Dr. Marques Sophie Office 519 Linear algebra Spring Semester 2015 marques@cims.nyu.edu Problem Set # 10 The following equations are considered over the reals numbers. All the answers should be justified unless mentioned differently. Problem 1 : Let λ be an eigenvalue of an invertible matrix A. Show that λ´1 is an eigenvalue of A´1 . Solution : Let λ be an eigenvalue of an invertible matrix A. Then there is a non zero vector x such that Ax “ λx So that, multiplying the equality by A´1 , x “ A´1 λx “ λA´1 x And multiplying the equality by λ´1 , we get A´1 x “ λ´1 x with x ‰ 0. That means that λ´1 is a eigenvalue for A´1 . Problem 2 : Show that if A2 is the zero matrix, then the only eigenvalue of A is 0. Solution : If A2 is the zero matrix, then detpA2 q “ detpAqdetpAq “ 0, so that detpAq “ 0. By the invertible matrix theorem, we have that that 0 is an eigenvalue for A. Now we prove that 0 is the only eigenvalue. Indeed, suppose that lambda is an eigenvalue for A, then there is a nonzero vector x such that Ax “ λx Then A2 x “ λAx So that A2 x “ λ2 x Since A2 “ 0, then λ2 x “ 0. And since x ‰ 0, this implies that λ2 “ 0, thus λ “ 0. Finally, we have proven that λ “ 0 is the only eigenvalue. Problem 3 : Show that λ is an eigenvalue of A if and only if λ is an eigenvalue of AT . 1 Solution : We know that λ is an eigenvalue for A if and only if λ is a root of the characteristic polynomial. We will prove that the characteristic polynomial of A equals the one of AT that will prove that A and AT have the same eigenvalues and this will solve the problem. We have prove in class that detpMq “ detpMT q for any matrix M, thus detpA ´ λIq “ detppA ´ λIqT q “ detpAT ´ λIq since pλIqT “ λI. So, we have proven that the characteristic polynomial of A detpA ´ λIq is equal to the characteristic polynomial of AT which is detpAT ´ λIq. Problemˆ4 : ˙ ˆ ˙ ˆ ˙ 0.6 0.3 3{7 0.5 Let A “ , v1 “ and x0 “ . 0.4 0.7 4{7 0.5 1. Find a basis for R2 , consisting of v1 and another eigenvector v2 of A. 2. Verify that x0 may be written in the form x0 “ v1 ` cv2 . 3. For k “ 1, 2, ¨ ¨ ¨ , define xk “ Ak x0 . Compute x1 and x2 , and write a formula for xk . Then show that xk ÞÑ v1 as k increases. Solution : 1. Since A is a 2 ˆ 2 matrix, the eigenvalues are easy to find, and factorizing the characteristic polynomial is easy when one of the two factors is known. ˆ ˙ 0.6 ´ λ 0.3 detp q “ p0.6´λqp0.7´λq´0.3¨0.4 “ λ2 ´1.3λ`0.3 “ pλ´1qpλ´3q 0.4 0.7 ´ λ The eigenvalues are 1 and 0.3. For the eigenvalue 0.3, solve pA ´ 0.3Iqx “ 0 : ˆ ˙ ˆ ˙ 0.6 ´ 0.3 0.3 0 0.3 0.3 0 “ 0.4 0.7 ´ 0.3 0 0.4 0.4 0 Here, x1 ` x2 “ 0, with xˆ The general solution is not needed. Set x2 “ 1 to 2 free. ˙ ´1 find an eigenvector v2 “ . A suitable basis for R2 is tv1 , v2 u. 1 ˆ ˙ ˆ ˙ ˆ ˙ 1{2 3{7 ´1 2. Write x0 “ v1 ` 2v2 : “ `c . 1{2 3{7 1 By inspection, c is ´1{14. (The value of c depends on how v2 is scaled.) 3. For k “ 1, 2, ¨ ¨ ¨ , define xk`1 “ Ak x0 . Then x1 “ Apv1 ` cv2 q “ Av1 ` cAv2 “ v1 ` cp0.3qv1 , because v1 and v2 are eigenvector. Again, x2 “ Ax1 “ Apv1 ` cp0.3qv2 q “ Av1 ` cAv2 “ v1 ` cp0.3qv2 Continuing, the general pattern is xk “ v1 ` cp0.3qk v2 , As k increase, the second term tends to 0 and so xk tends to v1 . Problem 5 : Diagonalize the matrices, if possible. 2 1. ¨ ˛ 2 0 ´2 A“˝ 1 3 2 ‚ 0 0 3 2. ¨ ˛ 1 2 ´3 B “ ˝ 2 5 ´2 ‚ 1 3 1 Solution : 1. ˛ 2´λ 0 ´2 3´λ 2 ‚ A ´ λI “ ˝ 1 0 0 3´λ ¨ The characteristic equation is ˇ ˇ 2´λ 0 ´2 ˇ ˇ 3´λ 2 detpA´λIq “ ˇ 1 ˇ 0 0 3´λ So, the eigenvalues ¨ of A 0 ˝ 1 For λ “ 2, A´2I “ 0 ˇ ˇ ˇ ˇ ˇ 0 ˇ “ p3´λq ˇ 2 ´ λ ˇ ˇ 1 3 ´ λ ˇ ˇ ˇ ˇ “ p3´λqp2´λqp3´λq “ p3´λq2 p2´ ˇ given ˛ ¨ ˛ to be 2 and 3. 1 1 0 0 ´2 2 ‚and row reducing rA´2I, 0s yields ˝ 0 0 1 0 ‚. 0 0 0 0 1 ˛ ¨ ˛ ´1 ´1 The general solution is x2 ˝ 1 ‚ and a basis for the eigenspace is v1 “ ˝ 1 ‚. 0 0 ˛ ¨ 2´λ 0 ´2 ˝ 1 3´λ 2 ‚ and row reducing rA ´ 3I, 0s yields For λ “ 3, A ´ 3λ “ 0 0 3´λ ¨ ˛ 1 0 2 0 ˝ 0 0 0 0 ‚. 0 0 0 0 ¨ ˛ ¨ ˛ 0 ´2 The general solution is x2 ˝ 1 ‚ ` x3 ˝ 0 ‚, and nice for the eigenspace is 0 1 ¨ ˛ ¨ ˛ 0 ´2 tv1 , v2 u “ t˝ 1 ‚, ˝ 0 ‚u. From v1 , v2 and v3 , construct 0 1 ¨ ˛ ´1 0 ´2 P “ rv1 , v2 , v3 s “ ˝ 1 1 0 ‚ 0 0 1 are 0 1 0 ¨ 3 Then set ¨ ˛ 2 0 0 D“˝ 0 3 0 ‚ 0 0 3 where the eigenvalues in D correspond to v1 , v2 and v3 , respectively. 2. ˛ 1´λ 2 ´3 5 ´ λ ´2 ‚ B ´ λI “ ˝ 2 1 3 1´λ ¨ Then the characteristic equation ˇ ˇ ˇ ˇ ˇ ˇ ˇ 5 ´ λ ´2 ˇ ˇ 2 ´2 ˇ ˇ 2 5´λ ˇ ˇ ´ 2ˇ ˇ ˇ ˇ detpB ´ λIq “ p1 ´ λq ˇˇ ˇ 1 1 ´ λ ˇ ´ 3ˇ 1 3 1´λ ˇ 3 ˇ “ p1 ´ λqpp5 ´ λqp1 ´ λq ` 6q ´ 2p2p1 ´ λq ` 2q ´ 3p6 ´ p5 ´ λqq “ 6 ´ 6λ ` p1 ´ λqp5 ´ 6λ ` λ2 q ´ 4 ´ 4 ` 2λ ´ 18 ` 15 ´ 9λ “ 6 ´ 6λ ` 5 ´ 6λ ` λ2 ´ 5λ ` 6λ2 ´ λ3 ´ 4 ´ 4 ` 2λ ´ 18 ` 15 ´ 9λ “ 7λ2 ´ λ3 ´ 24λ “ λp´λ2 ` 7λ ´ 24q ∆ “ 49 ´ 4 ¨ 24 ă 0 2 So the polynomial ´λ ` 7λ ´ 24 does not factorize into linear coefficient over R. Thus we know that B is not diagonalizable. Problem 6 : ˛ ¨ pp´1q Define T : P2 Ñ R3 by Tppq “ ˝ pp0q ‚ pp1q 1. Find the image under T of pptq “ 5 ` 3t. 2. Show that T is a linear transformation. 3. Find the matrix for T relative to the basis t1, t, t2 u for P2 and the standard basis for R3 . Solution : 1. The image of p under T is ¨ ˛ ¨ ˛ pp´1q 2 Tppq “ ˝ pp0q ‚ “ ˝ 5 ‚ pp1q 8 2. Let p and q be polynomials in P2 and let c be any scalar. Then, ¨ ˛ ¨ ˛ ¨ ˛ ¨ ˛ pp ` qqp´1q pp´1q ` qp´1q pp´1q qp´1q Tpp`qq “ ˝ pp ` qqp0q ‚ “ ˝ pp0q ` qp0q ‚ “ ˝ pp0q ‚`˝ qp0q ‚ “ Tppq`Tpqq pp ` qqp1q pp1q ` qp1q pp1q qp1q ¨ ˛ ¨ ˛ ¨ ˛ pcpqp´1q c ¨ pp´1q pp´1q Tpcpq “ ˝ pcppq0q ‚ “ ˝ c ¨ pp0q ‚ “ c ˝ pp0q ‚ “ c ¨ Tppq pcpqp1q c ¨ pp1q pp1q 4 3 3. Let B “ t1, t, t2 u and ˛ te1 , e2 , e3 u be the standard basis¨for R ˛. Since rTpb1 qs “ ¨ “ ´1 1 Tpb1 q “ Tp1q “ ˝ 1 ‚, rTpb2 qs “ Tpb2 q “ Tptq “ ˝ 0 ‚ and rTpb3 qs “ 1, 1, ˛ ¨ 1 Tpb3 q “ Tpt2 q “ ˝ 0 ‚, the matrix for T relative to B and is 1, ˛ ¨ 1 ´1 1 rrTpb1 qs , rTpb2 qs , rTpb3 qs s “ ˝ 1 0 0 ‚ 1 1 1 Problem 7 : Find the B-matrix for the transformation x ÞÑ Ax where B “ tb1 , b2 u with ˆ ˙ ˆ ˙ ˆ ˙ ´4 ´1 ´1 ´1 A“ , b1 “ , b2 “ 6 1 2 1 ˆ ˙ ´1 ´1 Solution : If P “ rb1 , b2 s “ , then the B- matrix is 2 1 ˆ ˙ˆ ˙ˆ ˙ ˆ ˙ 1 1 ´4 ´1 ´1 ´1 ´2 ´2 ´1 P AP “ “ ´2 ´1 6 1 2 1 0 ´1 Problem 8 : Let A ˆbe 2˙ˆ 2 matrix with eigenvalues 3 and 1{3 and corresponding eigenvectors ˆ ˙ 1 ´1 v1 “ and v2 “ . Let txk u be a solution of the difference equation 1 1 ˆ ˙ 9 xk`1 “ Axk , x0 “ . 1 1. Compute x1 “ Ax0 . 2. Find a formula for xk involving k and the eigenvectors v1 and v2 . Solution : 1. To find the action of A on x0 , express x0 in terms of v1 and v2 . That is, find c1 and c2 such that x0 “ c1 v1 ` c2 v2 . This is certainly possible because the eigenvectors v1 and v2 are linearly independent (by inspection and also because they correspond to distinct eigenvalues) and hence form a basis for R2 . (Two linearly independent vectors in R2 automatically span R2 .) . The row reduction ˆ ˙ ˆ ˙ 1 ´1 9 1 0 5 rv1 , v2 , v3 s “ „ 1 1 1 0 1 ´4 shows that x0 “ 5v1 ´ 4v2 . Since v1 and v2 are eigenvectors (for eigenvalues 3 and 1{3). ˆ ˙ 49{3 x1 “ Ax0 “ 5Av1 ´ 4Av2 “ 15v1 ´ 4{3v2 “ 41{3 5 2. Each time A acts on linear combination of v1 and v2 , the v1 term is multiplied by the eigenvalue 3 and the v2 term is multiplied by the eigenvalue 3 and the v2 term is multiplied by the eigenvalue 1{3 : x2 “ Ax1 “ Ar15v1 ´ 4{3v2 s “ 5p3q2 v1 ´ 4p1{3q2 v2 In general, xk “ 5p3qk v1 ´ 4p1{3qk v2 ,for k ě 0. Problem 9 : Classify the origin as an attractor, repeller, or saddle point of the dynamical system xk`1 “ Axk . Find the directions of greatest attraction and/or repulsion. ˆ ˙ 1.7 ´0.3 A“ ´1.2 0.8 Solution : detpA ´ λIq “ λ2 ´ 2.5λ ` 1 “ 0 Then ? 2.52 ´ 4 λ“ “ 2 and 0.5 2 The origin is a saddle point because one eigenvalue is greater than 1 and the other eigenvalue is less than 1 in magnitude. The direction of greatest repulsion is through the origin and the eigenvector v1 found below. Solve pA ´ 2Iqx “ 0 ˆ ˙ ˆ ˙ ´0.3 ´0.3 0 1 1 0 „ ´1.2 ´1.2 0 0 0 0 ˆ ˙ ´1 so x1 “ ´x2 and x2 is free. Take v1 “ . The direction of greatest attraction is 1 through the origin and the eigenvector v2 below. Solve pA ´ 0.5Iqx “ 0 ˆ ˙ ˆ ˙ 1.2 ´0.3 0 1 ´0.25 0 „ ´1.2 0.3 0 0 0 0 ˆ ˙ 1 So x1 “ 0.25x2 and x2 is free. Take v2 “ . 4 2.5 ˘ 6