatrix}.


The Minimal Polynomial#

Definition 10.30. The minimal polynomial of A∈Mn(F)A \in M_n(F), denoted mA(t)m_A(t), is the monic polynomial of least degree such that:mA(A)=0.m_A(A) = 0.

Theorem 10.31. For any matrix AA, the minimal polynomial exists and is unique.

Theorem 10.32. The minimal polynomial divides every polynomial that annihilates AA. In particular, mA(t)m_A(t) divides χA(t)\chi_A(t).

Theorem 10.33. If B=P−1APB = P^{-1}AP, then mB(t)=mA(t)m_B(t) = m_A(t). (Similar matrices have the same minimal polynomial.)

Theorem 10.34. λ∈F\lambda \in F is an eigenvalue of AA iff λ\lambda is a root of mA(t)m_A(t).

Theorem 10.35. The roots of the minimal polynomial are exactly the eigenvalues of AA.

Example 10.36. For A=(200020003)A = \begin{pmatrix} 2 & 0 & 0 \\ 0 & 2 & 0 \\ 0 & 0 & 3 \end{pmatrix}, χA(t)=(t−2)2(t−3)\chi_A(t) = (t-2)^2(t-3). The minimal polynomial is mA(t)=(t−2)(t−3)m_A(t) = (t-2)(t-3).

Example 10.37. For A=(210020003)A = \begin{pmatrix} 2 & 1 & 0 \\ 0 & 2 & 0 \\ 0 & 0 & 3 \end{pmatrix}, χA(t)=(t−2)2(t−3)\chi_A(t) = (t-2)^2(t-3). The minimal polynomial is mA(t)=(t−2)2(t−3)m_A(t) = (t-2)^2(t-3).

Exercise 10.38. Find the minimal polynomial of A=(1101)A = \begin{pmatrix} 1 & 1 \\ 0 & 1 \end{pmatrix}.

Exercise 10.39. Show that if AA is diagonalizable, then mA(t)m_A(t) has no repeated roots.

Exercise 10.40. Prove Theorem 10.31.


Diagonalizability#

Definition 10.41. A matrix A∈Mn(F)A \in M_n(F) is diagonalizable over FF if there exists an invertible matrix P∈Mn(F)P \in M_n(F) such that:P−1AP=DP^{-1}AP = Dis a diagonal matrix.

Theorem 10.42. AA is diagonalizable over FF iff FnF^n has a basis consisting of eigenvectors of AA.

Theorem 10.43. If AA has nn distinct eigenvalues in FF, then AA is diagonalizable over FF.

Theorem 10.44. AA is diagonalizable over FF iff the minimal polynomial mA(t)m_A(t) factors into distinct linear factors over FF.

Theorem 10.45. The geometric multiplicity of an eigenvalue λ\lambda (dimension of its eigenspace EλE_\lambda) is at most its algebraic multiplicity (multiplicity as a root of χA(t)\chi_A(t)):dim⁡Eλ≤mult⁡χA(λ).\dim E_\lambda \leq \operatorname{mult}_{\chi_A}(\lambda).

Theorem 10.46. AA is diagonalizable iff equality holds in Theorem 10.45 for every eigenvalue.

Example 10.47. A=(2102)A = \begin{pmatrix} 2 & 1 \\ 0 & 2 \end{pmatrix} is not diagonalizable over R\mathbb{R}. Its characteristic polynomial is (t−2)2(t-2)^2, but dim⁡E2=1<2\dim E_2 = 1 < 2.

Exercise 10.48. Determine whether A=(1221)A = \begin{pmatrix} 1 & 2 \\ 2 & 1 \end{pmatrix} is diagonalizable over R\mathbb{R}. If so, find PP and DD.

Exercise 10.49. Show that if AA is diagonalizable, then every eigenvalue has algebraic multiplicity equal to geometric multiplicity.

Exercise 10.50. Prove Theorem 10.44.


Invariant Subspaces#

Definition 10.51. A subspace W⊆VW \subseteq V is TT-invariant if T(W)⊆WT(W) \subseteq W.

Theorem 10.52. If WW is TT-invariant, then the restriction T∣W:W→WT|_W: W \to W is a linear transformation.

Theorem 10.53. Eigenspaces are invariant subspaces.

Theorem 10.54. If VV decomposes as a direct sum of TT-invariant subspaces V=W1⊕⋯⊕WkV = W_1 \oplus \cdots \oplus W_k, then the matrix of TT is block diagonal with respect to a basis adapted to this decomposition.

Exercise 10.55. Show that ker⁡(T−λI)\ker(T - \lambda I) and im⁡(T−λI)\operatorname{im}(T - \lambda I) are TT-invariant subspaces.

Exercise 10.56. Let V=R2V = \mathbb{R}^2, T(x,y)=(y,0)T(x, y) = (y, 0). Find all TT-invariant subspaces of VV.

Exercise 10.57. Prove that if WW is TT-invariant, then the characteristic polynomial of T∣WT|_W divides the characteristic polynomial of TT.


Schur's Triangularization Theorem#

Theorem 10.58 (Schur's Triangularization Theorem). If A∈Mn(C)A \in M_n(\mathbb{C}), then there exists a unitary matrix UU such that U−1AUU^{-1}AU is upper triangular.

Corollary 10.59. Every complex matrix is triangularizable.

Corollary 10.60. The eigenvalues of AA appear on the diagonal of the triangular matrix.

Corollary 10.61. Every complex matrix has at least one eigenvector.

Exercise 10.62. Why does Schur's theorem imply that every complex matrix has at least one eigenvector?

Exercise 10.63. Show that the eigenvalues of a triangular matrix are exactly its diagonal entries.


The Characteristic Polynomial of a Block Triangular Matrix#

Theorem 10.64. If A=(BC0D)A = \begin{pmatrix} B & C \\ 0 & D \end{pmatrix} is a block upper triangular matrix, then:χA(t)=χB(t)⋅χD(t).\chi_A(t) = \chi_B(t) \cdot \chi_D(t).

Corollary 10.65. The eigenvalues of AA are the eigenvalues of BB together with the eigenvalues of DD.

Exercise 10.66. Let AA be a block diagonal matrix with blocks A1,…,AkA_1, \ldots, A_k. Show that χA(t)=χA1(t)⋯χAk(t)\chi_A(t) = \chi_{A_1}(t) \cdots \chi_{A_k}(t).

Exercise 10.67. Let AA be upper triangular. Show that the eigenvalues of AA are the diagonal entries.


Summary#

We have introduced:

  • Linear transformations and their matrix representations
  • Eigenvalues and eigenvectors
  • Characteristic polynomials and their properties
  • The Cayley-Hamilton Theorem
  • Minimal polynomials
  • Diagonalizability criteria
  • Algebraic vs. geometric multiplicity
  • Invariant subspaces
  • Schur's Triangularization Theorem
  • Characteristic polynomial of block triangular matrices