1 Homework Solutions 18.335 - Fall 2004 1.1 jdet(A)j = 1. Show det(B), then A + B is Let A be an orthogonal matrix. Prove that that if B is also orthogonal and det(A) = singular. (det A)2 = det A det A = det A det AT = det AAT = det I = 1 A + B is singular i¤ AT (A + B) = I + AT B is. AT B is orthogonal so all its eigenvalues are 1 or -1. Since their product is equal to det AT B = 1 then at least one of the eigenvalues of AT B must be -1. Let the corresponding vector be x. Then (I + AT B)x = x x = 0, so I + AT B is singular and so is A + B. Second proof: det(A + B) = det(AT ) det(A + B) det(B T ) = det(AT AB T + AT BB T ) = det(AT + B T ) = det(A + B), so det(A + B) = 0. 1.2 Trefethen 2.5 (a) Let be an eigenvalue of S and v its corresponding eigenvector so that 2 Sv = v ) v Sv = v v = kvk : We also have v Sv = v S v = v Sv: This implies that = ) is imaginary. (b) If (I S) v = 0 for v 6= 0 then Sv = v and this means that 1 is an eigenvalue of S, a contradiction to (a). (c) We have: Q Q = h (I 1 S) = (I + S ) (I = (I i (I + S) (I S ) S) (I + S) = (I + S) 1 (I 1 1 (I S) (I (I 1 S) S) 1 (I + S) (I + S) S) 1 (I + S) S) 1 (I + S) = I where we have used that if AB = BA and B is invertible that AB B 1A 1.3 1 = Trefethen 3.2 kAxk : Choose an eigenvalue of A and let x 6= 0 x6=0 kxk kAx k k x k j j kx k such that Ax = x : Then = = = : Thus we have kx k kx k kx k kAxk kAk = sup j j : So kAk j j and since this is true for any eigenvalue x6=0 kxk of A we get kAk sup fj j ; eigenvalue of Ag = (A) : We know that kAk = sup 1 1.4 Trefethen 3.3 s (a) By de…nition kxk1 = max jxi j 1 i m m P j=1 x2j = kxk2 : Equality is achieved when we have a vector with only one non-zero component. s m q P p (b) Again, using the de…nition kxk2 = x2j m max jxi j = m kxk1 : 1 i m j=1 We have equality for a vector whose components are equal to each other. (c) Denoting by rj the j-th row of A we have kAk1 = max krj k1 : For some 1 j m p vector v 2 Cn , v = (1; :::; 1) = n andsusing the 2-norm de…nition we m P 1 2 get kAk2 = sup kAxk2 kAvk2 = p krj k1 : These yield kAk1 = n j=1 jxj=1 s m P p 2 max krj k1 krj k1 n kAk2 : Equality is achieved for a matrix 1 j m j=1 which is zero everywhere except along a row of ones. s s m m P P 1 2 2 (d) Using the notation from part (c), kAk2 = p krj k1 krj k1 n j=1 j=1 p p m max krj k1 = m kAk1 : We get equality for a square matrix which 1 j m is zero everywhere except along a column of ones. 1.5 Prove that kxy kF = s n P n P j=1 i=1 kxy kF = kxy k2 = kxk2 kyk2 2 jxi yj j = s n P i=1 2 jxi j s n P j=1 for any x and y 2 Cn : 2 jyj j = kxk2 kyk2 kxk2 jy zj kxy zk2 kxy k2 = sup = sup : This ratio is maximized if z==y; so n n kzk kzk2 z2 C z2 C 2 2 that jy zj = kyk2 ; thus completing the proof. 2