Page 65 - Matrices theory and applications
P. 65

3. Matrices with Real or Complex Entries
                              48
                              block-diagonal matrix whose diagonal blocks are direct similitudes. Hence
                              M 1 and M are unitarily similar to such a matrix.
                              Corollary 3.3.1 Real symmetric matrices are diagonalizable over IR,
                              through orthogonal conjugation. In other words, given M ∈ Sym (IR),
                                                                                         n
                                                                    −1
                                                                       is diagonal.
                              there exists an O ∈ O n (IR) such that OMO
                              In fact, since the eigenvalues of M are real, OMO −1  has only 1 × 1blocks.
                              We say that real symmetric matrices are orthogonally diagonalizable.
                                The interpretation of this statement in terms of quadratic forms is the
                                                                      n
                              following. For every quadratic form Q on IR , there exists an orthonor-
                              mal basis {e 1,... ,e n} in which this form can be written with at most n
                              squares: 4
                                                               n
                                                                    2
                                                       Q(x)=     a i x .
                                                                    i
                                                              i=1
                                                              1/2
                              Replacing the basis vector e j by |a j |  e j , one sees that there also exists
                              an orthogonal basis in which the quadratic form can be written
                                                           r       s
                                                              2       2
                                                   Q(x)=     x −     x   ,
                                                              i       j+r
                                                          i=1     j=1
                              with r+s ≤ n. This quadratic form is nondegenerate if and only if r+s = n.
                              The pair (r, s) is unique and called the signature or the Sylvester index of
                              the quadratic form. In such a basis, the matrix associated to Q is
                                              1
                                                                              
                                                 .
                                                .                             
                                                 .                      0     
                                                                              
                                                     1
                                                                              
                                                                              
                                                        −1
                                                                              
                                                                              
                                                             .
                                                            .                  .
                                                                               
                                                             .
                                                                              
                                                                −1            
                                                                              
                                                                    0         
                                                                              
                                                                        .
                                                                        .     
                                                0                        .    
                                                                             0
                              3.3.1 Rayleigh Quotients
                              Let M be a real n × n symmetric matrix, and let λ 1 ≤ ··· ≤ λ n be its
                              eigenvalues arranged in increasing order and counted with multiplicity. Let
                                4
                                 In solid mechanics, when Q is the matrix of inertia, the vectors of this basis are
                              along the inertia axes,and the a j , which then are positive, are the momenta of inertia.
   60   61   62   63   64   65   66   67   68   69   70