# 508 guarantees that every a c mm is unitarily similar

This preview shows page 1. Sign up to view the full content.

This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: nting algebraic multiplicities. In 1852 J. J. Sylvester (p. 80) discovered that the inertia of A is invariant under congruence transformations. Sylvester’s Law of Inertia Let A ∼ B denote the fact that real-symmetric matrices A and B = are congruent (i.e., CT AC = B for some nonsingular C ). Sylvester’s law of inertia states that: A∼B = Copyright c 2000 SIAM if and only if A and B have the same inertia. Buy online from SIAM http://www.ec-securehost.com/SIAM/ot71.html Buy from AMAZON.com 7.6 Positive Deﬁnite Matrices http://www.amazon.com/exec/obidos/ASIN/0898714540 Proof. 77 569 Observe that if An×n is real and symmetric with inertia (p, j, s), then ⎛ A∼⎝ = ⎞ Ip×p ⎠ = E, −Ij ×j (7.6.15) 0s×s It is illegal to print, duplicate, or distribute this material Please report violations to [email protected] because if {λ1 , . . . , λp , −λp+1 , . . . , −λp+j , 0, . . . , 0} are the eigenvalues of A (counting multiplicities) with each λi > 0, there is an orthogonal matrix P such that PT AP = diag (λ1 , . . . , λp , −λp+1 , . . . , −λp+j , 0, . . . , 0) , so C = PD, −1/2 −1/2 where D = diag λ1 , . . . , λp+j , 1, . . . , 1 , is nonsingular and CT AC = E. Let B be a real-symmetric matrix with inertia (q, k, t) so that ⎛ B∼⎝ = D E ⎞ Iq×q ⎠ = F. −Ik×k T H 0t×t If B ∼ A, then F ∼ E (congruence is transitive), so rank (F) = rank (E), and = = hence s = t. To show that p = q, assume to the contrary that p > q, and write F = KT EK for some nonsingular K = Xn×q | Yn×n−q . If M = R (Y) ⊆ n and N = span {e1 , . . . , ep } ⊆ n , then using the formula (4.4.19) for the dimension of a sum (p. 205) yields IG R dim(M ∩ N ) = dim M + dim N − dim(M + N ) = (n − q ) + p − dim(M + N ) > 0. Y P Consequently, there exists a nonzero vector x ∈ M ∩ N . For such a vector, x ∈ M =⇒ x = Yy = K and O C x∈N 0 y =⇒ xT Ex = 0T | yT F =⇒ x = (x1 , . . . , xp , 0, . . . , 0)T 0 y ≤ 0, =⇒ xT Ex > 0, which is impossible. Therefore, we can’t have p > q. A similar a...
View Full Document

{[ snackBarMessage ]}

Ask a homework question - tutors are online