Course 111: Algebra, 2 March 2007 1. Apply the orthogonal diagonalisation process to 5 0 6 0 11 6 6 6 −2 The eigenvalues are determined from the characteristic equation: (5−λ) ((11 − λ)(−2 − λ) − 36)+6(−6(11−λ)) = λ3 −14λ2 −49λ+686 By trial and error (enumerating the factors of 686 and finding one that satisfies the polynomial in λ) λ = 7 is a factor. Then by polynomial division you get (λ − 7)(λ2 − 7λ − 98) = λ3 − 14λ2 − 49λ + 686 and λ2 − 7λ − 98 = (λ − 14)(λ + 7) The eigenvalues are −7, 7, 14. The corresponding eigenvectors are solutions, x1 , x2 and x3 , to the matrix-vector equations (A − 7I)x1 = 0, (A + 7I)x2 = 0 and (A − 14I)x3 = 0, where I is the identity matrix. The solutions are for λ = 7 x = (3, 3/2, 1) for λ = −7 x = (−1/2, −1/3, 1) for λ = 14 x = (2/3, 2, 1) Then since the eigenvectors are all distinct the eigenvalues are orthogonal (check that this is true) and to make them orthonormal we just need to divide each one by its norm. ||x1 || = q 32 ||x2 || = q (−1/2)2 + (−1/3)2 + 12 = 7/6, ||x3 || = q (2/3)2 + 22 + 12 = 7/3 + (3/2)2 + 12 = q 23/2, Then A is orthogonalised by the matrix q 3/ 23/2 −1/2/(7/6) 2/3/(7/3) q 3/2/ 23/2 −1/3/(7/6) = 2/(7/3) q 1/ 23/2 1/(7/6) 1/(7/3) and √ 3 2 √ 23 √3 √46 √2 23 − 37 − 27 6 7 2 7 6 7 3 7 7 0 0 P T AP = D = 0 −7 0 0 0 14 2. Suppose that A and B are both orthogonal n × n square matrices. Prove that • A−1 is an orthogonal matrix Proof Since A is orthogonal its column vectors form an orthogonal (in fact they are orthonormal, but we only need orthogonal for this) set of vectors. Now, by the definition of orthogonal matrices, A−1 = AT . But this means that the rows of A−1 are the columns of A and so are an orthogonal set of vectors and therefore (as proved in the notes) A−1 is also an orthogonal matrix. • AB is an orthogonal matrix Hint: the norm preserving property can be useful here Proof: Consider ||ABx|| = ||A(Bx)||, where x ∈ Rn . A is an orthogonal matrix so it preserves norms (proved in lectures). Therefore ||ABx|| = ||A(Bx)|| = ||Bx|| Now, B is also orthogonal and also preserves norms, so ||ABx|| = ||Bx|| = ||x|| Therefore the product AB preserves norms and therefore is itself an orthogonal matrix. • Either det(A) = 1 or det(A) = −1 Proof Since A is orthogonal, AAT = I where I is the identity, n × n matrix. Taking determinants on both sides det(AAT ) = det(I) = 1 (1) Then using properties of the determinant we rewrite the right hand side of Eq.(1) as det(A)det(AT ) = 1 det(A)det(A) = 1 [det(A)]2 = 1 and therefore det(A) = ±1.