MATH 2121 Linear algebra (Fall 2017) Lecture 1

Check the course website

http://www.math.ust.hk/~emarberg/Math2121/

for the syllabus and other course details.

1 Notation
Todays lecture corresponds to Section 1.1 in the textbook. See the book for a more detailed discussion!

Throughout, well using the following notation:

C denotes the complex numbers a + b 1.
R denotes the real numbers.
Q denotes the rational numbers p/q.
Z denotes the integers {. . . , 2, 1, 0, 1, 2, . . . }.
N denotes the nonnegative integers {0, 1, 2, . . . }.
Ellipsis (. . . ) notation: we write a1 , a2 , . . . , a7 instead of the full list a1 , a2 , a3 , a4 , a5 , a6 , a7 .
We use the same convention to write a1 , a2 , . . . , an even when n is variable integer.

2 Systems of linear equations

Let x1 , x2 , . . . , xn be variables, where n 1 is some integer.
Let a1 , a2 , . . . , an , b be numbers in R or C.
(Well usually work with real numbers, but nothing is any harder with complex numbers.)
Unlike in calculus, where our favorite variables are x, y, z, in linear algebra we prefer x1 , x2 , x3 , . . . .
Reason: we want to go beyond 3 dimensions.

Definition. We refer to
a1 x1 + a2 x2 + + an xn = b
as a linear equation in the variables x1 , x2 , . . . , xn .
Pn
Notation. Another way of writing this equation is i=1 ai xi = b.
P
The symbol is the Greek letter sigma, for sum.
There are many other equivalent ways of writing the same equation. For example:

a1 x1 + a2 x2 + + an xn b = 0
b = a1 x1 + a2 x2 + + an xn
a1 x1 + a3 x3 + a5 x5 + . . . = b a2 x2 a4 x4 . . .

We treat all of these equations as the same object.

A system of linear equations or linear system is a list of linear equations.

1
MATH 2121 Linear algebra (Fall 2017) Lecture 1

Example.
2x1 x2 + 3x3 = 8
x1 4x4 = 8
x2 = 0
is a linear system in the variables x1 , x2 , x3 .

Definition. A solution of a linear system in variables x1 , x2 , . . . , xn is a list of n numbers (s1 , s2 , . . . , sn ),

with the property that if we plug in x1 = s1 , x2 = s2 , . . . , xn = sn in our equations, we get all true
statements.
Note: if our system contains any false equations like 0 = 1, then it cannot have any solutions.
Two linear systems are equivalent if they have the same set of solutions.

Example. How many solutions can a linear system have?

1. The system
x1 2x2 = 1
x1 + 3x2 = 3
has one solution (s1 , s2 ) = (3, 2).
2. But the system
x1 2x2 = 1
3x1 6x2 = 3
has many solutions: (s1 , s2 ) = (1, 1) or (3, 2) or (5, 3) or . . . .
3. Whereas the system
x1 2x2 = 1
x1 2x2 = 0
has no solutions.

Theorem. A linear system in two variables x1 and x2 has either 0, 1, or solutions.

Remark. The symbol is pronounced infinity. Saying a linear system has solutions is bad style,
since isnt a number. When we say this, we really mean: does not have finitely many solutions.

Proof by geometry. A solution to one equation ax1 + bx2 = c represents a point on a line after we identify
the pair of numbers (x1 , x2 ) with a point in the Cartesian plane.
A solution to a system of 2-variable linear equations represents a point where the lines corresponding to
the equations all intersect.
But a collection of lines all intersect either at 0 points (they dont have a common intersection), 1 point
(the unique point of intersection) or at infinitely many points (in the case when the lines are all the same
line, though they might come from different equations).

Proof by algebra. Suppose the linear system has two different solutions (s1 , s2 ) and (r1 , r2 ).
Define 1 = s1 r1 and 2 = s2 r2 .
The symbol is the Greek letter lambda.
If ax1 + bx2 = c was one of the equations in our system, then by definition as1 + bs2 = c and ar1 + br2 = c.
Taking the difference of these equations gives a(s1 r1 ) + b(s2 r2 ) = 0. In other words, a1 + b2 = 0.

2
MATH 2121 Linear algebra (Fall 2017) Lecture 1

It follows that a(s1 + z1 ) + b(s2 + z2 ) = as1 + bs2 = c for all z.

This works for all the equations in our system.
Therefore (s1 + z1 , s2 + z2 ) is a new solution to our system, for any choice of z.
So the system has infinitely many solutions.

A linear system is consistent if it has one or infinitely many solutions, and inconsistent if it has zero
solutions. Both the algebraic and geometric proofs generalize to any number of variables. (Think about
how to do this!) Therefore:

Theorem. A linear system in n variables is either consistent or inconsistent, i.e., has 0, 1, or infinitely
many solutions.

3 Matrices
A matrix is just a rectangular array of numbers, like these ones:
   
  5 3 7 6 4 3
1 or or .
2 2 1 1 0

We denote a general matrix by

A11 A12 A13 A14
A = A21 A22 A23 A24
A31 A32 A33 A34
Here A23 is pronounced A, two, three. This matrix is 3-by-4: it has 3 rows and 4 columns.
Say that a matrix A is m-by-n or m n if has m rows and n columns.
We usually write Aij (pronounced A, i, j) for the entry in the ith row and jth column of the matrix.
Matrices are useful as a compact way of writing a linear system.
Consider the linear system
x1 2x2 + x3 = 0
2x2 8x3 = 8
5x1 5x3 = 10
Define the coefficient matrix of this system to be

1 2 1
0 2 8
5 0 1

In other words, the matrix A where Aij is the coefficient of xj in the ith equation.
The augmented matrix of the system is

1 2 1 0
0 2 8 8 .
5 0 1 10

Exercise: how would you generalize this definition to any linear system?

3
MATH 2121 Linear algebra (Fall 2017) Lecture 1

4 Solving linear systems

We solve linear systems by adding equations together to cancel variables.

Example. To solve
x1 2x2 + x3 = 0
1 2 1 0

2x2 8x3 = 8 0 2 8 8
5x1 5x3 = 10 5 0 1 10
we first add 5 time eq. 1 to eq. 3 to get
x1 2x2 + x3 = 0

1 2 1 0
2x2 8x3 = 8 0 2 8 8 .
10x2 10x3 = 10 0 10 10 10

We then multiply eq. 2 by 1/2 to get

x1 2x2 + x3 = 0
1 2 1 0

x2 4x3 = 4 0 1 4 4 .
10x2 10x3 = 10 0 10 10 10

We then add 10 times eq. 2 to eq. 3:

x1 2x2 + x3 = 0
1 2 1 0

x2 4x3 = 4 0 1 4 4 .
30x3 = 30 0 0 30 30

Multiple eq. 3 by 1/30:

x1 2x2 + x3 = 0
1 2 1 0

x2 4x3 = 4 0 1 4 4 .
x3 = 1 0 0 1 1

The argument matrix of the last system if triangular : all entries in positions (i, j) with i > j are zero.
(Remember: i is the row, j is the column.)
We can easily solve for x1 , x2 , x3 from a triangular system, working from the bottom up:
The last equation x3 = 1 is already as simple as possible.

Substitute into second equation: x2 4x3 = x1 4(1) = 4 x2 = 0 .

Substitute into first equation: x1 2x2 + x1 = x1 2(0) + (1) = 0 x1 = 1 .

Definition. In solving this system of equations, we performed the following (elementary) row operations
on the augmented matrix of the system:
1. Replacement: replace one row by the sum of itself and a multiple of another row.
2. Scaling: multiple all entries in a row by a nonzero number.
3. Interchange: swap two rows.
Note: we add rows by add the corresponding entries:
     
1 2 3 4 + 7 0 8 4 6 = 1 8+2 7 4+3 7 6+4 7 .

Two matrices are row equivalent if one can be transformed to the other by a sequence of row operations.
Note that each row operation is reversible.

4
MATH 2121 Linear algebra (Fall 2017) Lecture 1

Theorem. If the augmented matrices of two linear systems are row equivalent, then the systems are
equivalent (i.e., have same solutions).

Proof. Heres the idea, minus the details: check that performing one row operation does not change
whether a given (s1 , s2 , . . . , sn ) is a solution to the linear system.

Given a linear system with augmented matrix A, suppose we perform row operations to A until we get
a matrix T with the property that whenever Tij 6= 0 but Ti1 = Ti2 = = Ti,j1 = 0, it holds that
Ti+1,j = Ti+2,j = = Tm,j = 0.
This means: if Tij is the first nonzero entry in the ith row of T going left to right, then Tij is the last
nonzero entry in the jth column of T going top to bottom. For example:

1 6 8 9 0 1 6 8 9 0
T = 0 0 3 2 1 or T = 0 0 3 2 1
0 0 0 4 2 0 0 0 0 2
From T in this form, we can easily determine if the system we started out with is consistent or inconsistent.
If T is the left matrix, the system is consistent: we have

x4 = 4, 3x3 + 2x4 = 1, and x 1 + 6x2 + 8x3 + 9x4 = 0.

Exercise: find a solution!

If T is the right matrix, the system is inconsistent: it includes the equation 0 = 2, from the last row.
In general, a linear system is inconsistent if and only if its augmented matrix can be transformed by row
operations to a matrix with a row of the form
 
0 0 ... 0 q

where q 6= 0. Well prove this next time, after introducing the courses most important algorithm, row
reduction to echelon form.