Chapter 3 Elementary Matrix Operations and Determinants 3-1 Elementary Operations Elementary row (or column) operations: 1 2 3 4 1. Interchanging any two adjacent rows (or columns). Eg. 3 4 1 2 2 1 2 1 2. Multiplying any row (or column) by a nonzero constant. Eg. 3 4 6 8 3. Adding any constant multiple of a row (or column) to another row (or column). 1 2 5 6 nd st Eg. : Adding minus double of the 2 row to the 1 row. 3 4 3 4 Rank of A, Rank(A): Rank(A)=dim(Range(LA)), where LA: Fn→Fm. Theorem A Mm×n(F), P Mm×m(F), and Q Mn×n(F). If P and Q are both invertible, then Rank(A)=Rank(PA)=Rank(AQ)=Rank(PAQ). Theorem A, B Mm×n(F), then Rank(A+B) Rank(A)+ Rank(B). [2000 台大電研] (Proof) y Range(LA+LB). x such that y=(LA+LB)(x)= LA(x)+LB(x) ∵ LA(x) Range(LA), LB(x) Range(LB), (LA+LB)(x) Range(LA)+Range(LB) Range( LA LB ) Range( LA ) Range( LB ) Rank(A+B) Rank(A)+ Rank(B) Rank ( A) Theorem A Mm×n(F), B Mn×p(F), then Rank(AB) . [1999 台大電研] Rank (B ) (Proof) y Range(LALB). x such that y=(LALB)(x) ∵ z=LB(x) Range(LB), LA(z) Range(LA), y=(LALB)(x)=LA(LB (x))=LA(z) Range(LA), R( LA LB ) R( LA ) Rank ( AB) Rank ( A) Note: Rank(AB) Rank(B) can be proved by Rank(A)=Rank(At) Question: How to determine the rank of a matrix? 1 0 1 1 0 1 Eg. Given A= 0 1 1 , Rank(A)=dim(Span{ 0, 1, 1 })=2 1 0 1 1 0 1 1 2 Eg. By elementary operation, B= 1 1 1 0 1 1 0 2 0 2 0 1 0 1 1 0 1 0 1 0 1 0 0 0 , Rank(B)=1 0 0 0 0 0 0 1 0 1 Eg. For A= 2 1 4 , find Rank(A). [1993 台大電研] 3 0 a 0 0 1 0 1 1 0 1 0 1. a 3 rank ( A) 2 (Sol.) 2 1 4 2 1 2 2 1 0 , 2. a 3 rank ( A) 3 3 0 a 3 0 a 3 3 0 a 3 1 3 2 1 3 Eg. For A= , determine the dimension of N(A) and the basis for N(A). What 2 5 4 2 4 is Rank(A)? [1993 中山應數所] 1 3 2 1 3 1 3 0 0 3 1 0 0 0 3 1 0 0 0 0 (Sol.) 2 5 4 2 4 2 5 0 0 4 2 1 0 0 4 2 1 0 0 0 Rank(A)=2, Nullity(A)=dim(V)-Rank(A) =5-2=3. x y 1 3 2 1 3 2 5 4 2 4 z 0 v w x 2 1 3 2 1 3 y 0 0 2 0 0 2 z r 1 s 0 t 0 . The basis of N(A) is { 1 , 0, 0} v 0 1 0 0 1 0 w 0 0 1 0 0 1 In Matlab language, we can use the following instructions to obtain the rank of a matrix: >>A=[2 1;4 3]; >>rank(A) A= 2 1 4 3 ans = 2 3-2 Inverse of a Matrix Inverse of a matrix A-1: A A-1= A-1A =I Eg. Show that dA 1 dA 1 A 1 A . [交大電子所] dx dx (Proof) A A-1= A-1A =I, A dA 1 dA 1 dA 1 dA 1 A 0, ∴ A 1 A dx dx dx dx Eg. Show that (AB)-1=B-1A-1. (Proof) AB(AB)-1=I=(AB)-1AB and ABB-1A-1=I=B-1A-1AB, ∴ (AB)-1=B-1A-1 Augmented matrix: [A|B]=(A(1),…,A(n),B(1),…,B(n)) Method of obtaining the inverse of a matrix A: M=Ep…E2E1,. If (A|I)=(MA|M)=(I|M), then M=A-1. 1 1 2 Eg. A= 3 0 1 , find A-1. 1 0 2 1 1 2 1 0 0 1 1 2 1 0 0 1 1 2 1 0 0 0 1 0 1 (Sol.) 3 0 1 0 1 0 0 3 5 3 1 0 0 1 1 0 2 0 0 1 0 1 0 0 5 0 1 3 0 1 0 1 2 1 1 2 1 0 0 1 1 0 1 5 0 1 0 1 0 1 0 1 0 1 0 1 0 0 5 0 1 3 0 0 1 0 5 6 1 0 0 0 2 5 1 5 5 0 1 , 1 0 1 0 1 3 0 0 1 0 1 5 3 5 5 0 2 5 1 5 ∴ A 1 0 1 0 1 5 3 5 1 0 2 4 Eg. A= 2 4 2 , find A-1. 3 3 1 0 2 4 1 0 0 2 4 2 0 1 0 1 2 1 0 2 4 2 0 1 0 0 2 4 1 0 0 0 2 4 1 (Sol.) 3 3 1 0 0 1 3 3 1 0 0 1 3 3 1 0 1 1 2 0 1 2 0 3 2 0 1 2 1 1 0 0 81 0 1 0 4 0 0 1 3 8 0 1 1 0 1 2 1 0 0 2 2 0 0 0 2 4 1 0 0 3 0 1 0 3 2 0 1 2 1 1 0 1 0 1 2 2 1 0 3 1 1 0 3 1 0 0 0 1 2 0 0 0 1 2 0 2 2 3 3 0 0 4 3 3 0 0 1 3 1 1 2 2 2 8 8 1 2 5 8 3 4 3 8 3 1 5 4 8 8 1 1 3 1 2 , A 4 1 4 3 3 4 8 8 3 4 1 2 1 4 0 0 1 4 1 4 4 Eg. (a) Find the inverse of A, where A= 0 1 1 . (b) Find the null space of A. [交大控制、光 1 2 1 1 4 0 0 電研究所] (Ans.) A-1= 1 3 1 , N(A)={ 0 } 1 2 1 0 1 4 7 Eg. Compute the inverse matrix of A= 2 5 8 by some elementary operations. [台科大電研] 3 6 10 a b 0 Eg. Compute the inverse matrix of A= c d 0 . [2003 台科大電研] 0 0 1 d / b / 0 (Ans.) A-1= c / a / 0 , where △=ad-bc 0 0 1 In Matlab language, we can use the following instructions to obtain the inverse of a matrix: >>A=[2,1;4,3] A= 2 1 4 3 >>inv(A) ans = 1.5000 -0.5000 -2.0000 1.0000 3-3 System of Linear Equations Matrix representation for system of linear equations: x1 2 x1 3x2 x3 1 2 3 1 1 Eg. x 2 6 Ax B . x x 2 x 6 1 1 2 x 2 3 1 3 Theorem A Mn×n(F). AX=B has only one solution A-1B if A is invertible. Conversely, if the system has exactly one solution, then A is invertible. Theorem AX=B has at least one solution Rank(A)=Rank(A|B). Gaussian elimination method: x1 2 x 2 x3 1 Eg. Solve 2 x1 2 x 2 x3 1 . 3x 5 x 2 x 1 2 3 1 1 2 1 (Sol.) 2 2 1 3 5 2 1 0 1 0 1 1 0 0 1 1 1 1 3 2 1 1 2 1 0 2 3 0 1 1 1 0 0 0 1 0 0 0 1 1 2 1 1 3 0 1 1 0 2 3 2 1 2 3 1 2 1 0 1 1 0 2 3 1 2 3 4 3 , ∴ x1=4, x2=-3, x3=-1. 1 In Matlab language, we can use the following instructions to solve a linear system of equations: >>A=[1,3,5;2,4,6;1,2,1]; >>B=[2,1,3]'; % [2,1,3]’ is the transpose of [2,1,3] x 3 y 5z 2 >>rref([A,B]) % solve 2 x 4 y 6 z 1. x 2y z 3 ans = 1.0000 0 0 0 1.0000 0 0 0 1.0000 -3.7500 4.0000 -1.2500 Homogeneous linear system: AX =0. 1 5 0 0 Eg. Solve AX=0 for A= 0 0 0 0 1 5 0 0 (Sol.) 0 0 0 0 0 2 0 4 1 0 0 3 0 0 4 8 0 0 0 0 0 2 0 4 1 0 0 3 . [2003 台科大電研] 0 0 4 8 0 0 0 0 x1 x 0 2 x 3 = 0 , x 0 x4 5 0 x 6 x1 4 2 5 x 2 0 0 1 x 3 =r 3 +s 0 +t 0 x 0 1 0 x 4 2 0 0 5 x 6 1 0 0 Theorem A Mm×n(F). AX=0 has a nontrivial solution if m<n and Rank(A)=m. Theorem Let K be the set of all solutions to AX=B, and let KH be the set of the solutions of AX=0. Then for any solution S to AX=B, K={S}+ KH={S+h: h KH }. x1 1 2 1 7 Eg. Solve x2 . 1 1 1 x 4 3 x1 1 1 2 1 0 (Sol.) x 2 0 K H {t 2, t F } 1 1 1 x 3 3 1 x1 1 1 1 2 1 7 S 1 fulfills x2 4 K {1 t 2, t F } 1 1 1 x 4 4 3 3 x1 4 x2 x3 x4 3 Eg. Solve the system of linear equations: 2 x1 8 x2 x3 4 x4 9 . [中正電研] x 4 x 2 x 5 x 6 2 3 4 1 1 4 1 1 (Sol.) 2 8 1 4 1 4 2 5 1 4 1 1 0 0 1 2 0 0 1 2 3 1 1 3 9 6 1 4 1 1 0 0 3 6 0 0 3 6 1 4 1 1 0 0 1 2 0 0 0 0 3 3 3 3 1 0 x1 4 x2 x3 x4 3 x3 2 x 4 1 x3 1 2 x 4 x1 4x2 x4 4 00 x1 4 4 1 x 0 1 0 Let x2=r, x4=s 2 r s x3 1 0 2 x 4 0 0 1 3-4 Determinants of Matrices n Determinant: det(A)= 1 i 1 i j ~ ~ Aij det Aij , where Aij is the (n-1)×(n-1) matrix obtained from A by deleting row i and column j. 1 1 Eg. det( 0 1 1 1 2 1 0 1 4 0 1 1 ~ i2 )= 1 Ai 2 det Ai 2 0 1 1 i 1 1 1 1 1 1 1 1 0 1 1 0 1 1 0 1 2 2 3 2 4 2 1 det 0 1 1 1 0 det 0 1 1 1 0 det 1 1 1 1 1 det 1 1 1 1 1 1 1 1 1 1 1 1 0 1 1 Theorem A Mn×n(F) and Rank(A)<n, then det(A)=0. Theorem (a) For A, B Mn×n(F), then det(AB)=det(A)det(B). (b) If A is invertible, det(A-1)=1/det(A). 1 2 1 3 (c) det(At)=det(A). Eg. det( )=-2= det( ) 3 4 2 4 n (d) det(A)= Aii i 1 0 0 2 if A is a triangular matrix. Eg. det( 11.4 3 0 )=2×(-3)×(-4)=24. 6.2 5.3 4 Theorem Q is invertible det(Q)≠0. Orthogonal matrix B: BtB=BBt=I. Eg. B= 1 2 3 2 3 2 is orthogonal because 1 2 1 2 3 2 3 2 1 2 1 2 3 2 3 2 = 1 0 . 1 0 1 2 Unitary matrix B: BB*=I, where (B*)ij=Bji*. Theorem (a) If B is orthogonal, then det(B)=±1. (b) If B is unitary, then |det(B)|=1. (Proof) (a) det(I)=1=det(BBt)=det(B) det(Bt)=[det(B)]2 det(B)=±1. (b) det(I)=1=det(BB*)=det(B)det(B*)=det(B)det(B)*=|det(B)|2 |det(B)|=1. Summary of properties of the determinants: 1. If B is obtained by exchanging two adjacent rows (or two adjacent columns) of A, then 1 2 3 4 det(B)=-det(A). Eg. det( )=-2 and det( )=2. 3 4 1 2 2. If B is obtained by multiplying each entry of some row (or column) of A by c, then det(B)=c.det(A). 2 1 2 1 Eg. det( )=-2 and det( )=(-12)-(-18)=6=(-3)×(-2). 3 4 9 12 3. If B is obtained from A by adding a multiple of row i (or column i) to row j (or column j), where i≠j, 1 2 5 6 then det(A)=det(B). Eg. det( )=-2 and det( )=-2. 4 3 4 3 4. det(I)=1. 0 0 1 if A is a triangular matrix. Eg. det( 11.4 3 0 )=(1)×(-3)×(2)=-6. 6.2 5.3 2 n 5. det(A)= Aii i 1 4 2 2 0 0 1 1 2 Eg. Find the determinants of A= 3 1 (Sol.) Triangularize A, we obtain det(A)= 1 4 2 3 2 2 0 4 3 0 1 2 1 2 2 3 2 2 1 4 2 3 0 6 4 2 0 12 5 7 2 4 6 0 and A-1. 3 3 4 2 1 4 0 6 0 0 0 0 2 3 4 2 3 3 8 3 16 3 1 4 0 6 0 0 0 0 ∴ det(A-1)= -1/144 0 3 1 1 2 1 1 1 , det(A)=? Eg. A= 1 2 3 1 3 1 1 2 0 3 0 3 0 3 1 1 1 1 1 1 0 1 1 5 0 1 1 5 0 1 1 5 (Sol.) A 0 3 0 0 0 13 0 0 3 2 3 13 0 13 0 4 1 7 0 0 3 13 0 0 det(A)=-[1.(-1).3.(-13)]=-39 2 3 4 2 3 3 0 8 =-144, Eg. Digits 1 to 9 can be arranged into 3×3 matrices in 9! ways. Find the sum of the determinants of these matrices. [台大電研] a b (Sol.) [det( d e g h c d e f )+det( a b g h i f a c )]+[det( b c i x2 x3 1 x1 x 1 x2 x3 1 -1 Eg. Find det(A ) for A= x1 x2 1 x3 ..... ..... ..... x1 x2 x3 (Sol.) d e f g a h )+det( b c i g h i d e )]+…=0. f ..... xn ..... x n ..... x n . [1990 中央土木所] ..... ..... ..... 1 x n x2 1 1 1 x1 1 1 det ( ) det( ) det( x 1 x 0 1 x x ) 1 x1 x2 , 1 x2 2 1 2 x1 1 x2 x3 1 0 1 x1 1 1 det( x1 1 x2 x3 ) det( x1 1 x2 x3 ) det( 0 x1 x1 x1 x2 1 x3 x2 1 x3 1 1 x2 0 1 ) 1 x3 1 0 0 1 1 1 ) 1 x x x ,…, ∴ det A 1 1 1 1 1 ) det( 0 1 1 = det( 0 1 2 3 n det A 1 xi 0 x1 x2 1 x3 0 0 1 x1 x2 x3 i 1 2 1 1 3 Eg. Triangularize the matrix: 3 0 2 2 0 3 2 4 . [台大電研] 2 1 4 6 In Matlab language, we can use the following instructions to obtain the determinant of a matrix: >>A=[1,3,0;-1,5,2;1,2,1]; >>det(A) ans = 10 Eg. Let f1(t), f2(t), …, fn(t) be the polynomials of degrees n-2. Prove that det(A)=0 f1 a1 ..... A= ..... ..... f n a1 f1 a 2 ..... ..... ..... f n a 2 f1 a3 ..... ..... ..... f n a3 ..... ..... ..... ..... ..... f1 a n ..... ..... , where a1, a2, …, an are arbitrary constants. [1991 台大 ..... f n a n 電研] (Sol.) i n, f i t ai 0 ai1t ..... ai ,n2 t n2 Span 1, t , ...., t n2 ∵ dim Span 1, t , ...., t n 2 n 1 n , ∴ { f1(t), f2(t), …, fn(t)} is linearly dependent. That is, j f j t c1 f1 t c 2 f 2 t ... c j 1 f j 1 (t ) c j 1 f j 1 (t ) ... c n f n t , 1 j n s.t for arbitrary ai , f j ai c1 f1 ai c 2 f 2 ai ... c j 1 f j 1 (ai ) c j 1 f j 1 (ai ) ... c n f n ai f1 a1 ..... A f j a1 ..... f n a1 f1 a1 ..... ci f i a1 i j ..... f n a1 f1 a 2 ..... f j a 2 ..... f n a 2 f1 a3 ..... ..... ..... f n a3 f1 a 2 ..... ci f i a2 ..... ..... ..... ..... ..... f1 a n ..... f j a n ..... f n a n f1 a3 ..... ..... ..... ..... ..... i j ..... f n a 2 for ..... f n a3 f1 a n ..... ci f i an rank A n detA 0 i j ..... ..... ..... f n a n 3-5 Classical Adjoint Matrix n i j Cofactor: For det(A)= 1 is the cofactor of A . ~ ~ i j Aij det Aij . , the scalar factor 1 det Aij ij 1 2 1 Eg. For A= 2 3 1 1 5 1 4 1 , find the cofactor of A41. 0 3 1 6 1 2 (Sol.) The cofactor of A41 = 1 4 1 5 1 1 det 1 4 1 23 0 3 1 Classical adjoint matrix: adj(A)= det(A)A-1 C is called the classical adjoint matrix of A if det(A)≠0, and Cij=cofactor of Aji. Note: adj(A)‧ A = det(A)‧ I and A-1=adj(A)/det(A) 1 2 3 Eg. For A= 1 3 4 , find adj(A) and A-1. 1 4 3 3 4 1 (Sol.) det(A)=-2, adj(A) = 1 1 1 4 3 4 3 3 4 2 3 4 3 1 3 1 3 1 2 1 4 7 7 6 1 2 adj ( A) 1 1 A1 1 0 1 d e tA 2 2 1 2 1 1 2 2 3 3 4 7 6 1 1 3 0 1 = 1 1 4 1 2 1 1 2 1 3 3 0 1 1 2 1 2 1 2 Theorem A-1 does not exist det(A)=0. Eg. For A = 0 0 0 3 2 , find β such that A-1 does not exist. [1990 清大物理所] 3 (Sol.) det(A)=9β-β3=0, β=0, ±3. Theorem If C=adj(A) Mn×n(F), then det(C)=[det(A)]n-1. 0 det A 0 0 0 ..... (Proof) C adj ( A) det AA 1 CA det AI n 0 det A 0 det CA det C det A det A det(C ) det A n 1 n . 2 2 0 Eg. For adj(A)= 0 2 1 , find det(A) and A. [1990 台大機研] 0 0 1 (Sol.) n=3, det(adj(A))=[det(A)]n-1=4 det(A)=±2 0 1 1 1 1 1 1 A 1 adj A A 1 0 1 0.5 A 0 1 1 . det A 0 0 0 0 2 0.5 1 c01 1 1 c1 Vander monde matrix: ... ... ... ... 1 c 1 n n 2 c0 2 c1 ... ... cn 2 n ... c0 n ... c1 ... ... ... ... n ... c n j 1 Theorem det(M)= (c j ci ) if M is the Vander monde matrix. j 1 i 0 (Proof) If some j=i, then cj=ci det(M)=0 (cj-ci) factor n j 1 Let det(M)= k (c j ci ) . For n=1, det(M)= (c1-c0)=k(c1-c0), k=1 j 1 i 0 1 1 1 2 Eg. Let M= 1 3 1 1 1 1 4 8 , determine det(M). 9 27 1 1 (Sol.) n=3, c0=1, c1=2, c2=-3, c3=-1, n j 1 det(M)= (c j ci ) =(c1-c0)(c2-c0)(c2-c1)(c3-c0)(c3-c1)(c3-c2) j 1 i 0 =(2-1)(-3-1)(-3-2)(-1-1)(-1-2)(-1+3)=240