‘lementary Matrices In this special handout of material not contained in the text, we introduce the concept of elementary matrix. Elementary matrices are useful in several ways that will be shown in this, handout. One important use that we will examine first is that elementary matrices can be used to carry out elementary row operations. This means that if you want (o interchange two rows, or multiply a row by a constant and add it to another row, or multiply some row by a non-zero constant, it can be done using an elementary matrix. The definition given next shows that elementary matrices can be created by applying elementary row operations to the identity matrix. Definition: An nxn matrix E is an elementary matrix if it can be obtained by performing a single elementary row operation on the identity matrix In. Some examples of elementary matrices for n = 3 and for each of the elementary row operations are 001 B= ° A 0 |» hich is obtained by interchanging the first row and third rows of fs, 100 By =| 0 10} which is obtained by multiplying the frst ow off by 1, and 100 B= | 219] whicnis obtained by multiplying the fist row of ly by -2 and adding it t0 the second row of 13. ‘The subscripts on E have no particular meaning but are just used to distinguish one elementary matrix from the next, An elementary row operation can be carried out by its corresponding elementary matrix through matrix multiplication. For example, suppose we wish fo interchange the first and third rows ofthe matrix A= This can be done by multiplying A on the left by the elementary matrix E), given above, to yield ‘Suppose we then wish to perform the elementary row operation of multiplying the first row by 1, we can do this by multiplying by E which then gives ‘We can then perform on this result the elementary row operation of multiplying the first row by -2 and adding it to the second row using E3 to obtain 130 FyEpE,A- | 09 2 3 EQ Ey old [four goal is to obtain rref for A, then we would continue with the following elementary matrices to 00 © 1 F and applying tis to the coe previous result we get0 | 1 then to multiply row 2 by 3 and add it to row 1 we do this to fy to obtain E 5 = [ 4 and applying this to the previous result we obtain con a es Es Bg B3 Ey Ey A = [ eno a then to multiply row 3 by 1/2 we use the elementary matrix Fy = [ . ‘ 00 1-1 oor coe ene Fg Bs Eq F3 EE, A = [ | to obtain. coe —_ eon 1 B7 Eg Bs By BB) By A = [: as ene and finally, to add row 3 to row 2 we use Eg = [ CS coe o-ooNote that the last equation is. Eg E7 Bg Es Eq 3 Eg Ey A = 1g. Thus the product Eg E7 Eg Bs Eq E3 E> Ey must be the inverse of A and this gives us another way of computing, the inverse of A. Note that carrying out the product, 2 Eg E7 Bg Bs Bg Ey Ey Ey = [ a = al Elementary matrices have another important property given in the next theorem. ‘Theorem 1. Every elementary matrix has an inverse, which is also an elementary matrix. Proof: We won't give a formal proof but will suggest two reasons why it is true. First, since every elementary row operation can be reversed, one would expect that this can be represented by an elementary row operation which is an elementary matrix. Second, itis easy to construct the inverse of an elementary matrix. Here are the three cases: 1. IfE interchanges two rows, then E is its own inverse. (Why?) 2. If E multiplies a row by a non-zero constant c, then its inverse is the elementary matrix that multiplies a row by Ic. * 3. If E multiplies row i by c and adds it to row j, the its inverse is the elementary matrix that multiplies row i by -c and adds it to row j For the elementary matrices used in the example above here are their inverses which are also elementary matrices100 130 , [too ryt =| 001 s O10 Egy! =| 010 2 orof’ “s) coi} 002 103 1 E,y'=| 010 B oral. 2 oor}? @8) oor ‘Theorem 2, (a). Every mxn matrix A can be decomposed as a product of elementary matrices and a matrix R that is rref, that is, a where Ej are clementary matries and R is ref. (b). If Ais nxn and has an inverse, then Ris the identity matrix and A is, decomposable into the-product of elementary matrices, that is, Q Proof of (a). Since every matrix A can be put into reef by row operations and since every row ‘operation can be represented by an elementary matrix we have, where R is rref, Ey... BAS R @ By Theorem 1 each Ej has an inverse that is an elementary matrix. Ifwe let E ; denote its inverse and multiply equation (3) on both sides by Ey... Egy tain equation (1), Proof of (b). If A has an inverse, then R is the 2). End of proof jentity in equation (1) and so we obtain equation I From the example before Theorem 2, we have that (Ey)! ... (Eg)! = AElementary Matrices and Determinants ‘The determinants of elementary matrices are very easy to compute and together with Theorem 2 an be used to prove some very important properties of nxn matrices and their determinants, The ‘main theorem is, Theorem 3. The determinant of an elementary matrix E is given as follows: (a). det (E) = -1, if E interchanges two rows. (b). det (E) = ¢, if E multiplies a row by a non-zero constant c. (©). det (E) = 1, if IfE multiplies row i by ¢ and adds it to row j Proof of (a). This follows from Theorem 3.2 in the text. However, it is easy to prove it based on the fact that det(l,) = a} 1 492. .--yq » Where ajj ~ 1, so when we interchange rows i and j, the term ay] 897 + Ajj Aj yy i the determinant of E except possibly forts sign, But we see that the columns have been permuted and thus the sign would change to -1 Proof of (b). E is just the same as Ij, except for a non-zero constant ¢ in place of aj togive det(E)= ay agg ....¢... apq =e since all of the other terms equal 1 Proof of (c). The elementary matrix E is that same as Iy except for one element ¢ that is not on the main diagonal. This value ¢ is multiplied by a 0 in any term that appear in the determinant of E since the only other non-zero terms in Fare the I's on the main diagonal. End of theorem. The proof above will be clearer if you take a few minutes to compute the determinants of typical 3x3 elementary matrices, Theorem 4, If E is an elementary matrix and B is an nxn matrix, then det(E B) = det(B)det(B). Proof. Case 1. If E interchanges two rows, then by Theorem 3.2 in the text, det(E B) = - det(B),Since det(E) =-1 by Theorem 3, then det(E B)= det(E) det(B). Case 2. If E multiplies a row by a non-zero constant, then by Theorem 3.5 in the text, det(E B) =edet(B), Again, by Theorem 3, det(E) =e, so det(EB) = det(E) det(B). Case 3. If E multiplies row i by c and adds it to row j, then by Theorem 3.6 in the text, det(E B) = dei(B). Again, by Theorem 3, det(E) = 1, so det(E B) = det(E) det(B). Elementary matrices can now be used to prove that det(AB) = det(A) det(B). ‘Theorem 5. (Theorem 3.8 in the text). If A and B are nxn matrices, then det(AB) = det(A) det(B). Proof. We will give part of the proof in a special case where A and B are non-singular and leave the remainder of the proof to the reader. Suppose A and B are non-singular. ‘Then by Theorem 20) A= Ey... 8, and B= B,.. By andso AB= £).. BoB, .. By where all of the E's are elementary matrices. By Theorem 4, da(AB) = det (Ej) dey. EB... Bp We can repeat using Theorem 4 to obtain det(AB) = det (Ej) det( 2)... det(f, ydet( ,)... det But then itis easy to show that det(A)= det (E 1) det(f 2)... det(E ,) and det(B) = det(f ;)... det(Ei,) which proves the result when A and B are non-singular. Exercises1. a, Determine the elementary matrix for each step in Example 5 in Sec. 1.6 of the text. b. Show A! asa product of elementary matrices and verify by multiplying out the matrices that the product is the inverse of A. ©. Show A as a product of elementary matrices and verify by multiplying out the matrices that the product is A. 2. a. Determine the elementary matrix for each step in Example 6 in Sec. 1.6 of the text. b. Verify Theorem 2(a) for A. 3. Show that Theorem 3 holds for the elementary matrices E ... Eg in the main example in this handout 4. Compute the determinant of A in the main example in this handout by using only elementary ‘matrices in conjunction with the Theorems in this handout,