You are on page 1of 6

6-6-2008

Mathematical Induction
Induction is a method for proving an infinite sequence of statements. In its easiest form, you start by
proving that the first statement is true. Next, you show that if an arbitrary statement in the list is true,
then the next statement in the list is true.

Induction is a consequence of the Well-Ordering Axiom for the positive integers.


The Fibonacci numbers are the numbers in the sequence 1, 1, 2, 3, 5, 8, 13, . . . . Each number (after
the first two) is the sum of the two numbers before it.

Mathematical induction is a method for proving a result which consists of a sequence of statements:

statement1 , statement2 , statement3 , . . . .

It applies when the statements in the sequence are related or similar. The idea is that if you can get
started, and if you can make each step from one statement to the next, you can prove all of them.
Induction depends on the following fundamental property of the positive integers:

Well-Ordering. Every nonempty set of positive integers has a smallest element.

This may seem obvious, and it is. There is no question of proving the Well-Ordering Principle; its
essentially an axiom which defines the positive integers. As simple as it sounds, it has very important
consequences: Induction, which Ill discuss here, is one, and the Division Algorithm is another.
Induction will work like this. Suppose you have a sequence of statements:

statement1 , statement2 , statement3 , . . . .

Sometimes the first statement will be statement0 rather than statement1 . The numbering can start with
any integer, but often the first statement is numbered 0 or 1.
First, do the basis step:

Prove the first statement lets say that its statement1 .

(Sometimes the basis step involves proving more than one statement. Ill give an example of this later.)
Next, do the induction step:

Assume statementn is true, and try to prove statementn+1 .

When youve completed both steps, all the statements statement1 , statement2 , statement3 , . . . are true,
by induction.
The statement statementn which youre assuming to be true is called the induction hypothesis or
induction assumption.
The procedure Ive given is often called simple induction. You may also do a general induction
(also called a strong induction), in which case the induction step above is replaced by:

Assume that statement1 , statement2 , . . . , statementn are true, and try to prove statementn+1 .

That is, in simple induction you assume the last statement is true and try to prove the next state-
ment. In general induction, you assume that all the preceding statements are true, and try to prove
the next statement. Which one you use depends on the situation.

Remark. The labelling of the statements is arbitrary, so some people prefer to do the induction step this
way:

1
Assume statementn1 is true, and try to prove statementn .
In this case, the induction step for a general or strong induction would be:
Assume that statement1 , statement2 , . . . , statementn1 are true, and try to prove statementn .
Before I give some examples, let me explain how induction is related to well-ordering. Why does proving
the basis and induction steps to prove all the statements?
Consider the case of a simple induction.
Well, if not all the statements are true, there must be a first statement which is not true. Why? The
numbers of the untrue statements form a subset of the positive integers, so by Well-Ordering, this subset
has the smallest element. This smallest element is the number of the first untrue statement.
Suppose then that the first untrue statement is statement53 . Now Ive proved statement1 (the basis
step), and I also proved that if I know a given statement is true, then I know the next statement is true
(the induction step). Applying the induction step to statement1 , I know that statement2 is true. Applying
induction step to statement2 , I know that statement3 is true. And so on. Eventually, I find out that
statement53 is true, but I assumed it was untrue. This contradiction means that all the statements must be
true.
Before I give some induction proofs, Ill give examples of statements that you might try to prove by
induction.

Example. (The Binomial Theorem)


n  
X n k nk
(x + y)n = x y ,
k
k=0

where n = 0, 1, 2, . . ..
This is a sequence of statements:
0  
X 0 k 0k
(x + y)0 = x y ,
k
k=0

1  
1
X 1 k 1k
(x + y) = x y ,
k
k=0
2  
X 2 k 2k
(x + y)2 = x y ,
k
k=0

and so on.
The idea is that if I know that a given statement in the sequence is true, I can use it to prove the next
one.

Example. (Fibonacci numbers) Let x0 = 1, x1 = 1, x2 = 2, x3 = 3, and in general

xn+1 = xn + xn1 for n 2.

This is a recursive formula for the Fibonacci numbers. There is an explicit formula for xn which is
somewhat unexpected:
!n+1 !n+1
1 1+ 5 1 1 5
xn = for n 0.
5 2 5 2

2
Again, this is a sequence of statements:
! !
1 1+ 5 1 1 5
x0 = ,
5 2 5 2

!2 !2
1 1+ 5 1 1 5
x1 = ,
5 2 5 2
3
! !3
1 1+ 5 1 1 5
x2 = ,
5 2 5 2

and so on.

Example. (A statement you wouldnt prove by induction) In calculus, you learned the following
result: A continuous function on a closed interval has an absolute maximum and an absolute minimum on
the interval. This is not a statement youd try to prove using induction.
There is no obvious way to express the result as a sequence of statements, where each one depends on
the one (or ones) before it. In fact, the result is a statement about all continuous functions and all closed
intervals, and these are both uncountable sets. If anything, the result seems to be an uncountable collection
of statements (whereas a sequence or list of statements must be countable).

Here are some examples of induction proofs.

Example. Show that


n
X n(n + 1)
k= for n 1.
2
k=1

Heres a geometric argument which makes the result plausible.

In this picture, there are

1 + 2 + 3 + 4 boxes.

If I add another copy of the original figure, I


have a 4 x 5 rectangle, which has 20 boxes.

Thus,
1 + 2 + 3 + 4 is half of 4 x 5.

The example shows that in general, 1 + 2 + + n boxes should be half of an n by n + 1 rectangle.

3
However convincing the pictures may be, they arent really a proof. Ill prove the result using induction.
First, I prove the result for n = 1. Plugging n = 1 into the formula, I find that
1
X 1 (1 + 1)
k = 1 while = 1.
2
k=1

Its true for n = 1.


Ill use the first form of the induction step. Assume the result is true for n:
n
X n(n + 1)
k= .
2
k=1

Note that Im assuming that this is true! I want to prove the next statement, which is the statement
for n + 1:
n+1
X (n + 1)(n + 2)
k= .
2
k=1

Start with the thing you know to be true.


n
X n(n + 1)
k= .
2
k=1

Add n to both sides:


n
X n(n + 1)
k + (n + 1) = + (n + 1).
2
k=1

The left side is the sum from 1 to n + 1, so


n+1
X n(n + 1)
k= + (n + 1).
2
k=1

Combine the terms on the right over a common denominator:


n+1
X n(n + 1) n(n + 1) + 2(n + 1) n2 + n + 2n + 2 n2 + 3n + 2 (n + 1)(n + 2)
k= + (n + 1) = = = = .
2 2 2 2 2
k=1

This is the statement for n + 1, which is what I wanted to prove. By induction, the result is true for all
n 1.

n
X n
Y
Example. (The definition of sums and product) Here are precise definitions of ai and ai .
i=1 i=1
For n = 1, define
1
X
ai = a1 .
i=1

For n > 1, define


n
X n1
X
ai = ai + an .
i=1 i=1

For example,
2
X 1
X
ai = ai + a2 = a1 + a2 .
i=1 i=1

4
Likewise, for n = 1, define
1
Y
ai = a1 .
i=1

For n > 1, define !


n
Y n1
Y
ai = ai an .
i=1 i=1
X Y
Using these definitions, you can prove properties of and .
n
Y
For example, Ill prove that if c is a constant, then c = cn for every positive integer n.
i=1
If n = 1, then
n
Y 1
Y
c= c = c = c1 .
i=1 i=1

Thus, the result is true for n = 1.


Assume that the result is true for n: n
Y
c = cn .
i=1

Then !
n+1
Y n
Y
c= c c = cn c = cn+1 .
i=1 i=1
Y
(The first equality used the definition of , while the second equality used the induction hypothesis.)
Since the result holds for n + 1, the result is true for all n 1, by induction.

Example. For that if n 5, then 3n > (n + 1)3 .


For n = 5, I have 3n = 35 = 243 and (n + 1)3 = 63 = 216. The result is true for n = 5.
Assume that the result is true for n:
3n > (n + 1)3 .
Ill try to prove the result for n + 1.

3n+1 = 3 3n = 3n + 2 3n > (n + 1)3 + 2 3n . ()

To go on, Ill prove two inequalities, which Ill add to get the one I want. You should be able to follow
the steps, but you might wonder how I knew what to do. As is often the case in proofs, I worked backwards
on scratch paper to figure out what I had to prove. What youre seeing is the cleaned-up version, written
forward.
First, I have n 5, so 3n 243 > 7.
Next,
3n > (n + 1)3 = n3 + 3n2 + 3n + 1 > n3 + 3n2 + 3n.
Now n 5, so n2 25 > 6, or n2 6 > 0. Multiplying the last inequality by n 5, I get

n(n2 6) > 0
n3 6n > 0
n3 + 3n2 + 3n > 3n2 + 9n

So
3n > (n + 1)3 = n3 + 3n2 + 3n + 1 > n3 + 3n2 + 3n > 3n2 + 9n.

5
Adding this inequality to 3n > 7, I get
2 3n > 3n2 + 9n + 7 = (n3 + 6n2 + 12n + 8) (n3 + 3n2 + 3n + 1) = (n + 2)3 (n + 1)3 .
Hence,
(n + 1)3 + 2 3n > (n + 2)3 .
Combining this with (*), I get
3n+1 > (n + 2)3 .
This proves the result for n + 1, so the result is true for all n 5, by induction.

Ill do the next example using general or strong induction that is, Ill assume the truth of all the
statements preceding the nth , rather than just the (n 1)st . Why? Because an is defined in terms of an1
and an2 , rather than just an1 , so I need to know that the induction hypothesis holds for n 2 and n 1,
not just n 1.
In addition, my basis step involves proving the result for n = 0 and for n = 1, not just n = 0. Why?
Because the recursive definition for an doesnt apply unless n 2.

Example. Define
a0 = 2, a1 = 2, and an = 2an1 + 8an2 for n 2.
Prove that
an = 4n + (2)n for n 0.
Check the result for n = 0:
a0 = 2 and 40 + (2)0 = 1 + 1 = 2.
It works.
Check the result for n = 1:
a1 = 2 and 41 + (2)1 = 4 2 = 2.
It works.
Next, take n > 1. Assume the result is true for all numbers less than n. I want to prove the result for
n, which is
an = 4n + (2)n .
By definition,
an = 2an1 + 8an2 .
By my inductive assumption, the result is true for n 1 and for n 2, because they are less than n. So
these statements are true:
an1 = 4n1 + (2)n1 , an2 = 4n2 + (2)n2 .
Plug these in above and simplify:
an = 2 4n1 + (2)n1 + 8 4n2 + (2)n2 = 2 4n1 + 8 4n2 + 2 (2)n1 + 8 (2)n2 =
     

4n2 (8 + 8) + 2n2 (4 + 8) = 4n + (2)n .


The keys to the simplification are to group like terms together (the powers of 4 and the powers of 2
to get the second equality) and to make what youve got look like what you want (factoring out the powers
of 4 and 2 to get the third equality).
This proves the statement for n, so the result is true for all n, by induction.


c 2008 by Bruce Ikenaga 6