Trace and determinant from the structure
The trace and the determinant of a matrix are computed from its entries, and a change of basis changes every entry. Why do the two numbers survive, and what do they say about the operator?
Where you are 3 in Structure of operators
Why this exists
The characteristic equation and finding eigenvalues found that the eigenvalues of a 2 \times 2 matrix add to its trace and multiply to its determinant, and sketched the same for larger matrices. This course built eigenvalues, triangular form and the Jordan form without determinants. With that structure in hand, the two numbers can be explained: both belong to the operator, not to the basis, and over \mathbb{C} they are the sum and the product of the eigenvalues, each counted \dim G_\lambda times.
Definition
Throughout, V is a finite-dimensional vector space over \mathbb{F}, n = \dim V and T \in \mathcal{L}(V).
The trace of a square matrix A = [a_{ij}] \in \mathbb{F}^{n \times n} is the sum of its diagonal entries,
The trace and the determinant of T are
for any basis \mathcal{B} of V; Result 2 shows that the basis does not matter. The determinant of an operator was already defined this way in Generalized eigenvectors.
Let V \neq \{\mathbf{0}\} be complex. Then m_T is a product of linear factors (The fundamental theorem of algebra), so Results 5 to 7 of Generalized eigenvectors apply. Let \lambda_1, \dots, \lambda_m be the distinct eigenvalues of T and d_j = \dim G_{\lambda_j} their multiplicities. The eigenvalues of T listed with multiplicity form the list (\mu_1, \dots, \mu_n) in which each \lambda_j appears d_j times; its length is n by Result 6(a) of Generalized eigenvectors, and by Result 7(a) there, d_j is also the multiplicity of \lambda_j as a zero of p_T. For a real or complex n \times n matrix A, the eigenvalues listed with multiplicity are the complex zeros of p_A(\lambda) = \det(A - \lambda I_n), each repeated as often as it is a zero.
Intuition
In a basis of generalized eigenvectors, Result 5(b) of Generalized eigenvectors makes [T] upper triangular, with each eigenvalue \lambda_j appearing d_j times on the diagonal. For a triangular matrix the trace is the sum of the diagonal and the determinant its product, so in that basis both numbers are visibly built from the eigenvalues. Another basis replaces the matrix A by P^{-1}AP, which changes the entries but neither number: the determinant because determinants multiply, the trace because \operatorname{tr}(XY) = \operatorname{tr}(YX) whenever both products exist. So what one good basis shows, every basis gives.
Results
V is finite-dimensional, n = \dim V, and S, T \in \mathcal{L}(V). Results 1 and 2 hold over any field.
Result 1 (the trace of a product). For A \in \mathbb{F}^{m \times n} and B \in \mathbb{F}^{n \times m}, \operatorname{tr}(AB) = \operatorname{tr}(BA). On \mathbb{F}^{n \times n} the trace is linear: \operatorname{tr}(aA + bB) = a\operatorname{tr} A + b\operatorname{tr} B.
Result 2 (trace and determinant of an operator). (a) Similar matrices have the same trace and the same determinant, so \operatorname{tr} T and \det T do not depend on the basis. (b) \operatorname{tr} is a linear functional on \mathcal{L}(V), and \operatorname{tr}(ST) = \operatorname{tr}(TS). (c) \det(ST) = \det S\det T and \det I = 1. (d) T is invertible if and only if \det T \neq 0, and then \det T^{-1} = 1/\det T.
Result 3 (sum and product of the eigenvalues). Let V \neq \{\mathbf{0}\} be complex, and let T have distinct eigenvalues \lambda_1, \dots, \lambda_m with multiplicities d_1, \dots, d_m. Then
Result 4 (real matrices). Let A \in \mathbb{R}^{n \times n} with n \ge 1. Then \operatorname{tr} A and \det A are the sum and the product of the complex eigenvalues of A listed with multiplicity. A nonreal eigenvalue \lambda and its conjugate \bar{\lambda} have the same multiplicity, so \det A is the product of the real eigenvalues, listed with multiplicity, times a positive number. The same holds for an operator on a real space, through its matrix in any basis.
Result 5 (coefficients of the characteristic polynomial). Let \mathbb{F} be \mathbb{R} or \mathbb{C} and n \ge 2 (no c terms when n = 2). Then for some scalars c_1, \dots, c_{n-2},
Result 6 (polynomials in T). Let V \neq \{\mathbf{0}\} be complex, with the notation of Result 3, and let q be a polynomial with complex coefficients. Then
In particular \operatorname{tr} T^k = d_1\lambda_1^k + \dots + d_m\lambda_m^k for every k \ge 0.
Why it is true
Result 1. Entry (i, i) of AB is \sum_{k=1}^{n} a_{ik}b_{ki}, and entry (k, k) of BA is \sum_{i=1}^{m} b_{ki}a_{ik}. A finite double sum may be added in either order and multiplication in \mathbb{F} is commutative, so
Linearity holds entry by entry along the diagonal.
Result 2. (a) Let B = P^{-1}AP. Result 1 with the factors P^{-1} and AP gives \operatorname{tr} B = \operatorname{tr}(AP\,P^{-1}) = \operatorname{tr} A. By the product rule of Properties of determinants: row operations, products, transposes, \det B = \det P^{-1}\det A\det P = \det A, since \det P^{-1}\det P = \det I = 1. For two bases \mathcal{B} and \mathcal{C} of V, [T]_{\mathcal{C}} = P^{-1}[T]_{\mathcal{B}}P with P invertible (Result 5 of The matrix of a linear map). (b) Fix a basis \mathcal{B} = (\mathbf{b}_1, \dots, \mathbf{b}_n). Column k of [aS + bT]_{\mathcal{B}} is the coordinate vector of aS\mathbf{b}_k + bT\mathbf{b}_k, and the coordinate map is linear (Result 1 there), so [aS + bT]_{\mathcal{B}} = a[S]_{\mathcal{B}} + b[T]_{\mathcal{B}}; and [ST]_{\mathcal{B}} = [S]_{\mathcal{B}}[T]_{\mathcal{B}} (Result 4 there). Result 1 finishes. (c) The same matrix identity with the product rule, and [I]_{\mathcal{B}} = I_n. (d) T is invertible if and only if [T]_{\mathcal{B}} is (Result 6 of Invertibility and isomorphic spaces), if and only if \det[T]_{\mathcal{B}} \neq 0 (The invertible matrix theorem). Then (c) with S = T^{-1} gives \det T^{-1}\det T = 1.
Result 3. Result 5(b) of Generalized eigenvectors gives a basis \mathcal{B} in which [T]_{\mathcal{B}} is block diagonal with upper-triangular blocks, the block for \lambda_j of size d_j with every diagonal entry \lambda_j. So [T]_{\mathcal{B}} is upper triangular and its diagonal lists each \lambda_j exactly d_j times. The trace is the sum of the diagonal, and the determinant of a triangular matrix is the product of its diagonal (Determinants by cofactor expansion). By Result 2(a) these are \operatorname{tr} T and \det T.
Result 4. Let T_A(\mathbf{x}) = A\mathbf{x} on \mathbb{C}^n. Its matrix in the standard basis is A, so \operatorname{tr} T_A = \operatorname{tr} A, \det T_A = \det A and p_{T_A} = p_A. By Result 7(a) of Generalized eigenvectors, the eigenvalues of T_A listed with multiplicity are the zeros of p_A listed with multiplicity, and Result 3 gives the sum and the product. Now let \lambda be a nonreal zero of multiplicity d, so p_A(z) = (z - \lambda)^d s(z) for a polynomial s. Conjugating every coefficient leaves p_A unchanged, since its coefficients are real, so p_A(z) = (z - \bar{\lambda})^d\bar{s}(z), where \bar{s} has the conjugate coefficients. Hence \bar{\lambda} has multiplicity at least d, and by the same argument from \bar{\lambda}, exactly d. Grouped in conjugate pairs, the nonreal eigenvalues contribute factors (\lambda\bar{\lambda})^d = \lvert\lambda\rvert^{2d} > 0 to the product. For an operator on a real space, apply this to [T]_{\mathcal{B}} and use Result 2(a).
Result 5. The constant term is p_T(0) = \det(T - 0I) = \det T, over any field. Over \mathbb{C}, Result 7(a) of Generalized eigenvectors gives p_T(\lambda) = (\mu_1 - \lambda)\cdots(\mu_n - \lambda), with the eigenvalues listed with multiplicity. Expanding the product, \lambda^n arises only by taking -\lambda from every factor, with coefficient (-1)^n, and \lambda^{n-1} by taking -\lambda from all factors but one and \mu_i from that one, with coefficient (-1)^{n-1}(\mu_1 + \dots + \mu_n), which is (-1)^{n-1}\operatorname{tr} T by Result 3. Over \mathbb{R}, let A = [T]_{\mathcal{B}}: then p_T = p_A = p_{T_A} with T_A as in the proof of Result 4, and \operatorname{tr} T_A = \operatorname{tr} A = \operatorname{tr} T, so the complex case applies.
Result 6. In the basis \mathcal{B} of the proof of Result 3, [T]_{\mathcal{B}} is upper triangular with each \lambda_j on the diagonal d_j times. By Result 5 of Upper-triangular matrices over the complex numbers, [q(T)]_{\mathcal{B}} is upper triangular with each q(\lambda_j) on the diagonal d_j times. Read off the trace and determinant as in Result 3; for the last line take q(z) = z^k.
Worked examples
Two bases for a reflection of polynomials
Let (Tp)(x) = p(1 - x) on \mathcal{P}_2(\mathbb{R}). Find \operatorname{tr} T and \det T from the matrix in the basis \mathcal{B} = (1, x, x^2), again in \mathcal{C} = \big(1, x - \tfrac12, (x - \tfrac12)^2\big), and match them with the eigenvalues.
SolutionIn \mathcal{B}: T1 = 1, Tx = 1 - x and Tx^2 = 1 - 2x + x^2. Their coordinate vectors are the columns of
It is triangular, so \operatorname{tr} T = 1 - 1 + 1 = 1 and \det T = 1 \cdot (-1) \cdot 1 = -1.
In \mathcal{C}: substituting 1 - x turns x - \tfrac12 into -(x - \tfrac12), and (x - \tfrac12)^2 into (\tfrac12 - x)^2 = (x - \tfrac12)^2. So [T]_{\mathcal{C}} = \operatorname{diag}(1, -1, 1), again with trace 1 and determinant -1, as Result 2(a) requires. \mathcal{C} is a basis of eigenvectors, with \dim E_1 = 2 and \dim E_{-1} = 1, so the eigenvalues listed with multiplicity are 1, 1, -1: sum 1 and product -1, as Result 4 says.
Finding eigenvalues from the trace and determinant
Let A = \begin{bmatrix} 1 & -2 & -2 \\ 0 & 2 & 0 \\ 2 & -1 & 1 \end{bmatrix}. (a) Show that 2 is an eigenvalue of A and find the others from \operatorname{tr} A and \det A alone. (b) For q(z) = z^2 - 2z + 5, find \operatorname{tr} q(A) and \det q(A).
Solution(a) The second row of A - 2I_3 is zero, so \det(A - 2I_3) = 0 and 2 is an eigenvalue. Next, \operatorname{tr} A = 1 + 2 + 1 = 4. Expanding along the second row, whose only nonzero entry is the 2 in position (2, 2),
List the eigenvalues with multiplicity as 2, \mu, \nu. By Result 4, 2 + \mu + \nu = 4 and 2\mu\nu = 10, so \mu + \nu = 2 and \mu\nu = 5. Then (z - \mu)(z - \nu) = z^2 - 2z + 5, whose zeros are 1 \pm 2i: a conjugate pair beside a real eigenvalue, as Result 4 predicts.
(b) q(z) = (z - 1 - 2i)(z - 1 + 2i), so q(1 \pm 2i) = 0, and q(2) = 5. Each eigenvalue has multiplicity 1, so Result 6 for \mathbf{x} \mapsto A\mathbf{x} on \mathbb{C}^3 gives \operatorname{tr} q(A) = 5 and \det q(A) = 0. Directly,
Multiplying matrices on the left
Fix A \in \mathbb{F}^{2 \times 2} and let T(X) = AX on \mathbb{F}^{2 \times 2}. Show that \operatorname{tr} T = 2\operatorname{tr} A and \det T = (\det A)^2, and evaluate both for A = \begin{bmatrix} 2 & -1 \\ 5 & 3 \end{bmatrix}.
SolutionLet E_{jk} have 1 in position (j, k) and 0 elsewhere, and use the basis (E_{11}, E_{21}, E_{12}, E_{22}): the coordinates of X are its first column followed by its second. Column k of AX is A times column k of X, so in block form
Its trace is 2\operatorname{tr} A. Block diagonal matrices multiply block by block, so [T] = \operatorname{diag}(A, I_2)\operatorname{diag}(I_2, A). Expanding \det\operatorname{diag}(I_2, A) along its first two rows leaves \det A, likewise for \operatorname{diag}(A, I_2) along its last two, and the product rule gives \det T = (\det A)^2.
For the given A, \operatorname{tr} A = 5 and \det A = 6 + 5 = 11, so \operatorname{tr} T = 10 and \det T = 121.
Naomi’s lacewing insectary
Naomi rears green lacewings at an insectary that supplies farms. Every two weeks she counts eggs e_k, larvae \ell_k and adults a_k. In each two-week period half of the eggs become larvae, a third of the larvae become adults, and each adult lays 6 eggs and dies, so \mathbf{x}_{k+1} = L\mathbf{x}_k with \mathbf{x}_k = (e_k, \ell_k, a_k) and
Find the eigenvalues of L from \operatorname{tr} L, \operatorname{tr} L^2 and \det L, show that L^3 = I_3, and find how many eggs per adult would make every count double every six weeks.
SolutionMultiplying,
so \operatorname{tr} L = \operatorname{tr} L^2 = 0. Expanding along the first row, \det L = 6\det\operatorname{diag}(\tfrac12, \tfrac13) = 1. Let \mu_1, \mu_2, \mu_3 be the eigenvalues listed with multiplicity. By Results 4 and 6 (with q(z) = z^2), \sum\mu_i = 0, \sum\mu_i^2 = 0 and \mu_1\mu_2\mu_3 = 1. Squaring the first sum,
so (z - \mu_1)(z - \mu_2)(z - \mu_3) = z^3 - 1. The eigenvalues are the cube roots of 1: 1, \omega and \bar{\omega}, with \displaystyle \omega = -\tfrac12 + \tfrac{\sqrt{3}}{2}i.
They are distinct, so L = PDP^{-1} with D = \operatorname{diag}(1, \omega, \bar{\omega}) (Diagonalization), and D^3 = I_3 gives L^3 = PD^3P^{-1} = I_3. Every count repeats every six weeks: 600 eggs become 300 larvae, then 100 adults, then 600 eggs. With f eggs per adult the traces stay 0 and \det L = f/6, so the same steps give z^3 - f/6 and L^3 = \tfrac{f}{6}I_3. Doubling every six weeks needs f/6 = 2: 12 eggs per adult.
A skew-symmetric matrix of odd size is singular
Let A \in \mathbb{R}^{n \times n} with A^{T} = -A and n odd. Prove that \det A = 0 and \operatorname{tr} A = 0. Then find a nonzero vector in \ker A for A = \begin{bmatrix} 0 & 2 & -1 \\ -2 & 0 & 3 \\ 1 & -3 & 0 \end{bmatrix}.
SolutionBy Properties of determinants: row operations, products, transposes, \det A^{T} = \det A, and \det(-A) = (-1)^n\det A, since -A is A with each of its n rows multiplied by -1. As n is odd,
so \det A = 0. Each diagonal entry satisfies a_{ii} = -a_{ii}, so it is 0 and \operatorname{tr} A = 0.
The given matrix is skew-symmetric of size 3, so \ker A \neq \{\mathbf{0}\}. Its first row gives 2y - z = 0 and its second -2x + 3z = 0; with y = 1, z = 2 and x = 3, and the third row checks: 3 - 3 + 0 = 0. So (3, 1, 2) \in \ker A. Oddness is needed: \begin{bmatrix} 0 & 1 \\ -1 & 0 \end{bmatrix} is skew-symmetric with determinant 1.
Traces of powers detect nilpotence
Let V be complex with n = \dim V \ge 1. Prove that T is nilpotent if and only if \operatorname{tr} T^k = 0 for k = 1, \dots, n.
SolutionSuppose T^r = 0. If T\mathbf{v} = \lambda\mathbf{v} with \mathbf{v} \neq \mathbf{0}, then \mathbf{0} = T^r\mathbf{v} = \lambda^r\mathbf{v}, so \lambda = 0. Thus 0 is the only eigenvalue, with multiplicity n, and Result 6 gives \operatorname{tr} T^k = n \cdot 0^k = 0 for k \ge 1.
Conversely, suppose the traces vanish but T has a nonzero eigenvalue. Let \mu_1, \dots, \mu_r be its distinct nonzero eigenvalues, with multiplicities c_1, \dots, c_r; then 1 \le r \le n. Define
read as z/\mu_1 when r = 1. It has degree r \le n, q(\mu_1) = 1, and q(0) = q(\mu_2) = \dots = q(\mu_r) = 0. Writing q(z) = \sum_{k=1}^{r} a_kz^k, linearity of the trace (Result 2(b)) gives \operatorname{tr} q(T) = \sum_k a_k\operatorname{tr} T^k = 0. But Result 6 gives \operatorname{tr} q(T) = c_1q(\mu_1) = c_1 \ge 1, the other terms vanishing. This contradiction shows that 0 is the only eigenvalue. Then V = G_0 (Result 5(a) of Generalized eigenvectors), and T^n = 0 by Result 3(d) there.
One trace is not enough: \operatorname{diag}(1, -1) has trace 0, but its square has trace 2.
Boundaries
- Real eigenvalues may be missing. The quarter turn (x, y) \mapsto (-y, x) of \mathbb{R}^2 has trace 0, determinant 1 and no real eigenvalue; Result 4 counts its complex eigenvalues \pm i.
- Multiplicity means \dim G_\lambda, not \dim E_\lambda. J_2(1) has \dim E_1 = 1, yet its trace is 2: the eigenvalue 1 is counted twice.
- Trace and determinant do not decide similarity. I_2 and J_2(1) both have trace 2 and determinant 1, and only I_2 is diagonalizable. From n = 3 on they do not even decide the eigenvalues: \operatorname{diag}(1, -1, 0) and the zero matrix share trace 0 and determinant 0.
- Only cyclic reorderings keep the trace. Result 1 applied to (AB)C and A(BC) gives \operatorname{tr}(ABC) = \operatorname{tr}(CAB) = \operatorname{tr}(BCA), but other orders can differ (Exercise 4).
Where this goes next
This lesson closes the course. Results 3 and 4 finish what The characteristic equation and finding eigenvalues began for 2 \times 2 matrices: at every size the trace and determinant are the sum and product of the eigenvalues, each counted \dim G_\lambda times, the total size of its blocks in the Jordan canonical form. Over \mathbb{R}, \lvert\det T\rvert is the volume factor of What the determinant means, now known to belong to the operator itself.
Exercises
Trace and determinant of operators
- Let (Tp)(x) = (x + 1)p'(x) on \mathcal{P}_2(\mathbb{R}). Find \operatorname{tr} T, \det T and the eigenvalues of T, and confirm Result 4.
- Let T(X) = X^{T} on \mathbb{C}^{2 \times 2}. Find \operatorname{tr} T, \det T and the eigenvalues of T listed with multiplicity.
- Let B = \begin{bmatrix} 1 & 4 \\ 2 & 3 \end{bmatrix} and T(X) = XB on \mathbb{R}^{2 \times 2}. Find \operatorname{tr} T and \det T.
- Let A = \begin{bmatrix} 1 & 0 \\ 0 & 0 \end{bmatrix}, B = \begin{bmatrix} 0 & 1 \\ 0 & 0 \end{bmatrix} and C = \begin{bmatrix} 0 & 0 \\ 1 & 0 \end{bmatrix}. Compute \operatorname{tr}(ABC), \operatorname{tr}(CAB) and \operatorname{tr}(BAC). Which equality among them does Result 1 promise?
Eigenvalues from traces and determinants
- Let A = \begin{bmatrix} 1 & -3 & -3 \\ -3 & 1 & -3 \\ -3 & 3 & 1 \end{bmatrix}. Given that 1 is an eigenvalue, find the other two from \operatorname{tr} A and \det A. Then find \operatorname{tr} A^2 and \det(A - 3I_3) without multiplying matrices.
- An operator T on \mathbb{C}^5 has only the eigenvalues 2 and -1, and \operatorname{tr} T = 7. Find the multiplicities, \det T, and the coefficients of \lambda^4 and \lambda^0 in p_T.
- A real 3 \times 3 matrix A has the eigenvalue 1 + i and \operatorname{tr} A = 1. Find its other eigenvalues and \det A.
- A matrix A \in \mathbb{C}^{2 \times 2} has \operatorname{tr} A = 5 and \operatorname{tr} A^2 = 13. Find \det A and the eigenvalues of A.
- An operator T on \mathbb{C}^5 has Jordan form \operatorname{diag}(J_2(3), J_1(-2), J_1(i), J_1(-i)). Find \operatorname{tr} T, \det T, \operatorname{tr} T^2 and \det(T - I).
- Priya runs a second insectary, where half of the eggs become larvae, half of the larvae become adults, and each adult lays 32 eggs and dies, so L = \begin{bmatrix} 0 & 0 & 32 \\ \tfrac12 & 0 & 0 \\ 0 & \tfrac12 & 0 \end{bmatrix}. Find \operatorname{tr} L, \operatorname{tr} L^2, \det L, the eigenvalues of L and L^3. Starting with 10 adults and nothing else, how many adults are there after six periods?
Proofs
- Let A \in \mathbb{R}^{n \times n} with \det A < 0. Prove that A has a negative real eigenvalue.
- Let V be a real vector space with n \ge 1, and let T^2 = -I. Prove that n is even. Show that over \mathbb{C} such an operator exists in every dimension.
- Let \mathbb{F} be \mathbb{R} or \mathbb{C} and A, B \in \mathbb{F}^{n \times n} with A invertible. Prove that AB and BA have the same characteristic polynomial, and hence, over \mathbb{C}, the same eigenvalues listed with multiplicity.
- Let V \neq \{\mathbf{0}\} be complex and T invertible, with the notation of Result 3. Prove that \operatorname{tr} T^{-1} = \sum_j d_j/\lambda_j and \det T^{-1} = \prod_j \lambda_j^{-d_j}.
- Prove that every A \in \mathbb{F}^{2 \times 2}, over any field, satisfies A^2 - (\operatorname{tr} A)A + (\det A)I_2 = 0.
Answers
- T1 = 0, Tx = x + 1 and Tx^2 = 2x^2 + 2x, so in (1, x, x^2) the matrix is \begin{bmatrix} 0 & 1 & 0 \\ 0 & 1 & 2 \\ 0 & 0 & 2 \end{bmatrix}. It is triangular: \operatorname{tr} T = 3, \det T = 0, and the eigenvalues are 0, 1, 2, each once (Result 4 of Upper-triangular matrices over the complex numbers). Indeed 0 + 1 + 2 = 3 and 0 \cdot 1 \cdot 2 = 0.
- In (E_{11}, E_{12}, E_{21}, E_{22}), T fixes E_{11} and E_{22} and swaps E_{12} and E_{21}, so its matrix is I_4 with two columns swapped: \operatorname{tr} T = 2 and \det T = -1. The symmetric matrices, a space of dimension 3, are fixed, and E_{12} - E_{21} is sent to its negative, so the eigenvalues are 1, 1, 1, -1, with sum 2 and product -1.
- Row j of XB is row j of X times B, so in (E_{11}, E_{12}, E_{21}, E_{22}) the matrix is \operatorname{diag}(B^{T}, B^{T}). As in Example 3, \operatorname{tr} T = 2\operatorname{tr} B = 8 and \det T = (\det B)^2 = (-5)^2 = 25.
- BC = A, so ABC = A^2 = A and \operatorname{tr}(ABC) = 1. CA = C and CB = \begin{bmatrix} 0 & 0 \\ 0 & 1 \end{bmatrix}, so \operatorname{tr}(CAB) = 1. BA = 0, so \operatorname{tr}(BAC) = 0. Result 1 with the factors AB and C promises \operatorname{tr}(ABC) = \operatorname{tr}(CAB), a cyclic reordering; BAC is not one.
- \operatorname{tr} A = 3 and, expanding along the first row, \det A = 10 - 36 + 18 = -8. With eigenvalues 1, \mu, \nu: \mu + \nu = 2 and \mu\nu = -8, so \mu, \nu are the zeros of z^2 - 2z - 8 = (z - 4)(z + 2), and the eigenvalues are 1, 4, -2. By Result 6, \operatorname{tr} A^2 = 1 + 16 + 4 = 21 and \det(A - 3I_3) = (-2)(1)(-5) = 10.
- With multiplicities a and b, a + b = 5 and 2a - b = 7, so a = 4 and b = 1. Then \det T = 2^4(-1) = -16. By Result 5 the coefficient of \lambda^4 is (-1)^4\operatorname{tr} T = 7 and the constant term is \det T = -16; indeed p_T(\lambda) = -(\lambda - 2)^4(\lambda + 1).
- By Result 4, 1 - i is also an eigenvalue. A nonreal third eigenvalue would need its conjugate too, so the third, r, is real: (1 + i) + (1 - i) + r = 1 gives r = -1. Then \det A = (1 + i)(1 - i)(-1) = -2.
- With eigenvalues \mu, \nu, Results 3 and 6 give \mu + \nu = 5 and \mu^2 + \nu^2 = 13, so \det A = \mu\nu = \tfrac12(25 - 13) = 6. The eigenvalues are the zeros of z^2 - 5z + 6: 2 and 3.
- The eigenvalues with multiplicity are 3, 3, -2, i, -i. \operatorname{tr} T = 4; \det T = 9 \cdot (-2) \cdot (i)(-i) = -18; \operatorname{tr} T^2 = 9 + 9 + 4 - 1 - 1 = 20. With q(z) = z - 1, \det(T - I) = 2^2(-3)(i - 1)(-i - 1), which is 4 \cdot (-3) \cdot 2, that is, -24.
- L^2 = \begin{bmatrix} 0 & 16 & 0 \\ 0 & 0 & 16 \\ \tfrac14 & 0 & 0 \end{bmatrix}, so \operatorname{tr} L = \operatorname{tr} L^2 = 0, and \det L = 32 \cdot \tfrac14 = 8. As in Example 4 the eigenvalues are the zeros of z^3 - 8: 2, 2\omega and 2\bar{\omega}, with \displaystyle \omega = -\tfrac12 + \tfrac{\sqrt{3}}{2}i. They are distinct, so L is diagonalizable and L^3 = 8I_3. Then L^6 = 64I_3, and after six periods there are 640 adults (and no eggs or larvae).
- List the eigenvalues of A with multiplicity. By Result 4, \det A = r_1\cdots r_s \cdot c, where r_1, \dots, r_s are the real ones and c > 0 comes from the conjugate pairs. So r_1\cdots r_s = \det A/c < 0. An empty product is 1, so s \ge 1, and a product of nonnegative numbers is nonnegative, so some r_i < 0. As \det(A - r_iI_n) = 0 and A - r_iI_n is real, it has a nonzero real null vector (The invertible matrix theorem), an eigenvector of A in \mathbb{R}^n for r_i.
- By Result 2(c), (\det T)^2 = \det T^2 = \det(-I) = (-1)^n, the determinant of the diagonal matrix -I_n. Since \det T is real, (\det T)^2 \ge 0, so (-1)^n = 1 and n is even. Over \mathbb{C}, T = iI on \mathbb{C}^n has T^2 = -I for every n.
- BA = A^{-1}(AB)A, so BA - \lambda I = A^{-1}(AB - \lambda I)A for every scalar \lambda. As \det A^{-1}\det A = 1, the product rule gives \det(BA - \lambda I) = \det(AB - \lambda I), so p_{AB} = p_{BA}. Over \mathbb{C}, eigenvalues listed with multiplicity are the zeros of p with multiplicity, so they agree.
- By Result 4 of The minimal polynomial, T^{-1} = q(T) for a polynomial q. Let T\mathbf{v} = \lambda_j\mathbf{v} with \mathbf{v} \neq \mathbf{0}; \lambda_j \neq 0 as T is invertible. Applying T^{-1} gives T^{-1}\mathbf{v} = \lambda_j^{-1}\mathbf{v}, while T^k\mathbf{v} = \lambda_j^k\mathbf{v} gives q(T)\mathbf{v} = q(\lambda_j)\mathbf{v}. So q(\lambda_j) = 1/\lambda_j for every j, and Result 6 gives \operatorname{tr} T^{-1} = \sum_j d_j/\lambda_j and \det T^{-1} = \prod_j \lambda_j^{-d_j}.
- Let A = \begin{bmatrix} a & b \\ c & d \end{bmatrix}. Then A^2 = \begin{bmatrix} a^2 + bc & ab + bd \\ ac + cd & bc + d^2 \end{bmatrix} and (a + d)A = \begin{bmatrix} a^2 + ad & ab + bd \\ ac + cd & ad + d^2 \end{bmatrix}, using that multiplication in a field is commutative. Their difference is (bc - ad)I_2 = -(\det A)I_2, so A^2 - (\operatorname{tr} A)A + (\det A)I_2 = 0.