Page 303 - DMTH502_LINEAR_ALGEBRA
P. 303

Unit 28: Positive Forms and More on Forms




                                                                                                Notes
                                       =  (   x x k  ( f  j ,  k )
                                               j
                                          j  k
                                       =  (   A x x k )                            ...(1)
                                               kj j
                                          j  k
          So we see that f is non-negative if and only if
          and
                                     A = A*

               A x x   0  for all scalars x , x , ... x                            ..(2)
                kj j  k              1  2  n
           j  k
          For positive f, the relation should be true for all (x , x , ... x )   0. The above conditions on positive
                                                 1  2   n
          f form are true if
                                    g(X, Y) = Y*AX                                 ...(3)
          is a positive form on the space of n × 1 column matrices over the scalar field.

          Theorem 1: Let F be the field of real number or the field of complex numbers. Let A be an n × n
          matrix over F. The function g defined by
                                    g(X, Y) = Y*AX                                 ...(4)

                                     n×1
          is a positive form on the space F  if and only if there exists an invertible n × n matrix P with
          entries in F such that A = P*P.
          Proof: For any n × n matrix A, the function g in (4) is a form on the space of column matrices. We
          are trying to prove that g is positive if and only if A = P*P. First, suppose A = P*P. Then g is
          Hermitian and
                                g(X, X) = X*P*PX
                                       = (PX)*PX
                                            0.
          If P is invertible and X   0, then (PX)*PX > 0.

          Now, suppose that g is a positive form on the space of column matrices. Then it is an inner
          product and hence there exist column matrices Q , ..., Q  such that
                                                 1     n
                                       = g(Q , Q )
                                     jk    1  k
                                       = Q AQ .
                                          *
                                          k  j
          But this just says that, if Q is the matrix with columns Q , ..., Q , then A*AQ = I. Since {Q , ..., Q }
                                                       1    n                   1    n
                                      –1
          is a basis, Q is invertible. Let P = Q  and we have A = P*P.
          In practice, it is not easy to verify that a given matrix A satisfies the criteria for positivity which
          we have given thus far. One consequence of the last theorem is that if  g is positive then det
          A > 0, because det A = det (P*P) = det P* det P  = |det P| . The fact that det A > 0 is by no means
                                                       2
          sufficient to guarantee that g is positive; however, there are n determinants associated with A
          which have this property: If A = A* and if each of those determinants is positive, then  g is a
          positive form.
          Definition: Let A be an n × n matrix over field F. The principal minors of A are the scalars  (A)
                                                                                    k
          defined by






                                           LOVELY PROFESSIONAL UNIVERSITY                                   297
   298   299   300   301   302   303   304   305   306   307   308