1.4 Properties of Matrix Operations THM 1 (Properties of Matrix Addition) Let A, B, C and D be m × n matrices. (a) A + B = B + A. (b) A + (B + C) = (A + B) + C. (c) There is a unique m × n matrix O such that A+O =A for any A. The matrix O is called the additive identity or zero matrix. HMHsueh 1 (d) For each A, there is a unique matrix D such that A + D = O. Write D as (−A) and it is called the additive inverse or negative of A. • Note: It can be shown that −A = (−1)A. HMHsueh 2 Proof. (a) Since for 1 ≤ i ≤ m, 1 ≤ j ≤ n, (A + B)ij = aij + bij = bij + aij = (B + A)ij , thus A + B = B + A. (c) Let U = [uij ]m×n which satisfies A + U = A. Since for all 1 ≤ i ≤ m, 1 ≤ j ≤ n, (A + U )ij = aij + uij = aij then uij = 0. Thus, U = [0]m×n. Denote U by O. HMHsueh 3 EX 1 (Ex. 2) The 2 × 3 zero matrix is à O= If à A= 2 3 4 −4 5 −2 0 0 0 0 0 0 ! . à ! then (−A) = −2 −3 −4 4 −5 2 ! , and A + (−A) = O. HMHsueh 4 THM 2 (Properties of Matrix Multiplication) A, B, C are matrices of appropriate sizes. (a) A(BC) = (AB)C. (b) A(B + C) = AB + AC. (c) (A + B)C = AC + BC. HMHsueh 5 Proof. (a) Let A = [aij ]m×n, B = [bij ]n×p, C = [cij ]p×k . Then for 1 ≤ i ≤ m, 1 ≤ j ≤ k, (A(BC))ij = rowi(A) · colj (BC). Since colj (BC) = Bcolj (C), thus (A(BC))ij = rowi(A)Bcolj (C). Similarly, since ((AB)C)ij = rowi(AB) · colj (C), rowi(AB) = rowi(A)B, and ((AB)C)ij = rowi(A)Bcolj (C). Thus A(BC) = (AB)C. HMHsueh 6 EX 2 (Ex. 5) Let à A= 2 2 3 3 −1 2 ! à and à AB + AC = HMHsueh −1 2 0 0 . 2 , C = 1 2 −2 3 −1 1 , B= 2 Then A(B + C) = 2 2 3 3 −1 2 15 1 7 −4 ! ! + à ! ! à 0 2 21 −1 3 2 = 7 −2 5 −3 6 −2 0 2 à = 21 −1 7 −2 ! . 7 DEF 1 The scalar matrix In = 1 0 ··· 0 0 1 ··· 0 ... . . . ... 0 0 ··· 1 is called the identity matrix of order n. • For any Am×n, ImAm×n = Am×nIn = Am×n. • Every scalar matrix Sn = rIn, for some r. HMHsueh 8 • If A is a square matrix, p is a positive integer, then define the powers of a matrix as Ap = AA · · · A. Moreover, define A0 = I. • For nonnegative integers p and q, square matrix A, ApAq = Ap+q , (Ap)q = Apq . • Note that (AB)p 6= ApB p, unless AB = BA. HMHsueh 9 • There exist A, B 6= O but AB = O. EX 3 (Ex. 7) If à A= 1 2 2 4 ! à 4 −6 −2 3 , B= ! , then A, B are not zero matrix, but à AB = HMHsueh 0 0 0 0 ! . 10 • There exist A, B, C such that AB = AC but B 6= C. EX 4 (Ex. 8) If à A= 1 2 2 4 ! à , B= 2 1 3 2 then à AB = AC = HMHsueh ! à , C= 8 5 16 10 −2 7 5 −1 ! ! . 11 EX 5 (Ex. 9) Suppose R and S are manufacturers of a product. Each year, R keeps (1/4) of its customers and (3/4) switch to S. S keeps (2/3) of its customers and (1/3) switch to R. Let R be the company 1 and S be the company 2. Define the matrix A with entry (i,j), aij = the chance of the customers of company j switch to company i. Then à A= 1/4 1/3 3/4 2/3 ! . If the initial distribution of the market is à x0 = 3 5 2 5 ! , then one year later, the distribution of the market is à x1 = Ax0 = HMHsueh 1 4 3 4 1 3 2 3 !à 3 5 2 5 ! à = 3 ) + ( 1 )( 2 ) )( (1 3 5 4 5 2) 3 3 )( ( 4 )( 5 ) + ( 2 3 5 ! à = 17 60 43 60 ! . 12 Similarly, at the end of 2 years, the distribution of the market will be x2 = Ax1 = A(Ax0) = A2x0. If à x0 = a b ! , where a + b = 1. Find a, b such that the distribution will be the same from year to year. That is, x0 = Ax0 = A2x0 = A3x0 = · · · . HMHsueh 13 x0 = Ax0 implies x0 = Apx0 for any integer p > 0, it suffices to find the solution of Ax0 = x0. Since à Ax0 = x0 ⇔ 1 4 3 4 1 3 2 3 !à a b ! à = a b ! . Then 2 3 1 1 a + b = b. a + b = a, 3 4 3 4 Subsequently, it can be found that 9 4 . , b= 13 13 Note : This is an example of a Markov Chain. a= HMHsueh 14 THM 3 (Properties of Scalar Multiplication) If r, s are real numbers and A, B are matrices, then (a) r(sA) = (rs)A. (b) (r + s)A = rA + sA. (c) r(A + B) = rA + rB. (d) A(rB) = r(AB) = (rA)B. Recall that by definition (rA)ij = raij . HMHsueh 15 THM 4 (Properties of Transpose) If r is a real number and A, B are matrices, then (a) (AT )T = A. (b) (A + B)T = AT + B T . (c) (AB)T = B T AT . (d) (rA)T = rAT . HMHsueh 16 Proof. (c) Let A = [aij ]m×p, B = [bij ]p×n, C = AB = [cij ]m×n and (AB)T = [cT ij ]n×m . Since cT ij = cji = rowj (A) · coli (B) = aj1b1i + aj2b2i + · · · + ajpbpi T + aT bT + · · · + aT bT b = aT pj ip 2j i2 1j i1 T + bT aT + · · · + bT aT a = bT ip pj i2 2j i1 1j = rowi(B T ) · colj (AT ) = (B T AT )ij . Thus, (AB)T = B T AT . HMHsueh 17 EX 6 (Ex. 11) Let à A= 1 3 2 2 −1 3 Then à (AB)T = and à B T AT = HMHsueh ! 0 1 and B = 2 2 . 3 −1 12 7 5 −3 ! 0 2 3 1 2 −1 , ! ! à 1 2 12 7 . 3 −1 = 5 −3 2 3 18 DEF 2 A = [aij ] is symmetric if AT = A. • A is symmetric if it is a square matrix with aij = aji. • If A is symmetric, the elements of A are symmetric with respect to the main diagonal of A. EX 7 (Ex. 12) The matrices 1 0 0 1 2 3 A = 2 4 5 and I3 = 0 1 0 0 0 1 3 5 6 are symmetric. EXERCISES. 9, 16, 19, T.1, T.7, T.8, T.23, T.32 HMHsueh 19