EXAM Exam 1 Math 5316, Fall 2012 December 2, 2012 • Write all of your answers on separate sheets of paper. You can keep the exam questions. This is a takehome exam, to be worked individually. You can use your notes. You may use Maple as indicated and turn in hardcopy Maple output with your written answers. There is a Maple worksheet on my website with the matrices already entered. • The exam is due Saturday, December 8 by 5pm. • You must show enough work to justify your answers. Unless otherwise instructed, give exact answers, not √ approximations (e.g., 2, not 1.414). • This exam has 8 problems. There are 370 points total. Good luck! 60 pts. Problem 1. In each part, find a matrix P so that P −1 AP is in Jordan Canonical Form. I would not undertake this by hand. There is a maple worksheet with the matrices already entered. A. 1400 1570 561 631 A= 294 330 495 555 −9793 −3927 −2057 −3465 86 34 18 32 B. A= 40 pts. −4899 3100 1110 1765 −14269 −16920 −36857 29120 72 48 −124 1222 −58 20 −236 −106 276 −543 −790 −203 115 195 22 −358 275 −265 294 −150 34 −9 −216 144 −435 182 −90 77 261 −432 −40 3575 294 −180 −713 0 828 −3405 4155 −688 345 −600 −51 2556 −7499 9033 −2116 1067 −1297 −171 −3096 2384 −7185 2619 −1290 1272 Problem 2. Find a matrix P so that P −1 AP = J, where J is in Jordan Canonical Form. Find a real matrix S so that S −1 AS = R, where R is in real Jordan form. Again, not a problem for hand computation! See the Maple worksheet. A= −49 −69 −501 −33 −2283 −489 −1686 −1971 5092766 −372123 1084 158927 5091364 −1697132 −70821 5288 151 −2223 −71006 23652 7754744 −564518 2197 241674 7752164 −2584179 1254085 −91603 318 39132 1253677 −417900 38991800 −2838877 10495 1215215 38979500 −12993735 3376128 −245787 1759 105244 3373954 −1124769 11189119 −814529 6002 348795 11181691 −3727636 14532072 −1051262 4913 451845 14527028 −4842842 1 −79220 1094 −120712 −19510 −606928 −52571 −174236 −226453 40 pts. 40 pts. Problem 3. Again, not a problem for hand computation. The following matrix A is diagonalizable, so we have P −1 AP = D, where D is diagonal. Find a matrix S so that At = S −1 AS. 14 23 −73 −39 117 80 −273 −351 . A= 39 26 −89 −117 0 5 −15 1 Problem 4. Again, not a problem for hand computation. You can use the Maple GramSchmidt command to check yourself. Read the help page for that command first. A Use the Gram-Schmidt process on the vectors v1 , v2 , v3 , v4 to find an orthonormal basis of R4 . Show the Gram-Schmidt process one step at a time. Find an orthogonal matrix Q and an upper triangular matrix R so that A = QR. 1 1 v1 = 0 , −1 1 1 A= 0 −1 0 1 v2 = 0 , 1 0 1 0 1 0 0 , 1 1 1 1 v3 = 1 , 1 1 1 1 1 0 0 v4 = 1 . 1 B Use the Gram-Schmidt process on the vectors v1 , v2 , v3 , v4 to find an orthonormal basis of C4 . Show the Gram-Schmidt process one step at a time. Find an unitary matrix U and an upper triangular matrix R so that A = U R. i 0 1 i i 0 0 i A= 0 1 1 −2 i 1 1 + 2i 1 1 i 0 1 i i 0 i 0 , v1 = v2 = v3 = v4 = 0 , 1 , −2i 1 1 1 + 2i 1 1 2 50 pts. 40 pts. Problem 5. Again, not a problem for hand computation. You can use the Maple GramSchmidt command to check yourself. Read the help page for that command first. The following matrix A is symmetric. Find an orthonormal basis for R5 consisting of eigenvectors of A. Find an orthogonal matrix Q so that Qt AQ is diagonal. −5 0 0 −10 5 0 20 0 0 0 0 0 −10 0 0 A= −10 0 0 10 −10 5 0 0 −10 −5 Problem 6. Let A be an m × n matrix over K. Multiplication by A is a linear transformation Kn → Km and multiplication by A∗ is linear transformation Km → Kn . We use the standard scalar product on Kn and Km . Show that Km = im(A) ⊕ ker(A∗ ) Kn = im(A∗ ) ⊕ ker(A), and that these are orthogonal direct sums, e.g., the subspaces im(A) and ker(A∗ ) are orthogonal. Hint: Do the first equation first, suppose that v ∈ im(A)⊥ , see if you can get it in ker(A∗ ). 50 pts. Problem 7. Let V be a vector space over K with a K-inner product. Recall that a linear transformation V → V is symmetric if hT (u), V i = hu, T (v)i, ∀u, v ∈ V. Recall that an operator P : V → V is a projection operator if P 2 = P (which is the same as saying P acts as the identity on im(P )). We know from previous work that in this case V = im(P ) ⊕ ker(P ). Show that this is an orthogonal direct sum if and only if P is symmetric. A word on Matrix Norms Let V be a vector space over K. Recall that a norm on V is a function V → K : v 7→ kvk that has the following properties. 3 1. kvk ≥ 0 for all v ∈ V and kvk = 0 ⇐⇒ v = 0. 2. kλvk = |λ|kvk for all λ ∈ K and v ∈ V . 3. The triangle inequality holds: for all v, w ∈ V , kv + wk ≤ kvk + kwk. In the following problem it is convenient to use the following norm on Kd : kxk = max{|xi | | i = 1, 2, . . . , d}, where x1 x2 x = . ∈ Kd . .. xd On the vector space Kd×d of d × d matrices, we use the following norm. If A = [aij ], then X d |aij | | i = 1, 2, . . . , d kAk = max j=1 In other words, for each row we form the sum of the absolute values of all the entries in the row, and then we take the largest of these row sums. It’s easy to see that if kAk is small, the entries in A must be small. Indeed, we can form the norm of Kd×d defined by kAkmax = max{|ai,j | | i, j = 1, 2, . . . , d}, where we take the maximum size of an entry. It’s easy to check that kAkmax ≤ kAk ≤ dkAkmax , so if one of these two norms is small, the other is forced to be small. The advantage of the norm k·k is the following two inequalities kAxk ≤ kAk kxk, x ∈ Kd , A ∈ Kd×d , kABk ≤ kAk kBk, A, B ∈ Kd×d . 4 50 pts. Problem 8. Let A be a d × d matrix over C. Let λ1 , λ2 , . . . , λk be the eigenvalues of A. Strictly for the purposes of this problem, we say that the matrix A is stable if |λj | < 1 for j = 1, 2, . . . , k, i.e., all the eigenvalues have modulus strictly less that one. In this problem, we’re working over C. A. Show that if D is a stable d × d diagonal matrix, then kDn k → 0, n → ∞. B. If S is a stable d × d matrix which is diagonalizable, then kS n k → 0, n → ∞. C. Let A be a stable d × d matrix which is not diagonalizable. Show that kAn k → 0, n → ∞. To do this, use the Jordan Decomposition A = S + N . Since S and N commute, you can compute (S + N )n by the Binomial Theorem. 5