IX Similarity

42. Similarity Classes

As already has been defined in Section 38: An (n-n)-matrix is similar to an (n,n)=matrix B if there exists a regular matrix P such that A = P-1BP. Similarity has the properties:

1. Reflexivity: Every matrix is similar to itself, for always E-1AE = A.
2.
Symmetry: If A is similar to B, then also B is similar to A, for it follows from A = P-1BP that B = P1-1AP1 with P1 = P-1.
3.
Transitivity: If A is similar to B and B is similar to C, then A is also similar to C, because A = P-1BP and B = Q-1BQ yields A = (QP)-1C(QP).

As a consequence of these properties, you obtain as you join mutually similar matrices in a class a subdivision of all (n,n)-Matrices into element free partial sets - classes of similarity.

A quantity which has for all similar matrices the same value is called an invariant of similarity. For example, according to the Rank Equation in Section 24, Rank is an invariant of similarity. Theorem 22 states that also the characteristic roots, the coefficients of the characteristic polynomial and, in particular, the determinant are similarity invariants. We will study the task to find a complete system of invariants of similarity, that is such a system that two matrices are exactly similar if their invariants agree. The fact that Rank and characteristic roots do not yet form a complete system of invariants is demonstrated by the two matrices

Both of them have the same characteristic roots and the same Rank, but are not similar, because due to P-1EP = E, only the unit matrix is similar to itself.

Let A and B be two matrices, both of which decompose in the same manner into diagonal minors

Hence, if corresponding diagonal components are similar, that is,

Ai = Pi-1BiPi (i = 1, 2, ··· , r)

then also A and B are similar, because with

you find A = P-1BP.

When you confine your considerations to matrices with elements out of a fixed body of numbers, the following observation is useful: If A and B are matrices with elements out of K which altogether are similar, there also always exists a regular matrix P with elements of K for which A = P-1BP, because this equation can also be written in the form PA = BP with the auxiliary condition
|P| 0. This shows that the determination of P results in the solution of a system of linear equations. However, every system of linear equations has already a complete system of solutions in every body which contains the coefficients.

last next