Unless otherwise stated, is a field and every space in a dimension formula is finite-dimensional.
Coordinates describe a vector relative to an ordered basis. They change when the basis changes, while the vector and the linear map do not.
Definition
Matrix of a map
For ordered bases of and of , the matrix has column equal to . Thus . For and an ordered basis of , .
A basis identifies with through its coordinate isomorphism. For bases , the matrix has columns . For an endomorphism, changing basis by gives
This is similarity, with the same basis change in domain and codomain. Independent changes of domain and codomain for a general map produce , a different equivalence problem.
Proof
Follow the coordinates
For every , the old output coordinates are . They also equal . Since the basis-change matrices are invertible, solving for the new output coordinates gives . In the endomorphism case .
Worked example
A nonstandard polynomial basis
On the real polynomials of degree at most one, let and . The derivative of is , and the derivative of is zero. Therefore . Relative to the matrix is ; and indeed is the stated matrix.
Similarity preserves characteristic polynomial, determinant, trace, rank and eigenvalues, since it represents the same endomorphism in another basis. Equal determinant and trace alone do not imply similarity. Row operations need not be similarity transformations; they generally change the represented endomorphism. To choose the direction of a change matrix, write the coordinate equation before multiplying.
Exercises
- Let on and . Find .
- Show similar matrices have the same characteristic polynomial, using .
- Are and similar? Justify.
Full solutions
- has coordinates in . has coordinates . The matrix is .
- If , then . Multiplicativity of determinant gives .
- No. For every invertible , , which is not the second matrix. Their equal characteristic polynomials do not suffice.
Connection
An eigenbasis is a basis in which an endomorphism has diagonal matrix. The next unit asks when such a basis exists.