264 Chapter 6 Matrix Algebra Objective 6.4 Solving Systems by Reducing Matrices ‘To show how to reduce a matrix and In this section we illustrate a method by which matrices can be used to solve a system eer, uoeON to soe a of linear equations. In developing this method of reduction, we will first solve a system by the usual method of elimination. Then we will obtain the same solution by using matrices. Let us consider the system f - ysl a x+I=5 @ consisting of two linear equations in two unknowns, x and y. Although this system can be solved by various algebraic methods, we willl solve it by a method that is readily adapted to matrices. For reasons that will be obvious later, we begin by replacing Equation (1) by Equation (2), and Equation (2) by Equation (1), thus obtaining the equivalent system? 8 ‘Multiplying both sides of Equation (3) by ~3 gives ~3x — 6y = ~15. Adding the left and right sides of this equation to the corresponding sides of Equation (4) produces an equivalent system in which x is eliminated from the second equation: x+2= 5 © Ox —7y = 14 © ‘Now we will eliminate y from the first equation. Multiplying both sides of Equation (6) by —! gives the equivalent system, xy ort y a 2 @) From Equation (8), y = 2, and, hence, ~2y = —4. Adding the sides of —2y = —4 to the corresponding sides of Equation (7), we get the equivalent system, x+0y=1 Ox + y=? ‘Therefore, x = 1 and y = 2, so the original system is solved. ‘Note thatin solving the original system, we successively replaced itby an equivalent system that was obtained by performing one of the following three operations (called elementary operations), which leave the solution unchanged 1, Interchanging two equations 2. Multiplying one equation by a nonzero constant 3. Adding a constant multiple ofthe sides of one equation to the corresponding sides of another equation Before showing a matrix method of solving the original system, 3x — Wwe first need to define some terms. Recall from Section 6.3 that the matrix 3-1 fen Socal rom Section 3.4 that wo systems are equivalent f they have the same set of solutions.Section 4 Solving Systems by Reducing Matrices 265 is the coefficient matrix of this system. The entries in the fist column correspond to the coefficients of the x’s in the equations. For example, the entry in the first row and first column corresponds to the coefficient of xin the first equation; and the entry in the second row and first column corresponds to the coefficient of x in the second equation Similarly, the entries in the second column correspond to the coefficients of the y's. Another matrix associated with this system is called the augmented coefficient ‘matrix and is given by 3-1/1 1 245 ‘The first and second columns are the first and second columns, respectively, of the coefficient matrix. The entries in the third column correspond to the constant terms in the system: The entry in the first row of this column is the constant term of the first equation, whereas the entry in the second row is the constant term of the second equation. Although it is not necessary to include the vertical line in the augmented coefficient matrix, it serves to remind us that the 1 and the 5 are the constant terms that appear on the right sides of the equations. The augmented coefficient matrix itself completely describes the system of equations. ‘The procedure that was used to solve the original system involved a number of equivalent systems. With each of these systems, we can associate its augmented coef- ficient matrix. Following are the systems that were involved, together with their corre- sponding augmented coefficient matrices, which we have labeled A, B, C, D, and B. Let us see how these matrices are related. ‘Matrix B can be obtained from A by interchanging the first and second rows of A. This operation corresponds to interchanging the two equations in the original system. Matrix C can be obtained from B by adding, to each entry in the second row of B, —3 times the corresponding entry in the first row of B: c 1 2 5 3+ (-3M(1) 1+ (-392)_ | 1+ (3966). _fi 2 5 =lo-7 | -14 This operation is described as follows: the addition of —3 times the first row of B to the second row of B. Matrix D can be obtained from C by multiplying each entry in the second row of C by —4. This operation is referred to as multiplying the second row of C by —4. “Matrix E can be obtained from D by adding ~2 times the second row of D to the first row of D.266 Chapter 6 Matrix Algebra Observe that E, which gives the solution, was obtained from A by successively performing one of three matrix operations, called elementary row operations: Elementary Row Operations 1, Interchanging two rows of a matrix 2. Multiplying a row of a matrix by a nonzero number 3. Adding a multiple of one row of a matrix to a different row of that matrix ‘These elementary row operations correspond to the three elementary operations used in the algebraic method of elimination, Whenever a matrix can be obtained from another by one or more elementary row operations, we say that the matrices are equivalent. Thus, A and E are equivalent, (We could also obtain A from E by performing simi- Jar row operations in the reverse order, so the term equivalent is appropriate.) When describing particular elementary row operations, we will use the following notation for convenience: Notation Corresponding Row Operation Ri@R; Interchange rows R, and R; IR; ‘Multiply row R; by the rionzero constant k. KR/+R, Add é times row R; to row R, (but leave R; unchanged). For example, writing 42 1 0-2 9 Pee eee eo 5 0 3 Sp mon a ‘means that the second matrix was obtained from the first by adding —4 times row 1 to row 2. Note that we can write (—K)R) as —KR;. | We are now ready to describe a matrix procedure for solving a system of linear} equations, First, we form the augmented coefficient matrix of the system; then, by ‘means of elementary row operations, we determine an equivalent matrix that clearly indicates the solution. Let us be specific as to what we mean by a matrix that clearly indicates the solution. This is a matrix, called a reduced matrix, which will be defined below. It is convenient to define first a zero-row of a matrix to be a row that consists entirely of zeros. A row that is not zero-row, meaning that it contains at least one nonzero entry, will be called @ nonzero-row. The first nonzero entry in a nonzero-row is called the leading entry. Reduced Matrix ‘A matrix is said to be a reduced matrix provided that al of the following are true: 1, All zero-rows are at the bottom of the matrix. 2. For each nonzero-row, the leading entry is 1, and all other entries in the column of the leading entry are 0. 3. The leading entry in each row is to the right of the leading entry in any row above it Itcan be shown that each matrix is equivalent to exactly one reduced matrix. To solve a system, we find she reduced matrix such that the augmented coefficient matrix is equivalent to it. In our previous discussion of elementary row operations, the matrix ef tl] is a reduced matrix.Section 64 Solving Systerns by Reducing Matrices 267 (EXAMPLE 1 Reduced Matrices For each of the following matrices, determine whether it is reduced or not reduced. [: 5] oft 0 a ff i] 03 010 10, 00 0 1 0 Ole OH 000 e {0 0 f£Jo 012 o1 0000 Solution: a, Not a reduced matrix, because the leading entry in the second row is not 1 b. Reduced matrix Not a reduced matrix, because the leading entry in the second row is not to the right of the leading entry in the first row a. Reduced matrix e. Not a reduced matrix, because the second row, which is a zero-row, is not at the bottom of the matrix eco £. Reduced matrix Now Work Problem 1 < (EXAMPLE 2 Reducing a Matrix O- Oo © 3-6 -3 0 6-122 211 Strategy To reduce the matrix, we must get the leading entry to be a I in the first row, the leading entry a1 in the second row, and so on, until we arrive ata zero-row, if there are any. Moreover, we must work from left to right, because the lea entry in each row must be to the left of all other leading entries in the rows below it. Reduce the matrix Solution: Since there are no zero-rows to move to the bottom, we proceed to find the first column that contains a nonzero entry; this turns out to be column 1. Accordingly, in the reduced matrix, the leading 1 in the first row must be in column 1. To accomplish this, we begin by interchanging the first two rows so that a nonzero entry is in row 1 of column 1 0 0 1 2 semen 3-6 -3 of BER loo 1 2 6-2 2 11 6-2 2 1 Next, we multiply row I by so that the leading entry is a ir, [lore2 cero) aS ee 6-12 2 11 Now, because we must have zeros below (and above) each leading entry, we add —6 times row 1 to row 3 euur, fl 2-1 0 SRLRS i 0 1 | oOmee Next, we move to the right of column 1 to find the first column that has a nonzero entry in row 2 or below; this is column 3. Consequently, in the reduced matrix, the268 Chapter 6 Matrix Algebra caution\ “The sequence of steps tat is used to rede aati isnt unique: however, the reduced mare unique APPLY IT > ‘8, An investment firm offers three stock portfolios: A, B, and C. The number of Dlocks of each type of stock in each of these portfolios is summarized in the following table: Portfolio ABC High 6 1 3 Risk) Moderate 3°23 low 1 «5 3 A client wants 35 blocks of high: risk stock, 22 blocks of moderate-isk ‘stock, and 18 blocks of low-risk stock. How many of each portfolio should be suggested? leading 1 in the second row must be in column 3. The foregoing matrix already does have a leading 1 there. Thus, all we need do to get zeros below and above the leading 1 is add 1 times row 2 to row 1 and add ~8 times row 2 to row 3: TSAO] ORR To 1 2 SR +R; [oo 005. Again, we move to the right to find the first column that has a nonzero entry in row 3; namely, column 4. To make the leading entry a1, we multiply row 3 by 1 1-20 2 0 012 O0°04 Finally, to get all other entries in column 4 to be zeros, we add ~2 times row 3 to both row I and row 2: 1-20-07 2 2WtR | og 0 00 IR +R: |g 1 ‘The last matrix is in reduced form, Now Work Problem 94 ‘The method of reducti eralized to systems con: system as described for solving our original system can be gen- ting of m linear equations in n unknowns. To solve such a Aux, + Anx + Aan + Ane + AmiX1 + Anoka ++ involves 4. determining the augmented coefficient matrix of the system, which is ‘An Aw Ain | Bi An An Aan | Ba Am) Ana Arm | Bu and 2, determining the reduced matrix to which the augmented coefficient matrix is equivalent. Frequently, step 2 is called reducing the augmented coefficient matrix. (GRAMPLES Solving a System by Reduction By using matrix reduction, solve the system 2e+3y a+ y xhyAPPLY IT > 9. A health spa customizes the diet and vitamin supplements of each ofits x+2y+4e-6 clients. The spa offers three different vitamin supplements, each containing different. percentages of the recom- Section 64 Sohing Systems by Reducing Matrices 269 Solution: Reducing the augmented coefficient matrix of the system, we have 2 3 fet 2 1 a] 5 lb a 3] ae 13 3]-1, ajr -1 1 1 cb f t Since the original system is equivalent to this system, it has a unique solution, namely, x= 4 -3 Now Work Problem 13-4 (EXAMPLE 4 Solving a System by Reduction Using matrix reduction, solve 2+ y-3 x+ yt22-1=0 ‘mended daily allowance (RDA) of Solution: Rewriting the system so that the variables are aligned and the constant terms vitamins A, C, and D. One tablet of appear on the right sides of the equations, we have supplement X provides 40% of the RDA OF A, 20% of the RDA of C, and 10% : +2944 of the RDA of D. One tablet of supple- yt tment ¥ provides 10% ofthe RDA ofA, xt y+22 10% of the RDA of C, and 30% ofthe RDA of D. One tablet of supplement Z Reducing the augmented coefficient matrix, we obtain provides 10% of the RDA of A, 50% 12 416 Teoeealeg of the RDA of C, and 20% of the RDA -RitR; ‘of D. The spa staff determines that one ep eee pe ee client should take 180% of the RDA of 11 2it 0. 1 =2]=5) vitamin A, 200% of the RDA of vitamin , and 190% of the RDA of vitamin D 10 0] 0) each day. How many tablets df each sup- 2Ra +R, plement should she take each day? (R2+Rs 7 pam e270 Chapter Matrix Algebra Whenever we get a row with all O's wo the left side ofthe vertical rule anda nonzero entry tothe right. no solution exist, APPLY IT 10. A. 200 veterinarian can purchase animal food of four different types: A, B,C, and D, Each food comes in the same size bag, and the number of grams of each of three nutrients in each bag is summarized in the following table: Food ABCD N 5 510 5 Nuiemt N; 10. 5 30 10 Ns 5 15 10 25 For one animal, the veterinarian dete ‘mines that she needs to combine the bags to get 10,000 g of Nj, 20,000 g of No, ‘and 20,000 g of Ns. How many bags of cach type of food should she order? cig, fio 0 =a | Omer le o 0 0 . 1 0 of 0 RR: lggafmo Gay of [pe ‘The last matrix is reduced and corresponds to Since 0 # 1, there are no values of x, y, and z for which all equations are satisfied simultaneously. Thus, the original system has no solution. Now Work Problem 15 < (EXAMPLES Parametric Form of a Solution Using matrix reduction, solve Day + Bee + Das + Gry = 10 wrt 2+ y= 2 3m =35 +6 = 9 Solution: Reducing the augmented coefficient matrix, we have 23 2 6/10 eg ea 01 21/2 o 1 2 30 -3 6/9 3.0 -3 3 1 3 4 BRR |p 7 0 -3 -6 ER. +R, o 1 2R+R, [i 0 2RSFR: | e 1 ‘This matrix is reduced and corresponds to the system Hon ENS,| | | ; } Section 6.4 Solving Systems by Reducing Matrices 271 Thus, na4—hy m=0 (19) asl—dy ay ‘The system imposes no restrictions on xs so that x5 may take on any real value. If swe append mS a2) to the preceding equations, then we have expressed all four of the unknowns in terms of xj and this is the general solution of the original system. For each particular value of xq, Equations (9)-(12) determine a particular solution of the original system. For example, if xy = 0, then a particular solution is n 4 m=0 4 4 =0 Ifxy =2, then u=0 is another particular solution, Since there are infinitely many possibilities for xs, there are infinitely many solutions of the original system. Recall (see Examples 3 and 6 of Section 3.4) that, if we like, we can write xj = r and refer to this new variable r as a parameter. (However, there is nothing special about the name r so that we could consider x4 as the parameter on which all the original variables depend, Note that we can write x2 = 0 + Oxs and x; = 0+ Lx.) Writing r for the parameter, the solution of the original system is given by xa =O4Ir where ris any real number, and we speak of having a one-parameter family of solutions. Now, with matrix addition and scalar multiplication at hand, we can say a little more about such families. Observe that 7 4 ” 0 1 0. B Residers familiar with analytic geometry will see that the solutions form a fine in 4 -xrmasste-space, passing through the point | 9 | and inthe direction ofthe line segment 0 0 3 joining | 9 | and}? 9. 1 Now Work Problem 17 < Examples 3- illustrate the fact thata system of linear equations may have a unique solution, no solution, or infinitely many solutions.272 Chapter Matrix Algebra PROBLEMS 6.4 In Problems 1-6, determine whether the matrix is reduced or not reduced. 100 12 1003 Ea =f ] afore 70 0012 aan 11 000 or o1 0100 100 4 loo oo10 Clond 00 0000 000. ‘In Problems 7-12, reduce the given masrix. 246 ef soy. ftas 123 on-n—un ntn—9—u—% nt tmomoa mtn ty ta —3 nt te a= nen te—x% co Ay bay 33 — Xa mom by =0 3 2344 0 0 2 1-6 1723 2 03 Wy4 af la 42 0] Ilo -1 0 1 7, oro. 0 44. Solve the systems in Problems 13-26 by the method of reduction a3, 2x-y=50 15, [r+3y+22—1=0 ” we (ES Ticioe ay t4n w. 1 4381 — xhyna=7 a. 2, \2x—3y—2e= 4 eo y-Se=B 5 +3 Bet 2y+ Iz co | WV s+ yt dz 2x 394 3c 0 0 Solve Problems 27-33 by using matrix reduction. 27. Taxes A company has taxable income of $312,000. The federal tax is 25% of that portion that is lft after the state tax has ‘been paid. The state taxis 10% of that portion thats let after the federal tax has been paid, Find the company's federal and state taxes. 28. Decision Making A manufacturer produces two products, ‘A.and B. For each unit of A sold, the profit is $8, and for each unit, ‘Of B sold, the profit is $11. From experience, it has been found that 25% more of A can be sold than of B. Next year the ‘manufacturer desires a total profit of $42,000. How many units of teach product must be sold? 29, Production Scheduling A manufacturer produces three products: A, B, and C. The profits for each unit of A, B, and C sold are $1, $2, and $3, respectively. Fixed costs are $17,000 per year, and the costs of producing each unit of A, B, and C are $4, $5, and ‘87, respectively. Next year, a total of 11,000 units ofall three products is to be produced and sold, and a total profit of $25,000 is to be realized. If total cost is to be $80,000, how many units of each ofthe products should be produced next yeat? 30, Production Allocation National Desk Co. has plants for ‘producing desks on both the East Coast and West Coast. At the East Coast plant, fixed costs are $20,000 per year and the cost of producing each desk is $90. At the West Coast plant, fixed costs are $18,000 per year and the cost of producing each desk is $95, Next year the company wants to produce a total of 800 desks. Determine the production order for each plant for the fortheoming year ifthe total cost for each plant is to be the same. 31, Vitamins A person is ordered by a doctor to take 10 units (of vitamin A, 9 units of vitamin D, and 19 units of vitamin E each day. The person can choose from three brands of vitamin pills. Brand X contains 2 units of vitamin A, 3 units of vitamin D, and 5 units of vitamin E; brand Y has 1, 3, and 4 units, respectively; and brand Z has 1 unit of vitamin A, none of vitamin D, and 1 of vitamin E. vray 8 (a) Find all possible combinations of pills that will provide exactly the required amounts of vitamins. (b) If brand X costs 1 centa pill, brand Y 6 cents, and brand Z 3 cents, are there any combinations in part (a) costing exactly 1S cents a day? (©) What isthe least expensive combination in part (a)? the most expensive? 32, Production A firm produces three products, A, B, and C, that require processing by three machines, I, Il, and II. The time in hours required for processing one unit of each product by the three machines is given by the following table: AES BFr Section 6.5 Soling Systems by Reducing Matrices (continued) 273 “Machine 1s available for 440 hours, machine I for 310 hours, and machine Il for 560 hours. Find how many units ofeach product shouldbe produced to make use of all the available time fon the machines. 33, Investments An investment company sells three types of pooled funds, Standard (S), Deluxe (D), and Gold Star (G). Each unit of contains 12 shares of stock A, 16 of stock B, and 8 Each unit of G contains 32 shares of stock A, 28 of stock B, and 36 oF stack C. ‘Suppose an investor wishes to purchase exactly 220 shares of stock A, 176 shares of stock B, and 264 shares of stock C by buying Units of the three funds. (a) Determine those combinations of units of S, D, and G that will, meet the investor’s requirements exactly. of stock C. Each unit of D contains 20 shares of stock A, 12 of stock B, and 28 of stock C. (b) Suppose the investor pays $300 for each unit of S, $400 for each unit of D, and $600 for each unit of G. Which of the ‘combinations from part a) will minimize the total cost to the investor? Objective 6.5 Solving Systems by Reducing To focus our attention on Matrices (continued) onhomageneous aystoms thet Involve more than one parameter in ‘the'r general solution; and to solve, ‘and Gonsider the theory of, homegeneous systems. ‘As we saw in Section 6.4, a system of linear equations may have a unique solution, no solution, or infinitely many solutions. When there are infinitely many, the general solution is expressed in terms of at least one parameter. For example, the general solution in Example 5 was given in terms of the parameter r: ‘Attimes, more than one parameter is necessary as the next example shows. (EXAMPLE 1 Using matrix reduction, solve [Pes ‘Two-Parameter Family of Solutions ab Xa 3x + 4a m- m- mt2y= 3 : The augmented coefficient matrix is 1 02 5 5/-3 10103 4)-1 1-1 -1 2] 3 whose reduced form is tO ules), Oe ei 2) . 0000] 0 Hence, xt xg + 3xy pb Osh y= from which it follows that a= I= a5-3% 2—245— 2X4 n= Since no restriction is placed on either 23 or x, they can be arbitrary real numbers, giving us a parametric family of solutions. Setting xy = r and x4 = s, we can give the 44Sec Example 7 of Section 3.274 Chapter 6 Matrix Algebra solution of the given system as a 35 m=-2-2r-s Ber mas where the parameters r and s can be any real numbers. By assigning specific values to rand s, we get particular solutions. For example, if r = 1 and s = 2, then the corresponding particular solution is x; = ~6,x2 = —6,x3 = 1, and x4 = 2. As in the ‘one-parameter case, we can now go further and write 1 1 -1 -3 |_| 2 -2 -1 Bs ee | 400) (ease |p| x 0 0. 1 1 hich can be shown to exhibit the family of solutions asa plane through | — | in 0 xitpxsts-space. Now Work Problem 1 <4 tis customary to classify a system of linear equations as being either homogeneous or nonhomogeneous, depending on whether the constant terms are all zero. Definition ‘The system Ami%i + Anata +--+ Anni = Bn is called a homogeneous system if By = By = By = 0. The system is a nonhomogeneous system if atleast one of the B; is not equal to 0. (EXAMPLE 2) Nonhomogeneous and Homogeneous Systems ‘The system is nonhomogeneous because of the 4 in the first equation. The system 2x +3; Be 4 is homogeneous. If the homogeneous systemi | Section 6.5 Solving Systems by Reducing Matrices (continued) 275 ‘were solved by the method of reduction, first the augmented coefficient matrix would be written 2 3 |0 3 -4 | 0 Observe thatthe last column consists entirely of zeros. This is typical of the augmented coefficient matrix of any homogeneous system. We would then reduce this matrix by using elementary row operations: fi ojo o 10 2 3 |0 3 -4|o0]7 ‘The last column of the reduced matrix also consists only of zeros. This does not occur by chance. When any elementary row operation is performed on a matrix that has a column consisting entirely of zeros, the corresponding column of the resulting matrix will also be all zeros. For convenience, it will be our custom when solving a homogeneous system by matrix reduction to delete the last column of the matrices involved. That is, wwe will reduce only the coefficient matrix of the system. For the preceding system, we would have fio o1 Here the reduced matrix, called the reduced coefficient matrix, correspondsto thesystem x+0y=0 ort y=0 so the solution is x = O and y =0. Let us now consider the number of solutions of the homogeneous system Aisi + Ana +++ + Atty =O Aaa, + Ama +--+ Army =0 Amit + Anaka +++ + Aim = 0 One solution always occurs when x) = 0,22 = 0,..., and x, = 0, since each equation is satisfied for these values. This solution, called the trivial solution, is a solution of every homogeneous system and follows from the matrix equation AD = Om where 0, is the n x 1 column vector (and Oy is the m x 1 column vector), ‘There is a theorem that allows us to determine whether a homogeneous system has unique solution (the trivial solution only) or infinitely many solutions. The theorem is based on the number of nonzero-rows that appear in the reduced coefficient matrix of the system. Recall that a nonzero-row is a row that does not consist entirely of zeros. Theorem Let Abe the reduced coefficient matrix of a homogeneous system of m linear equa- tions in n unknowns. If A has exactly k nonzero-rows, then k 12, Secret messages can be encoded by using a code and an encoding mari. Let the encoding matrix be E. Then we ‘ean encode a message by taking every two Ieters of the message, converting them to their corresponding numbers, creating a2 x 1 matrix, and then mul. tiplying each matrix by £.The message ‘can be unscrambled with a decoding ‘matrix that is the inverse of the coding ‘mattix—thatis, 2". Determine whether the encoding matrices 2 18 1 05, [5 3] as ae inverses of each other can be written in the matrix form AX = B, where -f 2) ye) Rp [i] xB) fl Motivation for what we now have in mind is provided by looking at the procedure for solving the algebraic equation ax = b. The later equation is solved by simply ‘multiplying both sides by the multiplicative inverse of a. (Recall that the multiplicative inverse of a nonzero number @ is denoted a~! [which is I/a] and has the property that ala = 1.) For example, if 3x = 11, then ul s'GN=3'G) 90 x= > If we are to apply a similar procedure to the matrix equation AX=B a then we need a multiplicative inverse of A—that is, a matrix C such that CA =I. If we have such a C, then we can simply multiply both sides of Equation (1) by C and get C(AX) = CB (CA)X = CB IX =CB X=CB ‘This shows us that if there isa solution of AX = B, then the only possible solution is the matrix CB. Since we know that a matrix equation can have no solutions, a unique solution, or infinitely many solutions, we see immediately that this strategy cannot possibly work unless the matrix equation has a unique solution. For CB to actually be a solution, we require that A(CB) = B, which isthe same as requiring that (AC)B = B. However, since matrix multiplication is not commutative, our assumption that CA does not give us AC = I. Consider the matrix products below, for example: trol[j]=me=n but [a]tol=[oo]4e However, if A and C are square matrices of the same order n, then it can be proved that AC = 1, follows from CA = I, 50 in this case we can finish the argument above and conclude that CB is a solution, necessarily the only one, of AX = B. For a square matrix A, when a matrix C exists satisfying CA = /, necessarily C is also square of the same size as A and we say that itis an inverse matrix (or simply an inverse) of A. Definition IFA is a square matrix and there exists a matrix C such that CA = I, then C is called an inverse of A, and A is said to be invertible. (SAMPLE 1 Inverse of a Matrix [! qese= [4 2}. sne a-[3 “IB J-B matrix C is an inverse of A. < Itcan be shown that an invertible matrix has one and only one inverse; that is, an inverse is unique. Thus, in Example 1, matrix C is the onfy matrix such that CA = Icaution’, For general functions, if g of = not follow that fg = it does £2 [Efe mieees message, Use the code from Apply it 1 215), 1-05 decode the message, broken into the following pieces: 28, 46,65.90 61,82 59,88,57,86 60, 84,21, 34,76, 102 sand the inverse F Section66 Inverses 279 For this reason, we can speak of the inverse of an invertible matrix A, which we denote by the symbol A“. Accordingly, A~'A = I. Moreover, although matrix multiplication is not generally commutative, it is a fact that A~! commutes with A: ay Aat Returning to the matrix equation AX = B, Equation (1), we can now state the following: IfA is an invertible matrix, then the matrix equation AX = B has the unique solution X=A7B, Probably the idea ofan inverse matrix has an air of déja vu about it. In Section 2.4 wwe discussed inverse fimctions, and these can be used to further understand inverse matrices. Let R" denote the set of x 1 column matrices (and let R™ denote the set of m x L column matrices) If A is an m Xx n matrix, then f(X) = AX defines a function :R" —> R".Ifm =n, itcan be shown that the function given by F(X) = AX has an inverse in the sense of Section 2.4 if and only if A has an inverse matrix A~!, in which case f'(X) = A“X, here is one caution to be observed here. In general, for a function f to have an inverse, say g, then we require both g of = I and f o g = I, where I is the identity function. It is a rather special fact about square matrices that CA = J implies also ACI. If is a function thet has an inverse, then any equation of the form f(x) = b has a unique solution, namely x = f-!(b). (EXAMPLE 2 Using the Inverse to Solve a System Solve the system ay +2 3x + Tez = 18 Solution: In matrix form, we have AX = B, where s-f 3] xf] » In Example 1, we showed that Therefore, 50% Now Work Problem 19 < In order that the method of Example 2 be applicable a system, two conditions must be met: 1, The system must have the same number of equations as there are unknowns. 2, The coefficient matrix must be invertible. As far as condition 2 is concemed, we caution that not all nonzero square matrices are invertible. For example, if 0 1 o1280 Chapter Matrix Algebra then @ b]fo 1]_fo ats] fr o i e a\{o 1]=|o c+a]*lo 1 for any values of a,b c, and d. Hence, there is no maltix that, when postmultiplied by A, yields the identity matrix. Thus, A is not invertible. Thereis an interesting mechanical procedure that allows us simultaneously to deter- mine whether or not a matix is invertible and find its inverse if iis so. The procedure is based on an observation whose proof would take us too far afield. First, recall that for any matix A there isa sequence Ey,E,..., of elementary row operations tha, cauTion’\ when applied to A, produce a reduced matrix. In other words, we have Every identity matrix is a reduced matrix, As A, 2 hy Be Ay tat ot every Caquare) reduced mt is at dent mass Forename any tee where Ay isa reduced matrix. We recal, to, that Ax is unigue and determined by A fiat Oi etced alone (even though there can be many sequences, of variable lengths, of elementary row operations that accomplish this reduction). IFA is square, say n x n, then we might have Ay = fy the m x n identity matrix. Theorem For square A and Ax as previously, A is invertible if and only if Ay = /. Moreover, if Ey, By,...,Ex is a sequence of elementary row operations that takes A to I, then the same sequence takes J to A~! (EXAMPLE3 | Determining the Invertibility of a Matrix Apply the theorem to determine if the matrix 10 aya Strategy We will augment A with a copy of the (2 x 2) identity matrix (just as, Wwe have often augmented a matrix by a column vector). The result will be 2.x 4. ‘We will apply elementary row operations to the entire 2 x 4 matrix until the first columns form a reduced matrix. Ifthe resultis , then, by the theorem, Ais invertible, bbut because we have applied the operations to the entire 2 x 4 matrix, the last columns will, also by the theorem, be transformed from I to A, if A is in fact invertible. A is invertible, \ Solution: We have ain [ ale ‘ 2+R fl of 1 0 pereme |Or2)| 20 aR fof 1 oF ef ot t=wia Since [A | 1] transforms with 1 to the left of the augmentation bar, the matrix A is invertible and the matrix B to the right of the augmentation bar is A~!. Specifically we conclude that Now Work Problem 1 1, We could extend the encoding scheme used in Apply it 1 to a3 x 3 matrix, encoding three leters of a mes- sage ata time. Find the inverses of the following 3 x 3 encoding matrices: 2127 323 434 Section6.6 lnvewses 281 Method to Find the Inverse of a Matrix IfAis an nxn matrix, form then x (2n) matrix (A | 1 and perform elementary row operations until the first n columns form a reduced matrix. Assume that the result is (R | B) so that we have (ALN +> IRB) UFR = 1, then A is invertible and A~ = B. If R # I, then A is not invertible, ‘meaning that A~' does not exist (and the matrix B is of no particular interest to our concerns here). (EXAMPLE 4 Finding the Inverse of a Matrix Determine A if A is invertible. 1 0 -2 aA=l4 20 1 1 2 ~10 : Following the foregoing procedure, we have 1 0 100 fai=|4 -2 1/0 1 0 12 oo1 HAR +Ry [iL 100 Sa f iid 0 101 san 6 4 Solution: We have win-fe 38 f wa=[e A282 Chapter Matrix Algebra APPLY IT 15.A group of investors has $500,000 10, invest inthe stocks of three companies Company A sells for $50 a share and thas an expected growth of 13% per year. ‘Company B sells for $20 per share and thas an expected growth of 15% per year. ‘Company C sells for $80 share and has ‘anexpected growth of 10% per year. The ‘group plans buy twice as many shares ‘of Company A as of Company C. I the ‘group's goal is 12% growth per year, hhow many shares ofeach stock should the investors buy? caution\ ‘The method of redoction in ection 6.4 and 65 faster computation tan that of nding a mati iene ‘The first two columns of the last matrix form a reduced matrix different from J. Thus, Ais not invertible. : Now Work Problem 74 Now we will solve a system by using the inverse. (EXAMPLE 5 Using the Inverse to Solve a System Solve the system a = 2, 4x 2+ x x1 + 2x2 — 10x) by finding the inverse of the coefficient matrix. Solution: In matrix form the system is AX = B, where 1 0 -2 A=|4 2 1 ees) is the coefficient matrix. From Example 4(a), 80x = ~7,x) = =17, and.x5 Now Work Problem 27 < It can be shown that a system of 7 linear equations in n unknowns has a unique solution if and only ifthe coefficient matrix s invertible. Indeed, inthe previous example the coefficient matrix is invertible, and a unique solution does in fact exist. When the coefficient matrix is not invertible, the system will have either no solution or infinitely ‘many solutions. While the solution of a system using a matrix inverse is very elegant, we must provide a caution. Given AX = B, the computational work required to find A~? is ‘greater than that required to reduce [A | B]. If there are several equations to solve, all with the same matrix of coefficients but variable right-hand sides, say AX = By, AX = By, ... ,AX = By, then for suitably large k it might be faster to compute A“ than to do & reductions, but a numerical analyst will in most cases still advocate in favor ofthe reductions. For even with A~" in hand, one still has to compute A7'B and, if the order of A is large, this too takes considerable time. (EXAMPLE 6 A Coefficient Matrix That Is Not Invertible Solve the systemSince 1-2 2-1 11 Section 6.8 lnverses 283 The coefficient matrix is 1-21 2-15 fa 100 103 01 0/+--+]o11 oe ooo 1-11 the coefficient matrix is not invertible, Hence, the system cannot be solved by inverses. Instead, another method must be used. In Example 4(a) of Section 6.5, the solution was found to be x = —3r, =-randz = +, where r is any real number (thus providing infinitely many solutions). PROBLEMS 6.6 [In Problems 1-18, ifthe given matrix is invertible find its inverse. (] Now Work Problem 31 <1 For Problems 21-34, ifthe coefficient matrix ofthe system is invertible, solve the system by using the inverse. Ifnot, solve the system by the method of reduction. lor 5y aby Deas 5 a. 5 arty de yt 43743 ae yt x-3y wo tye sw 2r = Dy — ye For Problems 35 and 36, find i(I — A)" for the given matrix A. Se wa-[f iheimene of he max imeved ee A requires [labor hour to paint and $ labor hour to polish; model B reite labor hour foreach proces During cach hour that he {ssembly nes open, tere ae 100 abr hous eel or painting en 8D labor hous fr polishing Hew mary ofeach ‘model can be produced each hour if all the labor hours available are to be utilized?