Row operations can be used to transform a matrix A into an equivalent matrix B that represents the same system of equations. Elementary row operations (EROs) include adding a multiple of one row to another, multiplying a row by a non-zero constant, and swapping rows. Applying EROs produces an equivalence relation - if A can be transformed into B through EROs, they are row equivalent and represent the same system. The solution sets of row equivalent matrices are also the same.
Row operations can be used to transform a matrix A into an equivalent matrix B that represents the same system of equations. Elementary row operations (EROs) include adding a multiple of one row to another, multiplying a row by a non-zero constant, and swapping rows. Applying EROs produces an equivalence relation - if A can be transformed into B through EROs, they are row equivalent and represent the same system. The solution sets of row equivalent matrices are also the same.
Row operations can be used to transform a matrix A into an equivalent matrix B that represents the same system of equations. Elementary row operations (EROs) include adding a multiple of one row to another, multiplying a row by a non-zero constant, and swapping rows. Applying EROs produces an equivalence relation - if A can be transformed into B through EROs, they are row equivalent and represent the same system. The solution sets of row equivalent matrices are also the same.
Row operations can be used to transform a matrix A into an equivalent matrix B that represents the same system of equations. Elementary row operations (EROs) include adding a multiple of one row to another, multiplying a row by a non-zero constant, and swapping rows. Applying EROs produces an equivalence relation - if A can be transformed into B through EROs, they are row equivalent and represent the same system. The solution sets of row equivalent matrices are also the same.
For simplicity, we shall first study homogeneous systems of equations. The secret is to configure the rows of the coefficient matrix A so as to (more or less) read off the solutions.
L4.1 Row operations. Consider a homogeneous linear system in matrix form
Ax = 0,
with
A Rm,n ,
x Rn,1 .
In this case, each equation is completely determined by the corresponding row of A, and we can encode the equations by the m rows
r1
r2
rm of A. In this notation, with m = 4, the scheme
r1
r2 2r1 r3
3r4
represents an equivalent system of equations; we have merely subtracted twice the first from the second and multiplied the last by 3. These changes will not affect the values of any solution (x1 , . . . , xn ). We are also at liberty to change the order in which we list the equations. Our aim is to use such changes to simplify the system. Definition. Let A be a matrix of size mn. An elementary row operation (ERO) is one of the following ways in which a new matrix of the same size is formed from A: (i) add to a given row a multiple of a different row, (ii) multiply a given row by a nonzero constant, (iii) swap or interchange two rows. In symbols, we can denote the operations that we have just described by (i) ri 7 ri + arj , i 6= j , (ii) ri 7 cri , c 6= 0, (iii) ri rj . In practice, it is often convenient to take a to be negative; in particular (i) includes the act of subtracting one row from another: r i 7 ri rj (but it is essential that i 6= j otherwise we would effectively have eliminated one of the equations).
L4.2 Solving a homogeneous system. Let us show how EROs can be used to solve the linear system
x1 + x2 + 2x3 + 3x4 = 0
5x1 + 8x2 + 13x3 + 21x4 = 0
(1)
34x1 + 55x2 + 89x3 + 144x4 = 0
written before. (The choice of Mole coefficients will keep the arithmetic manageable.) 13
We shall apply EROs to convert A into a matrix that is roughly triangular, and then solve the resulting system. A =
r2 5r1
1 3 r2
r3 34r1
1 21 r3
r3 r 2
1 5 34
1 0 34
1 0 34
1 0 0
1 0 0
1 0 0
1 2 3 8 13 21 55 89 144
1 2 3 3 3 6 55 89 144
1 2 3 1 1 2 55 89 144
1 2 3 1 1 2 21 21 42
1 2 3 1 1 2 1 1 2
1 2 3 1 1 2 0 0 0
On the left, we jot down (in abbreviated form) the operations used. It is not essential to do this, provided the operations are carried out one at a time; errors occur when one tries to be too ambitious! It follows from the last matrix that (1) has the same solutions as the system
x1 + x2 + 2x3 + 3x4 = 0 x2 + x3 + 2x4 = 0 But one can see at a glance how to solve this; we can assign any values to x 3 and x4 which will then determine x2 (from the second equation) and then x 1 (from the first). Suppose that we set x3 = s and x4 = t (it is a good idea to use different letters to indicate free variables); then x2 = s 2t
x1 = (s 2t) 2s 3t = s t.
The general solution is
(x1 , x2 , x3 , x4 ) = (s t, s 2t, s, t), or in column form
1 1 s t
s 2t = s 1 + t 2 , x= 0 1
s 1 0 t
where s, t are arbitrary. The set of solutions is therefore
L {u, v} ,
where
1 1
u= 1 , 0
14
1 2
v= 0 . 1
(2)
We shall see that the solution set of any homogeneous system is always a linear combination of this type. Exercise. Compute u v , and explain why this is also a solution.
L4.3 Equivalence relations. Applying EROs produces a natural relation on the set of matrices of any fixed size. Definition. A relationship between elements of a set is called an equivalence relation if (E1) A A is always true, (E2) A B always implies B A, (E3) A B and B C always implies A C . Observe that these three conditions are satisfied by equality =. On the set of real numbers, having the same absolute value is an equivalence relation, but 6 is not. But we are more interested in sets of matrices. Definition. From now on, we write A B to mean that B is a matrix obtained by applying one or more EROs in succession to A. Proposition. This does define an equivalence relation, and if A B we say that A and B are row equivalent. Proof. By definition A B and B C imply that A C . Obviously A A (take (i) with a = 0 or (ii) with c = 1, both of which are permitted). The condition (E2) is less obvious. But each of the three operations is invertible; it can be undone by the same type of operation. For example, r 1 7 r1 ar2 by r1 7 r1 + ar2 . So if A B then we can undo each ERO in the succession one at a time, and B A. QED Example. If B is a matrix which has the same rows as A but in a different order, then A B . This is because any permutation of the rows can be obtained by a succession of transpositions, i.e. EROs of type (iii). For example, let
1 2 3 4 5 6 4 5 6 A = 4 5 6 , B = 1 2 3 , C = 7 8 9 . 7 8 9 7 8 9 1 2 3 Then B is obtained from A by r1 r2 ;
C is obtained from B by r2 r3 ,
so A C even though it is not possible to pass from A to C by a single ERO. Of course, it is
also true that A B , B C , and C A. Exercise. In this example, can C be obtained from A by a succession of EROs of type (i)?
The theory of this lecture can be summarized by the
Proposition. Suppose that A B . Then the column vector x is a solution of Ax = 0 iff it is a solution of Bx = 0. In other words, the homogeneous linear systems associated to row equivalent matrices have the same solutions. This follows because each individual ERO leaves unchanged the solution space. 15
L4.4 Further exercises.
1. Use EROs to find the general solution of the linear system
2x 2y + z + 4t = 0
x y 4z + 2t = 0 x + y + 3z 2t = 0
3x 3y + z + 6t = 0. 2. Show that the matrix
1 2 1 2 A = 1 3 1 3 1 8 1 8
is row equivalent to one with one row null.
3. Recall that a square matrix P Rn,n is called invertible if there exists a matrix P 1 of the same size such that P 1P = In = P P 1 . Two matrices A, B are said to be similar if there exists an invertible matrix P such that P 1AP = B . Prove that being similar is an equivalence relation on the set Rn,n . (Hint: you will need the fact that (P 1 )1 = P .)
