1.2 Gaussian Elimination P.

Danziger

m Equations in n Unknowns

Given n variables x1, x2, . . . , xn and n+1 constants

a1, a2, . . . , an, b the equation

a 1 x1 + a 2 x2 + . . . + a n xn = b
represents an n − 1 dimensional object in n-space,
called a hyperplane.

We want to consider the situation where we have

m such equations
a11x1 + a12x2 + . . . + a1nxn = b1
a21x1 + a22x2 + . . . + a2nxn = b2
... ...
am1x1 + am2x2 + . . . + amnxn = bm

This is called a system of m (linear) equations in

n unknowns (or variables).

We want to find solutions of this system of equa-

tions.

1
1.2 Gaussian Elimination P. Danziger

Theorem 1 Given a system of m equations in n

unknowns:

• If m < n then the number of parameters in the

solution will be at least n − m.

(Thus if there is a unique solution we must

have m ≥ n.)

• If m > n the system is called overprescribed.

Overprescribed systems either have no solu-

tion or they contain reduncancy. redundancy
means that we can find (m−n) equations which
can be dropped without affecting the solution.

If a system of equations has no solution it is called

inconsistent

If a system of equations has at least one solution

it is called consistent

2
1.2 Gaussian Elimination P. Danziger

Coefficient Matrices and Aug-

mented Matrices
The xi actually carry no information, the system is
completely described by the aij and bi, i = 1, . . . m,
j = 1, . . . , n.

We thus use the matrix of coefficients, wich is

an m × n array containing the coefficients of the
equations.
 
a11 a12 . . . a1n

 a21 a22 . . . a2n 


 ... ... ... ... 

am1 am2 . . . amn

We also have the Augmented Matrix, which in-

cludes the bi on the right:
 
a11 a12 . . . a1n b1

 a21 a22 . . . a2n b2 


 ... ... ... ... ... 

am1 am2 . . . amn bm

The augmented matrix contains all the informa-

tion necessary to solve the system.
3
1.2 Gaussian Elimination P. Danziger

1. Find the matrix of coefficients and the aug-

mented matrix for the following system.
x + 2y − 3z = 1
+ y + z = 1
x + y + z = 0

This system of equations has coefficient ma-

trix:
 
1 2 −3
 0 1 1 
 
1 1 1
and Augmented matrix:
 
1 2 −3 1
 0 1 1 1 
 
1 1 1 0

2. Find the augmented matrix for the following

system.
x + − 2z = 1
+ y − z = 0
This system of equations has Augmented ma-
trix:
!
1 0 −2 1
0 1 −1 0
4
1.2 Gaussian Elimination P. Danziger

3. Given the following augmented matrix find the

original system of equations.
 
1 2 −3
 0 1 1 
 
1 1 0

The system is
x + 2y = −3
y = 1
x + y = 0

This is a system of 3 equations in 2 unknowns.

It is inconsistent (no solution), since by the second

equation y = 1, the third equation then tells us
that x = −1, but then the first equation states
(substituting in x = −1 and y = 1): −1 + 2 = 3,
which is not true.

5
1.2 Gaussian Elimination P. Danziger

Note that each ow of the augmented matrix cor-

responds to one of the original equations.

Each column contains the all the coefficients of

a given variable in the system. We say that this
column corresponds to this variable.

Example 2

 
x + 2y = −3 1 2 −3
y = 1  0 1 1 
 
x + y = 0 1 1 0

The first row corresponds to x, the second cor-

responds to y and the third corresponds to the
constants.

6
1.2 Gaussian Elimination P. Danziger

Elementary Row Operations

There are three basic operations we can preform
on equations, these correspond to Row Operations
on the corresponding matrices.

1. We can multiply an equation by a constant ≡

Multiply a row by a constant.

2. Add a multiple of one equation to another ≡

replace a row by itself plus a multiple of an-
other row.

3. Interchange the order of equations ≡ Inter-

change two rows.

Notation We generally denote the ith row of the

matrix by Ri. Let c be a constant, and 1 ≤ i, j ≤ m
then

Ri → Ri + cRj means replace Row i by row i plus

c times row j.
Ri → cRi means replace row i with c times
row i.
Ri ↔ Rj means interchange row i with row j.
7
 1.2 Gaussian Elimination P. Danziger

Note that preforming any of these operations does

not change the solution to the original system of
equations.
When using row operations always indicate
the operation you have used!
Example 3

1.
   
1 1 3 3 1 1 3 3
R2 → R2 − 2R1
 2 2 3 3  −→  0 0 −3 −3 
   
R3 → R3 − R1
1 1 1 1 0 0 −2 −2

2.
   
1 1 3 3 2 2 3 3
 2 2 3 3  R1 ↔ R2 −→  1 1 3 3 
   
1 1 1 1 1 1 1 1

3.
   
1 1 3 3 2 2 3 3
 2 2 3 3  R2 → 2R2 −→  4 4 6 6 
   
1 1 1 1 1 1 1 1

Never operate on the same row twice in one step.

8
1.2 Gaussian Elimination P. Danziger

Row Echelon Form

Definition 4 1. A matrix is in Row Echelon Form

(REF) if all of the following hold:

(a) Any rows consisting entirely of 0’s appear

at the bottom.

(b) In any non-zero row the first number, from

the left, is a one. Called the leading one or
pivot.

(c) In any two successive non-zero rows the

leading one on top is to the left of the one
on the bottom.

2. A matrix is in Reduced Row Echelon Form

(RREF) if it is in REF (all of the above hold)
and any column containing a leading one is
zero in all other entries.

9
1.2 Gaussian Elimination P. Danziger

Example 5

1. The following are in REF

   
1 1 3 0 1 3 3
 0 1 1   0 0 0 1 
   

0 0 1 0 0 0 0
 
! 1 2
1 1 0
 0 1 
 
0 0 1
0 0
1 indicates a pivot.

2. The following are NOT in REF

   
1 1 0 1 1 3 3
 0 1 1   0 0 0 0 
   
0 1 1 0 0 1 1
 
! 1 2
1 1 3
 0 0 
 
3 0 1
0 1

10
1.2 Gaussian Elimination P. Danziger

3. The following are in RREF

   
1 0 0 0 1 3 0
 0 1 0   0 0 0 1 
   

0 0 1 0 0 0 0
 
! 1 0
1 2 0
 0 1 
 
0 0 1
0 0
1 indicates a pivot. All of the 0’s in these
examples are forced.

4. The following are NOT in RREF

   
1 0 2 0 1 3 0
 0 1 0   0 0 0 1 
   

0 0 1 0 0 0 1
 
! 1 2
1 2 3
 0 1 
 
0 0 1
0 0

11
1.2 Gaussian Elimination P. Danziger

The Gaussian Algorithm

The following Algorithm reduces an m × n matrix
to REF by means of elementary row operations
alone.
1. For Each row i (Ri) from 1 to m

(a) If any row j below row i has non zero entries

to the left of the first non zero entry in row i
exchange row i and j (Ri ↔ Rj ) [Ensure We
are working on the leftmost nonzero entry.]

(b) Preform Ri → 1c Ri where c = the first non-

zero entry of row i. [This ensures that row
i starts with a one.]

(c) For each row j (Rj ) below row i (Each j > i)

i. Preform Rj → Rj − dRi where

d = the entry in row j which is directly
below the pivot in row i. [This ensures
that row j has a 0 below the pivot of row
i.]

(d) If any 0 rows have appeared exchange them

to the bottom of the matrix.
12
1.2 Gaussian Elimination P. Danziger

The Gaussian-Jordan
Algorithm

The following Algorithm reduces an n × m matrix

to RREF by means of elementary row operations
alone.

1. Preform Gaussian elimination to get the matrix

in REF

2. For each non zero row i (Ri) from n to 1 (bot-

tom to top)

(a) For each row j (Rj ) above row i (Each j <

i)

i. Preform Rj → Rj − bRi where

b = the value in row j directly above the
pivot in row i. [This ensures that row j
has a zero above the pivot in row i]

13
1.2 Gaussian Elimination P. Danziger

Gaussian Elimination
To Solve a system of equations we preform the
following steps:

1. Translate the system to its augmented matrix

A.

2. Use Gaussian elimination to reduce A to REF.

Note that the REF form of A has the same
solution set.

3. For each column which does not contain a

pivot introduce a parameter and set the cor-
responding variable equal to that parameter.

4. Substitute the parameters back into the re-

maining non zero equations, this will produce
a solution for the remaining variables.

The number of pivots in the REF of a matrix A is

called the rank of A and is denoted by r or r(A).
Note that the number of parameters in the solu-
tion is equal to n − r. 14
1.2 Gaussian Elimination P. Danziger

Example 6 Solve the following system of equa-

tions. x1 + 2x2 + x3 = 3
x1 + 3x2 + 2x3 = 5
2x2 + x3 = 6

Row reduce augmented matrix to REF

 
1 2 1 3
 1 3 2 5  R2 → R2 − R1
 

 0 2 1 6 
1 2 1 3
 0 1 1 2  R3 → R3 − 2R2
 

 0 2 1 6 
1 2 1 3
 0 1 1 2 
 
0 0 1 −2

For Gaussian elimination use back substitution:

x1 + 2x2 + x3 = 3 (1)
x 2 + x3 = 2 (2)
x3 = −2 (3)
From (3) x3 = −2,
From (2) x2 = 2 − x3 = 2 − (−2) = 4 and
From (1) x1 = 3−2x2 −x3 = 3−2(4)−(−2) = −3.
15
1.2 Gaussian Elimination P. Danziger

Gaussian-Jordan

Instead of using back substitution as in Gaussian

elimination, we can continue reducing until A is in
RREF.

As before, for each column which does not con-

tain a pivot introduce a parameter and set the
corresponding variable equal to that parameter.

But now we may read off the other variables with

no further work.

Example 7 Solve the following system of equa-

tions. x1 + x2 + 3x3 = 3
2x1 + 2x2 + 3x3 = 3
x 1 + x2 + x 3 = 1

We write out the Augmented matrix and use Gaussian-

Jordan to reduce it to RREF.

16
1.2 Gaussian Elimination P. Danziger

 
1 1 3 3
R2 → R2 − 2R1
 2 2 3 3 
 
R3 → R3 − R1
 1 1 1 1 
1 1 3 3
 0 0 −3 −3  R2 → − 1
3 R2
 

 0 0 −2 −2 
1 1 3 3
 0 0 1 1  R3 → R3 + 2R2
 

0 0 −2 −2
1 1 3 3
 0 0 1 1  R1 → R1 − 3R2
 

 0 0 0 0 
1 1 0 0
 0 0 1 1 
 
0 0 0 0

We let the variable corresponding to the column

not containing a pivot (the second column which
corresponds to x2) be the free variable.
Let t ∈ R, set x2 = t, then x3 = 1 (from row 2)
and
x1 = −x2 = −t (from row 1).
Or (x1, x2, x3) = (−t, t, 1)
17
1.2 Gaussian Elimination P. Danziger

Example 8 Solve the following system of equa-

tions.
x1 + 3x2 − 2x3 + 2x5 = 0
2x1 + 6x2 − 5x3 − 2x4 + 4x5 − 3x6 = −1
5x3 + 10x4 + 15x6 = 5
2x1 + 6x2 + 8x4 + 4x5 + 18x6 = 6

The Augmented Matrix is:

 
1 3 −2 0 2 0 0

 2 6 −5 −2 4 −3 −1 

0 0 5 10 0 15 5 
 

2 6 0 8 4 18 6
First leading 1 is in the 1,1 position, already 1.
Get all 0’s below this leading 1 position.
 
1 3 −2 0 2 0 0
R2 −→ R2 − 2R1 
 0 0 −1 −2 0 −3 −1 

R4 −→ R4 − 2R1 0 0 5 10 0 15 5 
 

0 0 4 8 0 18 6
Get leading 1 in second row.
 
1 3 −2 0 2 0 0
 0 0 1 2 0 3 1 
R2 −→ −R2 
 
 0 0 5 10 0 15 5


0 0 4 8 0 18 6
18
1.2 Gaussian Elimination P. Danziger

Get all 0’s below second leading 1.

 
1 3 −2 0 2 0 0
R3 −→ R3 − 5R2 
 0 0 1 2 0 3 1 

R4 −→ R4 − 4R2  0 0 0 0 0 0 0
 

0 0 0 0 0 6 2
Move row of 0’s to bottom:
 
1 3 −2 0 2 0 0
 0 0 1 2 0 3 1 
R3 ↔ R4 
 
0 0 0 0 0 6 2

 
0 0 0 0 0 0 0
Get next leading 1.
 
1 3 −2 0 2 0 0
1  0 0 1 2 0 3 1 
 
R3 −→ R3 
 0 0 1 

6  0 0 0 1 3 
0 0 0 0 0 0 0
Matrix is now in Row Echelon Form.

19
1.2 Gaussian Elimination P. Danziger

Gauss Elimination

We now use back substitution. The Matrix trans-

lates to the following system of equations:
x1 + 3x2 − 2x3 + 2x5 = 0
x3 + 2x4 + 3x6 = 1
1
x6 = 3
For each variable corresponding to a column not
containing a leading 1, we assign a free variable.
Let s, t, r ∈ R.
Let x2 = s, x4 = t, x5 = r.
Then the equations imply: x6 = 1 3
x3 = 1 − 2x4 − 3x6 = 1 − 2t − 1 = −2t So x3 = −2t.
x1 = −3x2 + 2x3 − 2x5 = −3s + 2(−2t) − 2r. So
x1 = −3s − 4t − 2r.
Thus the final solution is:
1)
(x1, x2, x3, x4, x5, x6) = (−3s − 4t − 2r, s, −2t, t, r, 3

20
1.2 Gaussian Elimination P. Danziger

Gauss-Jordan

We continue the algorithm to get the matrix in

Reduced Row Echelon Form.
Get 0’s above rightmost leading 1 (in column 6).
 
1 3 −2 0 2 0 0
 0 0 1 2 0 0 0 
 
R2 −→ R2 − 3R3 
 0 0 1 

 0 0 0 1 3 
0 0 0 0 0 0 0
Get 0’s above next leading 1 (in column 3).
 
1 3 0 4 2 0 0
 0 0 1 2 0 0 0
 

R1 −→ R1 + 2R2 
 0 0 0 0 0 1 1


 3 
0 0 0 0 0 0 0
The Matrix is now in Reduced Row Echelon Form.
The Matrix translates to the following system of
equations:
x1 + 3x2 + 4x4 + 2x5 = 0
x3 + 2x4 = 0
1
x6 = 3
For each variable corresponding to a column not
containing a leading 1, we assign a free variable.
21
1.2 Gaussian Elimination P. Danziger

Let s, t, r ∈ R.
Let x2 = s, x4 = t, x5 = r.
Then the matrix implies: x6 = 3 1

x3 = −2t
x1 = −3x2 − 4x4 − 2x5 = −3s − 4t − 2r.
Thus the final solution is
1 ).
(x1, x2, x3, x4, x5, x6) = (−3s − 4t − 2r, s, −2t, t, r, 3

22