3.1 Elem. Matrix Operations and Elem.

Matrices

Definition
Let A ∈ Mm×n (F ) . The following are called elementary row [column]
operations.
Type 1: interchanging any two rows [columns] of A
Type 2: multiplying a row [column] of A by a nonzero scalar
Type 3: adding any scalar multiple of a row [column] to another row [column]

Definition
An elementary matrix of Type 1, 2, or 3, is an n × n matrix obtained from In by
performing an elementary row operation of Type 1, 2, or 3, respectively.
3.1 Elem. Matrix Operations and Elem. Matrices

Theorem (3.1a)
Let A, B ∈ Mm×n (F ) , with B obtained from A by performing an elementary row
operation. Then there exists E ∈ Mm×m (F ) such that B = EA. In fact, E is
obtained from Im by performing the same elementary row operation as that
performed on A to obtain B. Conversely, if E is an elementary m × m matrix,
then EA is the matrix obtained from A by performing the same elementary
operation as that obtained from Im to get E .

Theorem (3.1b)
Let A, B ∈ Mm×n (F ) , with B obtained from A by performing an elementary
column operation. Then there exists E ∈ Mn×n (F ) such that B = AE . In fact, E
is obtained from In by performing the same elementary column operation as that
performed on A to obtain B. Conversely, if E is an elementary n × n matrix, then
AE is the matrix obtained from A by performing the same elementary operation
as that obtained from In to get E .
3.1 Elem. Matrix Operations and Elem. Matrices

Theorem (3.2)
Elementary matrices are invertible, and the inverse of an elementary matrix is an
elementary matrix of the same type.
3.2 Rank of a Matrix and Matrix Inverses

Definition
Let A ∈ Mm×n (F ). The rank of A is the rank of LA : F n → F m .

Thus, an n × n matrix A is invertible iff rank(A) = n.

Theorem (3.3)
Let V and W have finite dimension with ordered bases B and C , resp. If
T : V → W a linear transformation, then rank(T ) = rank([T ]C
B ).

Theorem (3.4)
Let A be an m × n matrix, P and Q invertible matrices of sizes m × m and n × n,
resp. Then
1 rank(AQ ) = rank(A)
2 rank(PA) = rank(A)
3 rank(PAQ ) = rank(A)
3.2 Rank of a Matrix and Matrix Inverses

Corollary
Elementary row and column operations are rank preserving.

Theorem (3.5)
Let A be an m × n matrix with columns a1 , a2 , . . . , an . Then
rank(A) = dim(span({a1 , a2 , . . . , an })).

Theorem (3.6)
Let A be an m × n matrix with rank r . Then r ≤ m, r ≤ n, and A can be
transformed into matrix  
Ir O1
D=
O2 O3
where O1 , O2 , and O3 are zero matrices.
3.2 Rank of a Matrix and Matrix Inverses
Corollary
Let A be an m × n matrix with rank r . Then there exist invertible matrices B and
C of sizes m × m and n × n, resp., such that D = BAC , where
 
Ir O1
D=
O2 O3

is the m × n matrix where O1 , O2 , and O3 are zero matrices.

Corollary
Let A be an m × n matrix. Then
1 rank(At ) = rank(A)
2 Let a1 , a2 , . . . , an be the rows of A. Then
rank(A) = dim(span({a1 , a2 , . . . , an })).

Corollary
Let A be an invertible n × n matrix. Then A can be written as the product of
elementary matrices.
3.2 Rank of a Matrix and Matrix Inverses

Theorem (3.7)
Let V , W and Z be finite dimensional vector spaces and T : V → W and
U : W → Z linear transformations. Let A and B be matrices such that AB is
defined. Then
1 rank(UT ) ≤ rank(U )
2 rank(UT ) ≤ rank(T )
3 rank(AB ) ≤ rank(A)
4 rank(AB ) ≤ rank(B )
3.2 Rank of a Matrix and Matrix Inverses

Definition
Let A and B be matrices with sizes m × n and m × p, resp. The augmented
matrix (A | B ) is the m × (n + p ) matrix formed with the first n columns equal
to the columns of A and the remaining p columns from the columns of B.

Theorem
Let A be an n × n matrix. Then A is invertible iff (A | In ) can be transformed by
elementary row operations into a matrix of the form (In | B ) ; in this case,
B = A−1 .Thus, A is not invertible iff upon applying elementary row operations on
(A | In ) , a row consisting entirely of 0’s is obtained on the left part of the
augmented matrix.
3.3 Systems of Linear Equations

An m × n linear system is a system of m linear equations in n unknowns over the

field F :
a11 x1 + a12 x2 + · · · + a1n xn = b1
a21 x1 + a22 x2 + · · · + a2n xn = b2
.. .. ..
. . .
am1 x1 + am2 x2 + · · · + amn xn = bm
where aij , bi are scalars in F for each i and j, 1 ≤ i ≤ m, 1 ≤ j ≤ n. The m × n
matrix  
a11 a12 · · · a1n
 a21 a22 · · · a2n 
A= .
 
.. .. .. 
 .. . . . 
am1 am2 ··· amn
is called the coefficient matrix.
3.3 Systems of Linear Equations

   
x1 b1
 x2   b2 
If x =   and b =  , then the linear system is equivalent to the
   
.. ..
 .   . 
xn bn
 
s1
 s2 
matrix equation Ax = b. A solution s =   is an element of F n for which
 
..
 . 
sn
As = b; the set of all solutions is called the solution set. The linear system is
said to be consistent if the solution set is not empty; inconsistent, otherwise.
3.3 Systems of Linear Equations

Definition
The m × n linear system Ax = b is homogeneous if b = 0; nonhomogeneous,
otherwise.
Note: A homogeneous system has at least one solution—the zero vector.

Theorem (3.8)
Given the homogeneous linear system Ax = 0. Let K be the solution set of the
system. Then K = N (LA ) ; hence, K is a subspace of F n , and
dim (K ) = n − rank(LA ) = n − rank(A).

Corollary
If m < n, then the m × n linear system Ax = 0 has a nonzero solution.
3.3 Systems of Linear Equations

Definition
Given a nonhomogenous system Ax = b (b 6= 0), we call the equation Ax = 0 the
associated homogeneous system.

Theorem (3.9)
Let K = {x | Ax = b } , KH = {x | Ax = 0}— that is, K is the solution set of
the linear system Ax = b while KH the solution set of the associate homogeneous
system. Then for any s ∈ K ,

K = { s } + KH = { s + k | k ∈ KH } .

Theorem (3.10)
Let Ax = b be an n × n linear system. Then A is invertible iff Ax = b has a
unique solution.
3.3 Systems of Linear Equations

Definition
Given the linear system Ax = b, the matrix (A | b ) is called the augmented
matrix of the linear system.

Theorem (3.11)
The linear system Ax = b is consistent iff rank (A) = rank (A | b ) .
3.4 Systems of Linear Equations - Computations

Definition
Two linear systems are said to be equivalent if they have the same solution set.

Theorem (3.13)
Let Ax = b be an m × n linear system, C an invertible m × m matrix. Then the
system (CA) x = Cb is equivalent to Ax = b.

Corollary
Let Ax = b be an m × n linear system. If (A0 | b 0 ) is obtained from (A | b ) by a
finite number of elementary row operations, then A0 x = b 0 is equivalent to
Ax = b.
3.4 Systems of Linear Equations - Computations
To solve a linear system Ax = b by Gaussian elimination:
1 From the augmented matrix (A | b ) , create a 1 in the first row of the
leftmost nonzero column, using Type 1 and/or Type 2 elementary row
operations.
2 Using Type 3 elem. row operations, use the first row to obtain 0’s in the
remaining positions of the leftmost nonzero column.
3 Create a 1 in the next row in the leftmost possible column, without using
previous row(s).
4 Use Type 3 elementary row operations to obtain 0’s below the 1 created in
the previous step.
5 Repeat steps 3 and 4 on each succeeding row until no nonzero rows remain.
6 Work upward beginning with the last nonzero row, and add multiples of each
row to the rows above.
7 Repeat step 6 for each preceding row until it is performed with the second
row.
8 The solution can then be obtained by inspection.
3.4 Systems of Linear Equations - Computations

Definition
A matrix is in reduced row echelon form (rref) if the following are satisfied:
1 Any row containing a nonzero entry precedes any row in which all the entries
are zero (if any).
2 The first nonzero entry in each row is the only nonzero entry in its column.
3 The first nonzero entry in each row is 1 and it occurs in a column to the right
of the first nonzero entry in the preceding row.

Theorem (3.14)
Gaussian elimination transforms any matrix into its rref.

Note: If at any point in applying the Gaussian elimination, one obtains a row
whose only nonzero entry is the one in the last column, then the system is
inconsistent.
3.4 Systems of Linear Equations - Computations

Theorem (3.15)
Let the linear system Ax = b consist of r nonzero equations in n unknowns. If
rank (A) = rank (A | b ) and (A | b ) is in rref, then
1 rank (A) = r
2 If the general solution of the system obtained by the Gaussian elimination is

s = s0 + t1 u1 + t2 u2 + · · · + tn−r un−r ,

then {u1 , u2 , . . . , un−r } is a basis for the solution set of the associated
homogeneous system, and s0 is a solution of Ax = b.
3.4 Systems of Linear Equations - Computations

Theorem (3.16)
Let A be an m × n matrix of rank r , r > 0, and B is the rref of A. Then
1 B has exactly r nonzero rows
2 For each i, 1 ≤ i ≤ r , there is a column bji of B such that bji = ei .
3 The columns of A numbered j1 , j2 , . . . , jr are linearly independent.
4 For each k, 1 ≤ k ≤ n, if the kth column of B is ∑ri=1 di ei , then the kth
column of A is ∑ri=1 di aji .

Corollary
The rref of a matrix is unique.