Lecture 02 Recap

A strategy to solve linear systems using

elementary row operations.
Definition of row-echelon form and reduced
row-echelon form of a matrix.

Gaussian Elimination and Gauss-Jordan Elimination.

Lecture 03

Gaussian Elimination (continued)

Homogeneous Linear Systems
What can row-echelon form tell us?
Remember that any linear system has either
(i) no solution (that is, inconsistent); or
(ii) exactly one solution (that is, an unique solution); or
(iii) infinitely many solutions.
By looking at a row-echelon form of the augmented
matrix of the linear system, we can determine which
of the above holds for the linear system.

  Gaussian  
   
   
  Elimination  
   
How can we tell? (consistent vs. inconsistent)
If the augmented matrix of a linear system has a
row-echelon form whose last column is a pivot column,
then the linear system is inconsistent.

    : leading entry
  
  
  In other words, there is a row
 0 0 ... 0  where every entry is zero except
 
the last entry, which is non zero.

last column
How can we tell? (consistent vs. inconsistent)
If the augmented matrix of a linear system has a
row-echelon form whose last column is a pivot column,
then the linear system is inconsistent.
If the augmented matrix of a linear system has a
row-echelon form whose last column is NOT
a pivot column, then the linear system is consistent.

  
  
  
 
 0 0 ... 0 0
 
Can we say more?
But a consistent linear Yes you are right,
system could still row-echelon forms can
behave in two different also tell us how many
ways right? solutions a linear system
have!
How can we tell? (unique vs. infinitely many
solutions)
If the augmented matrix of a consistent linear system
has a row-echelon form where every column
(except the last) is a pivot column, then the linear
system has a unique (that is, exactly one) solution.

 
   Every column here
  
  is a pivot column
 
  
 0 0 0 ... 0 0 

How can we tell? (unique vs. infinitely many
solutions)
If the augmented matrix of a consistent linear system
has a row-echelon form where some column
(other than the last) is NOT a pivot column, then
the linear system has infinitely many solutions.

 * 
 *  Some column here
 0  
  is NOT a pivot column
 
 0  
0 0 
 0 0 ... 0
 Some examples
1 0 2 0
 0 1 3 0
Linear system is inconsistent
  (last column is a pivot column)
0 0 1 0
0 0 1
 0

 1 0 1 4 0 Linear system is consistent

0 1 1 3 1 and has infinitely many
 
0 0 0 2 1 solutions
0 0 0 0 0 (some column other than the
 
0 0  last is a non pivot column)
 0 0 0
 Some examples
 1 0 1 4 0 Linear system is consistent
0 1 1 3 1 and has infinitely many
 
0 0 0 2 1 solutions
0 0 0 0 0 (some column other than the
 
0 0  last is a non pivot column)
 0 0 0

 1 0 1 4 0
Linear system is consistent
0 1 1 3 1
  and has a unique solution
0 0 3 2 1
0 (every column other than the
0 0 1 0
  last is a pivot column)
0 0 
 0 0 0
Notations to use
To ensure that we (your examiners) understand what
you are doing, please use the following notations when
performing elementary row operations:

1) When you want to multiply row i by a non zero

constant c , write cRi .

 1 0 1 2  12 R2  1 0 1 2
 0 2 0 1 0 1 0  12 
   
0 0 1 0 
2R3 0 0 1 0 
 2 
Notations to use
2) When you want interchange rows i and j ,
write Ri  R j .

 0 0 1 2 R1  R2  1 2 0 1
 1 2 0 1  0 0 1 2
   
0 0 1 0  0 0 1 0 
 2  2
Notations to use
3) When you want add k times of row i to row j ,
write R j  kRi .

Remember that in this case,

row j changes but row i does not.

 1 1 1 2 R2  R1  1 1 1 2
 1 2 0 1  0 3 1 1
  R3  2 R1  
 2 0 1 0  0 2 3 4 
 2  2
Can I ‘combine’ operations ‘under
one arrow’?
Answer: Only if there is no confusion or ambiguity.

  R2  R1
  No ambiguity
  R3  2 R1
 
 

  R2  R1 first then R3  2 R2
R2  R1
  OR
  R3  2 R2
  R3  2 R2 first then R2  R1
 
Can I ‘combine’ operations ‘under
one arrow’?
R2  R1 first then R3  2 R2

  R2  R1  
   
   
   
   

R3  2 R2

 
 
 
 
 
Example
What condition(s) must be satisfied by a , b , c such
that the linear system

 x  2 y  3z  a

2 x  6 y  11z  b
 x  2 y  7z  c


has at least one solution?

 Example
For what values of a will the following linear system

 x  2y  3z  4

3x  y  5z  2
4 x  y  (a 2  14) z  a  2


have no solution? Exactly one solution?

Infinitely many solutions?
Example
For what values of a and b will the following linear
system
ax  y  a

x  y  z  1
 y  az  b


have no solution? Exactly one solution?

Infinitely many solutions?
Example (quadric surface)
Find a formula for the quadric surface

ax 2  by 2  cz 2  d

that passes through the points (1,1, 1), (1, 3, 3)

and (2, 0, 2).
Solution (quadric surface)
Since the quadric surface

ax 2  by 2  cz 2  d (*)

passes through (1,1, 1), (1,3,3) and (2, 0, 2),

x  1, y  1, z  1,
x  1, y  3, z  3 and
x  2, y  0, z  2

must satisfy equation (*).

Solution (quadric surface)
ax 2  by 2  cz 2  d (*)

x  1, y  1, z  1, a  b  c  d

x  1, y  3, z  3 and  a  9b  9c  d
x  2, y  0, z  2  4a  4c  d


a  b  c  d  0

  a  9b  9c  d  0
 4a  4c  d  0

Solution (quadric surface)
a  b  c  d  0

 a  9b  9c  d  0
 4a  4c  d  0


 1 1 1 1 0 R2  R1  1 1 1 1 0
 1 9 9 1 0 0 8 8 0 0
   
 4 0 4 1 0  R3  4 R1  0 4 0 3 0 
 

R3  12 R2  1 1 1 1 0
0 8 8 0 0
 
0 0 4 3 0 

Solution (quadric surface)
 1 1 1 1 0 1
R2  1 1 1 1 0
0 8 8 0 0
8
0 1 1 0 0
   
0 0 4 3 0 
1
R3 0 0 1 3 0 
 4  4

R1  R3  1 1 0  74 0
0 1 0 3 0
R2  R3  4 
0 0 1 3 0 
 4

R1  R2  1 0 0 1 0
0 1 0 3 0
 4

0 0 1 3 0 
 4
Solution (quadric surface)
ax 2  by 2  cz 2  d

a  t
 1 0 0 1 0 b
0 1 0 3   34 t
0 
 4
   34 t
0 0 1 3 0  c
 d  t, t .
4

Let d  4, then a  4,b  3, c  3, d  4 is a solution.

An equation of the quadric surface is

4 x 2  3 y 2  3 z 2  4.
Think for a while…
ax 2  by 2  cz 2  d

Let d  4, then a  4,b  3, c  3, d  4 is a solution.

An equation of the quadric surface is

4 x 2  3 y 2  3z 2  4.

If I choose another value for d , does that mean

I get another equation of another quadric surface?
Let d  8, then a  8,b  6, c  6, d  8 is a solution.
8 x 2  6 y 2  6 z 2  8  4 x 2  3 y 2  3z 2  4.
Discussion
Suppose we have a consistent linear system involving
three unknowns x , y , z at most 3 leading entries
row-echelon form    
R   
   
   
   
inconsistent
1) if 3 non zero rows (that is, 3
leading entries)
 unique solution
 solution set is a point
Discussion
Suppose we have a consistent linear system involving
three unknowns x , y , z
     
R     
     
     
     
2) if 2 non zero rows (that is, 2
leading entries)
 infinitely many solutions
 general solution has one parameter
 solution set is a line
Discussion
Suppose we have a consistent linear system involving
three unknowns x , y , z
     
R     
     
     
     
3) if 1 non zero row (that is, 1
leading entry)
 infinitely many solutions
 general solution has two parameters
 solution set is a plane
Example
x  y  2z  1
 Three planes in xyz  space
x  y  z  0
x  y  z  2

1 1 2 1   1 0 0 23 
1 1 1 0  0 1 0 1 
   
1 1 1 2  0 0 1 1 
   3

3 leading entries

x  23 , y  1, z   13 is the unique solution

 Example
Four planes in xyz  space
 x  y  2z  1
x  y  z  0
 x  1
2  12 t
 
 z  1 y  1
 32 t
2 x
2
z  t
3 x  y  1  t

1 1 2 1  1 0 12 12 
 1 1 0 1 3 1 
1 0   
  2 2

2 0 1 1 0 0 0 0
 3 1 0 0 0 0
 0 1   
2 leading entries
Solution is a straight line consisting of points:
( 12  12 t , 12  23 t , t ) for all t 
 Example
 x  y  2z  1

3 x  3 y  6 z  3  x  1  s  2t

Two planes in xyz  space y  s
z  t s ,t 

x  y  2z  1
1 1 2 1 1 1 2 1
 3 3 6 3 0 0 0 0
   
1 leading entry
Solution is a plane consisting of points:
(1  s  2t , s , t ) for all s , t 
Definition (Homogeneous Linear Systems)
A linear system is said to be homogeneous if it
has the following form:

 a11 x1  a12 x2  ...  a1n xn  0

 a x  a x  ...  a x  0
 21 1 22 2 2n n

 : : : :
am1 x1  am 2 x2  ...  amn xn  0

a11 ,a12 ,...,amn are real constants.

Something special
Recall that any linear system behaves in exactly one of
the following three ways:
Linear system

no solutions exactly one infinitely many

(inconsistent) solution solutions

(consistent)
Something special
no solutions exactly one infinitely many
(inconsistent) solution solutions

(consistent)
 a11 x1  a12 x2  ...  a1n xn  0
 a x  a x  ...  a x  0
 21 1 22 2 2n n

 : : : :
am1 x1  am 2 x2  ...  amn xn  0

What if we let x1  0, x2  0,..., xn  0?

Something special
no solutions exactly one infinitely many
(inconsistent) solution solutions

(consistent)
 a11 0  a12 0  ...  a1n 0  0
a 0  a22 0  ...  a2 n 0  0
 21

 : : : :
am1 0  am 2 0  ...  amn 0  0

x1  0, x2  0,..., xn  0 is ALWAYS a solution!

Definition (Trivial and non-trivial solution)
A homogeneous linear system is always consistent.

x1  0, x2  0,..., xn  0 is called the trivial solution

of the homogeneous linear system.
Any other solution (if there exists) is called a
non-trivial solution.

no solutions exactly one infinitely many

(inconsistent) solution solutions

only trivial trivial + non-trivial

solution solutions
Example
How many solutions does the following homogeneous
linear system have?

2 x  y  3z  0

 x  2 y  3z  0
x  y  4z  0


Note that the above linear system represents 3 planes

in the three dimensional space, each plane contains
the origin (that is, contains the point x  0, y  0, z  0).
Example (2 variables)
l1 and l2 are two lines in the xy plane
passing through the origin.

 a1 x  b1 y  0 (l1)

a2 x  b2 y  0 (l2 )

y y

x x
Example (2 variables)
 a1 x  b1 y  0 (l1)

a2 x  b2 y  0 (l2 )

The linear system has only the trivial solution if

and only if the two lines are not the same.

The linear system has non-trivial solutions if

and only if the two lines are identical.
3 variables
a1 x  b1 y  c1 z  0 ( p1)

a2 x  b2 y  c2 z  0 ( p 2)

origin

Infinitely many solutions

(trivial + non-trivial)
3 variables and 3 planes
p1 , p2 and p3 are three planes in the three dimensional
space, each containing the origin (x  0, y  0, z  0).

 a1 x  b1 y  c1 z  0 ( p1)

a2 x  b2 y  c2 z  0 ( p2 )
a x  b y  c z  0
 3 3 3 ( p3 )

Discuss the relative positions of the planes such that

the linear system has (i) only the trivial solution;
(ii) non-trivial solutions.
What similarities can you find?
x  y  2z  0  1 1 2 0
  2 3 9
2 x  3 y  9 z  0  0 

 x1  x2  2 x3  4 x4  0

2 x1  x2  3x3  5 x5  0
3x  2 x  x  3x4  x5  0
 1 2 3

 1 1 2 4 0 0
 2 1 3 0 5 0
 
 3 2 1 3 1 0 

 What similarities can you find?
What is the maximum number of leading entries in
a row-echelon form of each of the following
augmented matrices?

 1 1 2 0  
 2 3 9 0   
  

 1 1 2 4 0 0  
 2 1 3 0 5 0  
   
 3 2 1 3 1 0   
  
 What similarities can you find?
 1 1 2 4 0 0  
 2 1 3 0 5 0  
   
 3 2 1 3 1 0   
  

If a homogeneous linear system has more unknowns

than equations...

... there will ALWAYS be non-pivot columns in a

row-echelon form of the augmented matrix
(other than the last column).
 infinitely many solutions  non-trivial solutions
What similarities can you find?
x  y  2z  0

2 x  3 y  9 z  0

 x1  x2  2 x3  4 x4  0

2 x1  x2  3x3  5 x5  0
3x  2 x  x  3x4  x5  0
 1 2 3

Homogeneous linear systems with more unknowns

than equations always has infinitely many solutions.
End of Lecture 03

Lecture 04
Introduction to matrices
Matrix operations (till Theorem 2.2.22)