If A and B are similar matrices, then show that A and B have the same rank, trace, characteristic polynomial and eigenvalues.
10 marks · directive: Show
The core of a good answer
If B = P⁻¹AP with P invertible, then multiplication by an invertible matrix does not change rank, tr(P⁻¹AP) = tr(APP⁻¹) = tr A, and det(B − λI) = det(P⁻¹(A − λI)P) = det(A − λI); equal characteristic polynomials give equal eigenvalues with the same algebraic multiplicities.
An outline written by upsc.help, not an official UPSC answer; every fact checked by a separate review. How we work
What the question asks
A proof question. Start from the definition of similarity and prove each of the four invariants in turn, using only standard facts: rank is unchanged by multiplication by an invertible matrix, tr(XY) = tr(YX), and det is multiplicative. Each step should be one or two lines with the reason stated.
Answer outline
Introduction
Definition: n × n matrices A and B over a field F are similar if there is an invertible n × n matrix P with B = P⁻¹AP. Throughout, I is the identity and λ a scalar.
Step 1: Rank
- Lemma: if Q is invertible then rank(QX) = rank X and rank(XQ) = rank X.
- Reason: X ↦ QX is injective on column vectors, so QXv = 0 iff Xv = 0; hence null(QX) = null(X) and by rank-nullity rank(QX) = rank X. For XQ, the column space of XQ equals the column space of X because Q is onto.
- So rank B = rank(P⁻¹(AP)) = rank(AP) = rank A.
Step 2: Trace
- For any n × n X, Y: tr(XY) = Σᵢ Σⱼ xᵢⱼ yⱼᵢ = Σⱼ Σᵢ yⱼᵢ xᵢⱼ = tr(YX).
- tr B = tr(P⁻¹(AP)) = tr((AP)P⁻¹) = tr(A(PP⁻¹)) = tr A.
Step 3: Characteristic polynomial
- B − λI = P⁻¹AP − λP⁻¹IP = P⁻¹(A − λI)P.
- det(B − λI) = det(P⁻¹) det(A − λI) det(P) = det(A − λI), because det(P⁻¹) det(P) = det(I) = 1.
- So χ_B(λ) = χ_A(λ) as polynomials in λ.
Step 4: Eigenvalues
- Eigenvalues are the roots of the characteristic polynomial, so A and B have the same eigenvalues with the same algebraic multiplicities.
- Direct link to eigenvectors: if Av = λv, then B(P⁻¹v) = P⁻¹APP⁻¹v = P⁻¹Av = λ(P⁻¹v), and P⁻¹v ≠ 0; so P⁻¹ maps the λ-eigenspace of A onto that of B and geometric multiplicities also agree.
Step 5: Corollaries worth adding
- Determinant: det B = det A (constant term of χ, or directly).
- The converse fails: N = [[1, 1], [0, 1]] and I both have rank 2, trace 2 and χ(λ) = (λ − 1)², yet they are not similar, since P⁻¹IP = I ≠ N for every invertible P.
Conclusion
Rank, trace, determinant, characteristic polynomial and eigenvalues are similarity invariants, which is why they can be attached to a linear operator independently of the basis chosen; equality of these invariants does not by itself prove similarity.
Examples and data to cite
- A = [[1, 1], [0, 1]] and B = I (2 × 2) share rank 2, trace 2 and characteristic polynomial (λ − 1)², but are not similar: this shows the invariants are necessary, not sufficient.
- Change of basis: if [T]_β = A and P is the change-of-basis matrix, then [T]_γ = P⁻¹AP, so the trace and eigenvalues of a linear operator T are well defined.
Facts to remember
- B = P⁻¹AP with P invertible defines similarity; it is an equivalence relation.
- tr(XY) = tr(YX) for square matrices of the same order.
- det(XY) = det X · det Y and det(P⁻¹) = 1/det P.
- Multiplying by an invertible matrix preserves rank.
- Similar matrices also have the same determinant and the same minimal polynomial.