Download as pdf or txt
Download as pdf or txt
You are on page 1of 38

Chapter 24

Techniques of Proof IV: Induction

You know, of course, that the devil is dangerous: if you give him the little finger,
he [. . . ] finally takes all your fingers even if you happen to have infinitely many.
– George Pólya, Mathematical Discovery

24.1 Prefatory Remarks


Induction is not an easy topic, especially if you are new to it. My first encounter with
induction occurred shortly before I entered college. I came across the following problem in
[9]:
x
If f0 (x) = and fn+1 = f0 ◦ fn for n = 0, 1, 2, . . . , find a formula for fn (x).
x+1
This problem annoyed me to no end, but after much practice in futility, I seemingly acquired
some foggy notion of how the proof by mathematical induction worked. Of course, I had
no idea why my proof was actually valid—I only knew the mechanics of it. It seemed like
magic to me that I could work out some statement to be true for only one number, assume
the statement was true for some arbitrary number k, and then imply that the statement was
true for the number k + 1, making the statement true for all numbers from 1 to k. Indeed,
proof by mathematical induction seemed like some sort of devilish trick in the vein of the
opening quote by Pólya in [8].
It was not until I was in college in a mathematics seminar on the history of mathematics
and problem-solving that I began to better understand induction. The seminar required a
term project and my project was on the origins of mathematical induction. I eventually had
to confront a fear that had been festering for quite some time—why was induction a valid
proof technique?
Unless you answer this question for yourself, any proof or argument that you come up
with that uses mathematical induction will hardly be satisfactory in your own eyes—your
proof’s validity will seem to be more due to magic than concrete mathematical logic. A brief
proof of induction’s validity is offered in [5]; however, if you found that the proof outlined
in [5] was too difficult to grasp, then see the “Closing Remarks” section at the end of this
chapter for a more detailed explanation concerning mathematical induction’s validity.

1
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 2

The template given below, adapted from [3], should prove to be quite useful when writing
up induction proofs. In the exercises that follow, I have tried to be as strict as possible in
following the template for the sake of clarity. The template is there to help you and to force
you to construct your proof in a logical fashion, but it does you absolutely no good if you do
not understand the basis on which the template was built. The template is given as follows,
where the onus is on the reader to change letters and numbers as needed:

Problem: Prove that for all n ≥ 2, . . . holds.


Solution: For any integer n ≥ 2, let S(n) denote the statement. . .
Base step (n = 2) : S(2) says . . . which is true because . . .
Inductive step S(k) → S(k + 1) : Fix some k ≥ 2. Assume that

S(k) : (write out what S(k) says)

holds. [S(k) is called the inductive hypothesis.] To be proved is that

S(k + 1) : (write out what S(k + 1) says)

follows. Beginning with the left side of S(k + 1),

LHS of S(k + 1) = simplify or rearrange


= ...
= ...
= ... (by S(k))
= ...
= RHS of S(k + 1),

one arrives at the right side of S(k + 1), thereby showing S(k + 1) is
also true, completing the inductive step.
Conclusion: By mathematical induction, it is proved that for all
n ≥ 2, the statement S(n) is true.

A few words of advice are in order before diving into the exercises:

1. Do as many problems as you can. Tom Lehrer, an American pianist and mathemati-
cian, once said, “But in the new (math) approach, the important thing is to understand
what you’re doing, rather than to get the right answer.” You learn by doing. There
are hundreds of thousands of induction problems that are available. Try some of them.
Indeed, I myself learned a great deal by writing this chapter.
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 3

2. If you can (it is worth it), purchase David Gunderson’s Handbook of Mathematical
Induction if you are eager to learn more about mathematical induction (e. g. origins
of the term and method, ways to use it, etc.). His book proved to be invaluable in
writing this chapter.

24.2 Complete Solutions


Exercise 24.2. Check the other two statements above for small n to help convince yourself
that they are true.
Pn 1
First Statement: i=1 i = 2 n(n + 1) for all n ∈ N:
1
X 1
n=1: i = 1 = (1)(1 + 1) X
i=1
2

2
X 1
n=2: i = 1 + 2 = 3 = (2)(2 + 1) X
i=1
2

3
X 1
n=3: i = 1 + 2 + 3 = 6 = (3)(3 + 1) X
i=1
2

..
.

10
X 1
n = 10 : i = 1 + 2 + 3 + 4 + 5 + 6 + 7 + 8 + 9 + 10 = 55 = (10)(10 + 1) X
i=1
2

..
.
The first statement appears to be reasonable based on the small sample of n-values that
have been verified.
Second Statement: 2n−1 ≤ n! is true for all n ∈ N:
n = 1 : 21−1 ≤ (1)! ⇐⇒ 20 ≤ 1! ⇐⇒ 1 ≤ 1 X
n = 2 : 22−1 ≤ (2)! ⇐⇒ 21 ≤ 2! ⇐⇒ 2 ≤ 2 X
n = 3 : 23−1 ≤ (3)! ⇐⇒ 22 ≤ 3! ⇐⇒ 4 ≤ 6 X
..
.

n = 10 : 210−1 ≤ (10)! ⇐⇒ 29 ≤ 10! ⇐⇒ 512 ≤ 3628800 X


..
.
The second statement also appears to be reasonable from the verification of the sample
n-values.
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 4

Exercise 24.10 (i). Prove by induction that


n
X n
i2 = (n + 1)(2n + 1).
i=1
6
Solve: For any integer n ≥ 1, let S(n) denote the statement
n
n X
S(n) : i2 =
(n + 1)(2n + 1).
i=1
6

Base step (n = 1): The statement S(1) says that 1i=1 i2 = 12 = 16 (1 + 1)(2 + 1) which is
P
true because 1 = 1.
Inductive step S(k) → S(k + 1): Fix some k ≥ 1, k ∈ N. Assume that
k
X k
S(k) : i2 = (k + 1)(2k + 1)
i=1
6
holds (this is the inductive hypothesis). To be proved is that
k+1
X (k + 1) (k + 1)(k + 2)(2k + 3)
S(k + 1) : i2 = ((k + 1) + 1)(2(k + 1) + 1) =
i=1
6 6
follows. Beginning with the left side of S(k + 1),
k+1
X k
X
2
i = i2 + (k + 1)2 (evaluate sum for i = k + 1)
i=1 i=1

k
= (k + 1)(2k + 1) + (k + 1)2 (by S(k))
6
k(k + 1)(2k + 1) + 6(k + 1)2
= (common denominator)
6
 
(k + 1) k+1
= [k(2k + 1) + 6(k + 1)] factor out
6 6
(k + 1) 2
= [2k + k + 6k + 6] (expand)
6
(k + 1) 2
= [2k + 7k + 6] (add)
6
(k + 1)
= [k(2k + 3) + 2(2k + 3)] (factor)
6
(k + 1)
= [(k + 2)(2k + 3)] (factor)
6
(k + 1)(k + 2)(2k + 3)
= , (rearrange)
6
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 5

one arrives at the right side of S(k + 1) , thereby showing S(k + 1) is also true, completing
the inductive step.
Conclusion: By mathematical induction, it is proved that for all n ≥ 1, n ∈ N, the
statement S(n) is true.
Understand and Plan: This is your first real induction problem, a problem that appears
in almost any text that covers mathematical induction. It is straightforward and easy for
anyone who has dealt with induction before, but if you are new to proving by mathematical
induction, then really buckle down and work through the nuts and bolts of this problem—no
shortcuts. The best way to work through this problem, and all the rest that follow, is to very
closely follow the template for proving by induction given in the “Prefatory Remarks” section
of this chapter∗ . Following the template will force you to understand how the inductive
hypothesis is to be constructed and where to use it. If you closely follow the template, then
your proof should look like, if not exactly the same, as the one given above.
Pn 2All of that being
said, the problem statement is straightforward: Verify that a sum, i=1 i in this case, is
equivalent to a closed-form expression, in this case n6 (n + 1)(2n + 1). As with most induction
problems, try to start by verifying that the statement in question holds when n = 1 (in some
cases, you may want to start with a different n-value). Once you are satisfied with verifying
your statement for a number of n-values, simply move on to the inductive step.

Exercise 24.10 (ii). Show that 2n ≤ 2n for all natural numbers n.


Solve: For any integer n ≥ 1, let S(n) denote the statement
S(n) : 2n ≤ 2n .
Base step (n = 1): The statement S(1) says that 2(1) ≤ 21 which is true because 2 ≤ 2.
Inductive step S(k) → S(k + 1): Fix some k ≥ 1, k ∈ N. Assume that
S(k) : 2k ≤ 2k
holds. To be proved is that
S(k + 1) : 2(k + 1) ≤ 2k+1
follows. Beginning with the left side of S(k + 1),
2(k + 1) = 2k + 2 (expand)

≤ 2k + 2 (by S(k))

≤ 2k + 2k (since k ≥ 1)

≤ 2(2k ) (since 2k + 2k = 2k (1 + 1) = 2(2k ))

= 2k+1 , (multiply)

The template forces you to logically construct your proof, but the template should not serve as a
substitute for your knowledge of why mathematical induction is a valid proof technique. Use the template
as an aid in writing up your inductive proof—not as a mathematical crutch.
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 6

one arrives at the right side of S(k + 1), thereby showing S(k + 1) is also true, completing
the inductive step.
Conclusion: By mathematical induction, it is proved that for all n ≥ 1, n ∈ N, the state-
ment S(n) is true.
Understand and Plan: Many inequalities are proved using induction, but the proofs often
require a moderate amount of ingenuity. Generally, there is a good bit of algebraic manip-
ulation involved as is made evident in the solution of this exercise. As with the previous
exercise, the problem statement is straightforward: Prove that the inequality 2n ≤ 2n holds
for all n ∈ N. Start by verifying that the statement holds for n = 1. Then assume the induc-
tive hypothesis. The hard part about this problem is mostly figuring out how to effectively
manipulate the inequalities that appear when trying to move from the left side of S(k + 1)
to the right side of S(k + 1). Always be thinking about what you can logically deduce in
order to make the problem easier. For example, it is a helpful observation to realize that
2k + 2 ≤ 2k + 2k (because k ≥ 1). Without this observation, trying to get to the right side
of S(k + 1) might be much more difficult, if not impossible. See (24.11) for a problem that
requires a great deal of ingenuity when it comes to manipulating inequalities.

Exercise 24.10 (iii). Prove that 32n − 1 is divisible† by 8 for all natural numbers n.
Solve: For any integer n ≥ 1, let S(n) denote the statement

S(n) : 8 | (32n − 1).

Base step (n = 1): The statement S(1) says that 8 | (32(1) − 1) which is true because 8 | 8.
Inductive step S(k) → S(k + 1): Fix some k ≥ 1, k ∈ N. Assume that

S(k) : 8 | (32k − 1)

holds, where m is some integer that satisfies

32k − 1 = 8m.

To be proved is that
S(k + 1) : 8 | (32(k+1) − 1)
follows. To see that S(k + 1) follows, show that 32(k+1) − 1 is divisble by 8:

32(k+1) − 1 = 32k+2 − 1 (expand exponent)


= 32k · 32 − 1 factor out 32


= 9(32k − 1) + 8 (rearrange and manipulate)


= 9(8m) + 8 (by S(k))
= 8(9m + 1). (factor out 8)

Recall that the “|” symbol means divides; thus, 8 | (32n − 1) is read as “8 divides 32n − 1.”
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 7

It has been shown that 8 | (32(k+1) − 1), thus concluding the proof of the inductive step.
Conclusion: By mathematical induction, it is proved that for all n ≥ 1, n ∈ N, the
statement S(n) is true.
Understand and Plan: This is a fairly straightforward problem, but one has to be intelligent
in setting it up. Verifying that the statement holds for n = 1 is easy enough, but this
problem demands something more before proving the inductive step. To be proved is that
the statement 32n − 1 is divisible by 8. What would a satisfactory proof of this look like?
Clearly, if 8 is always a factor of 32n − 1, then the statement is valid and the proof is done.
How can one be sure that 8 is, indeed, always a factor of 32n − 1? Writing that 32n − 1 = 8m
for some integer m indicates that 8 will always be a factor of 32n − 1. This knowledge is
required to effectively prove that the statement holds for all n ∈ N. Otherwise, one will
get stuck and not be able to use the inductive hypothesis when moving from the left side of
S(k + 1) to the right side of S(k + 1). If you do not believe you will get stuck, then give it
a try. When you arrive at the fact that 32(k+1) − 1 = 9(32k − 1) + 8, you will not be able
to effectively progress in your proof. This is the kind of thinking (making sure you have
everything set up properly before moving on to the inductive step) you want to practice
when proving divisibility problems by induction. The following exercise also makes use of
this line of reasoning.

Exercise 24.10 (iv). Prove that 17 divides 34n + 43n+2 for all n ∈ N.
Solve: For any integer n ≥ 1, let S(n) denote the statement
S(n) : 17 | (34n + 43n+2 ).
Base step (n = 1): The statement S(1) says that 17 | (34(1) +43(1)+2 ) which is true because
17 | 1105.
Inductive step S(k) → S(k + 1): Fix some k ≥ 1, k ∈ N. Assume that
S(k) : 17 | (34k + 43k+2 )
holds, where m is some integer that satisfies
34k + 43k+2 = 17m.
To be proved is that
S(k + 1) : 17 | (34(k+1) + 43(k+1)+2 )
follows. To see that S(k + 1) follows, show that 34(k+1) + 43(k+1)+2 is divisible by 17:
34(k+1) + 43(k+1)+2 = 34k · 34 + 43k · 43 · 42 (expand exponents)
= 81 · 34k + 64 · 43k+2 (evaluate select terms)
= 17 · 34k + 64 · 34k + 64 · 43k+2 (strategically rearrange)
= 64(34k + 43k+2 ) + 17 · 34k (factor out 64)
= 64(17m) + 17 · 34k (by S(k))
= 17(64m + 34k ). (factor out 17)
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 8

It has been shown that 17 | (34(k+1) + 43(k+1)+2 ), thus concluding the proof of the inductive
step.
Conclusion: By mathematical induction, it is proved that for all n ≥ 1, n ∈ N, the
statement S(n) is true.
Understand and Plan: This is essentially a more difficult version of the previous exercise,
where everything is identical in setting the problem up, but proving the inductive step is a
little more tricky. The ability to see that you should split up 81 · 34k into 17 · 34k + 64 · 34k
is something that comes with experience and lacking this insight can be a little infuriating
at first.

Exercise 24.10 (v). Show that sin nx ≤ n sin x for all natural numbers n and 0 ≤ x ≤ π2 ,
where x ∈ R is fixed.
Solve: For any integer n ≥ 1, let S(n) denote the statement‡

S(n) : | sin(nx)| ≤ n| sin(x)| for any fixed x ∈ R.

Base step (n = 1): The statement S(1) says that | sin(x)| ≤ | sin(x)|, which is true because
x ∈ R is fixed.
Inductive step S(k) → S(k + 1): Fix some k ≥ 1, k ∈ N. Assume that

S(k) : | sin(kx)| ≤ k| sin(x)|

holds. To be proved is that

S(k + 1) : | sin((k + 1)x)| ≤ (k + 1)| sin(x)|

follows. Beginning with the left side of S(k + 1),

| sin((k + 1)x)| = | sin(kx + x)| (expand)


= | sin(kx) cos(x) + cos(kx) sin(x)| (by (24.1))
≤ | sin(kx) cos(x)| + | cos(kx) sin(x)| (by (24.2))
= | sin(kx)| · | cos(x)| + | cos(kx)| · | sin(x)| (since |ab| = |a||b|)
≤ | sin(kx)| + | sin(x)| (since | cos(θ)| ≤ 1)
≤ k| sin(x)| + | sin(x)| (by S(k))
= (k + 1)| sin(x)|, (factor)

one arrives at the right side of S(k + 1), thereby showing S(k + 1) is also true, completing
the inductive step.

See the “Understand and Plan” portion directly following the conclusion of this problem to understand
why the notation has slightly changed.
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 9

Conclusion: By mathematical induction, it is proved that for all n ≥ 1, n ∈ N, the


statement S(n) is true.
Understand and Plan: There is a subtlety to this problem that, once picked up on, makes
it a good bit easier to deal with. The parameter 0 ≤ x ≤ π2 tells us that sin(x) must be
positive. This is easier to see if you can envision sin(x) graphically, where the portion of the
graph under question lies solely in the first quadrant (where all values of sin(x) are positive).
This observation, along with the fact that n ∈ N, means that we can rewrite the problem as
follows:
Show that | sin(nx)| ≤ n| sin(x)| holds for each n ≥ 1, n ∈ N, and x ∈ R is fixed.
Before we attempt to solve this problem, it would behoove us to recall the addition formula
for sin(α + β) and the Triangle Inequality, respectively:
sin(α + β) = sin α cos β + cos α sin β (24.1)
and
If a, b ∈ R, then |a + b| ≤ |a| + |b|. (24.2)
This version of the Triangle Inequality is only concerned with two numbers, a and b. The
generalized Triangle Inequality is dealt with in Exercise 24.10 (xv). The ability to recognize
that (24.1) and (24.2) must be used for this problem comes with experience. This is often
what makes some induction problems rather difficult—experience and know-how when it
comes to manipulating expressions appropriately to produce the desired result.

Exercise 24.10 (vi). Prove the Binomial Theorem, that is,


n  
n
X n n−r r
(x + y) = x y (24.3)
r=0
r
 
for all n ∈ N. Recall that nr = r!(n−r)!
n!

for all 0 ≤ r ≤ n.

Solve: For any integer n ≥ 1, let S(n) denote the statement


n  
n
X n n−r r
S(n) : (x + y) = x y .
r=0
r
P1 1

Base step (n = 1): The statement S(1) says that (x + y)1 = r=0 r x1−r y r which is true
because
1  
1
X 1 1−r r
(x + y) = x y
r=0
r
   
1 1−0 0 1 1−1 1
= x y + x y
0 1

= x + y.
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 10

Inductive step S(k) → S(k + 1): Fix some k ≥ 1, k ∈ N. Assume that


k  
k
X k k−r r
S(k) : (x + y) = x y
r=0
r

holds. To be proved is that


k+1  
k+1
X k + 1 k+1−r r
S(k + 1) : (x + y) = x y
r=0
r

follows. Beginning with the left side of S(k + 1),

(x + y)k+1 = (x + y)k (x + y)
"k  
#
X k k−r r
= x y (x + y) (by S(k))
r=0
r
" k   # " k   #
X k X k
k−r r k−r r
=x x y +y x y (expand)
r=0
r r=0
r

k   k  
X k k+1−r r X k k−r r+1
= x y + x y (multiply)
r=0
r r=0
r

k   k+1  
X k k+1−r r X k
= x y + xk+1−r y r (reindex the right sum)
r=0
r r=1
r − 1

k     k    
X k k+1−r r k k+1 0 X k k+1−r r k 0 k+1
= x y + x y + x y + xy (expand)
r=1
r 0 r=1
r−1 k

k    
k+1
X k k
=x + + xk+1−r y r + y k+1 (combine like terms)
r=1
r r−1

k  
k+1
X k+1
=x + xk+1−r y r + y k+1 (Pascal’s Identity)
r=1
r

  k    
k + 1 k+1 X k + 1 k+1−r r k + 1 k+1
= x + x y + y (prep. to reindex sum)
0 r=1
r k + 1

k+1  
X k + 1 k+1−r r
= x y , (reindex sum)
r=0
r
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 11

one arrives at the right side of S(k+1), thereby showing that S(k+1) is also true, completing
the inductive step.
Conclusion: By mathematical induction, it is proved that for all n ≥ 1, n ∈ N, the
statement S(n) is true.
Understand and Plan: This is a challenging exercise, as there is a lot going on, but it is
surprisingly manageable if one works through it in a piecemeal fashion. The problem starts
off simple, factoring (x + y)k+1 and then applying the inductive hypothesis. In general, if
the inductive hypothesis is used right away, as it is in this case, you can usually expect a
rather easy or somewhat difficult problem, where here we have the latter case. A lot of effort
will have to go into working from the left side of S(k + 1) to the right side of S(k + 1).
While doing this work, much experimentation will be required. For instance, once you have
reached the point that

(x + y)k+1 = (x + y)k (x + y)
..
.
k   k  
X k k+1−r r X k k−r r+1
= x y + x y ,
r=0
r r=0
r

you will then have to use some ingenuity to complete the proof. As seen in the solution to
this exercise, most of that ingenuity referred to involves seeing when to evaluate summations
appropriately and when/how to reindex summations§ .
In the context of this problem, when the expression
k   k  
X k k+1−r r X k k−r r+1
x y + x y (24.4)
r=0
r r=0
r

is encountered, the right summation should be looked at closely. Why? Notice that the sum
we are ultimately trying to obtain,
k+1  
X k + 1 k+1−r r
x y , (24.5)
r=0
r

has as its terms xk+1−r y r and not xk−r y r+1 , which appear in the right summation of (24.4).
This indicates that the right sum may need to be “tinkered with” or reindexed in order for
the desired sum to be obtained. If the terms were different for the left sum as well, then both
sums would likely need to be reindexed to solve the problem (in order to obtain summations
with equivalent terms as those in (24.5)).
A modicum of experience with binomial coefficients would prove to be invaluable in
solving this problem. For instance, the proof of this exercise was essentially finished once
§
The idea of “reindexing the sum” is a tricky one at first if you have never been exposed to it. This idea
is illustrated with two worked examples, one from this problem and one in general, in the “Closing Remarks”
section at the end of this chapter.
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 12

Pascal’s Identity? was used:


     
n+1 n n
= + (Holds for any fixed r ≥ 1, and all n ≥ r)
r r r−1

Lastly, one may find it useful to prove (24.3) by working without summation notation,
whereby (24.3) would be stated as follows:
         
n n n n n−1 n n−2 2 n n−1 n n
(x + y) = x + x y+ x y + ··· + xy + y
0 1 2 n−1 n
I, however, have found that working with such expanded notation makes everything more an-
noying and difficult to deal with. Working with summation, or sigma (Σ), notation provides
an easy and compact way to deal with binomial coefficients.

Exercise 24.10 (vii). Show that n2 − 1 is divisible by 8 when n is an odd natural number.
Solve: For any odd integer n ≥ 1, let S(n) denote the statement

S(n) : 8 | (n2 − 1).

Base step (n = 1): The statement S(1) says that 8 | (12 −1) which is trivially true because
8 | 0.
Inductive step: Fix an odd number k ≥ 1, where k = 2` − 1 for some arbitrary k, ` ∈ N.
Assume that
S(k) : 8 | (k 2 − 1)
holds; that is, suppose there is some integer m for which k 2 − 1 = 8m. The next odd number
after k is k + 2; thus, it remains to be shown that

S(k + 2) : 8 | ((k + 2)2 − 1)

follows. Beginning with the left side of S(k + 2),

(k + 2)2 − 1 = k 2 + 4k + 3 (expand)
= k 2 − 1 + 4k + 4 (strategically rearrange)
= 8m + 4k + 4 (by S(k))
= 8m + 4(2` − 1) + 4 (since k = 2` − 1)
= 8m + 8` (simplify)
= 8(m + `), (factor out 8)

one arrives at an expression which is divisible by 8, completing the inductive step.


Conclusion: By mathematical induction, it is proved that for all odd n ≥ 1, n ∈ N, the
?
This identity may be proved using induction. See [3] for such a proof.
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 13

statement S(n) is true.


Understand and Plan: This is our first somewhat “unusual” induction problem. Notice
that the inductive step is concerned with proving the implication S(k) → S(k + 2) and not
S(k) → S(k + 1) [in fact, realizing this is the key to solving this problem]. This is because we
are trying to show that the statement in question, S(n), is true for all odd natural numbers.
The inductive step may look bizarre at first, but it is still valid. If you think you can prove
the implication S(k) → S(k + 1), give it a shot. You will not be able to. The reason is that
k + 1 is an even number and S(n) does not hold for all even n ∈ N.

Exercise 24.10 (viii). Prove Leibniz’s Theorem for repeated differentiation of a product:
If u and v are functions of x, then prove that

dn
     
n n n
(uv) = u 0 vn + u 1 v n−1 + u2 vn−2 + · · · + ur vn−r + · · · + un v0 ,
dxn 1 2 r
di u dj v
for all n ∈ N, where ui and vj denote dxi
and dxj
, respectively.
(You will need to use      
k k k+1
+ =
r−1 r r
but proving this identity is true should not pose much difficulty£ .)
Solve: For any integer n ≥ 1, let S(n) denote the statement¶
n  
(n)
X n
S(n) : (f g) = f (r) g (n−r) . (24.6)
r=0
r

Base step (n = 1): The statement S(1) says that


1  
(1)
X 1 (r) (1−r)
(f g) = f g
r=0
r
   
1 (0) (1) 1 (1) (0)
= f g + f g
0 1

= f g (1) + f (1) g,

which is true because (f g)0 = f g 0 + f 0 g. [This is called the Product Rule and is often first
encountered in calculus.]
£
This identity is called Pascal’s Identity and is used in the proof of the Binomial Theorem in Exercise
24.10 (vii).

See the “Understand and Plan” portion directly following the conclusion of this problem to understand
why the notation has slightly changed.
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 14

Inductive step S(k) → S(k + 1): Fix some k ≥ 1, k ∈ N. Assume that


k  
(k)
X k (r) (k−r)
S(k) : (f g) = f g
r=0
r

holds. To be proved is that


k+1  
(k+1)
X k + 1 (r) (k+1−r)
S(k + 1) : (f g) = f g
r=0
r

follows. Beginning with the left side of S(k + 1),


0
(f g)(k+1) = (f g)(k)

k  
!0
X k (r) (k−r)
= f g (by S(k))
r=0
r

k   k  
X k (r+1) (k−r) X k (r) (k+1−r)
= f g + f g (by base case or Product Rule)
r=0
r r=0
r

k+1   k  
X k (r) (k+1−r)
X k (r) (k+1−r)
= f g + f g (reindex left sum)
r=1
r − 1 r=0
r

k     k    
X k (r) (k+1−r) k (k+1) (0) X k (r) (k+1−r) k (0) (k+1)
= f g + f g + f g + f g (expand)
r=1
r − 1 k r=1
r 0

k    
(k+1)
X k k
= fg + + f (r) g (k+1−r) + f (k+1) g (rearrange; combine like terms)
r=1
r − 1 r

k  
(k+1)
X k + 1 (r) (k+1−r)
= fg + f g + f (k+1) g (Pascal’s Identity)
r=1
r

  k    
k+1 (k+1)
X k + 1 (r) (k+1−r) k + 1 (k+1)
= fg + f g + f g (prep. to reindex sum)
0 r=1
r r+1

k+1  
X k+1
= f (r) g (k+1−r) , (reindex sum)
r=0
r

one arrives at the right side of S(k+1), thereby showing that S(k+1) is also true, completing
the inductive step.
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 15

Conclusion: By mathematical induction, it is proved that for all n ≥ 1, n ∈ N, the


statement S(n) is true.
Understand and Plan: Part of the trick of solving certain problems is knowing how to adopt
whatever notation best suits your needs—we saw this in practice for the proof of (24.3). As
noted in [9], if we have that y = f (x), where x is the independent variable and y is the
dependent variable, then there are some alternative notations for the derivative of y = f (x):
dy df d
f 0 (x) = y 0 = = = f (x) = Df (x) = Dx f (x)
dx dx dx
d
Notice that the symbols D and dx are called differential operators since they indicate the
operations of differentiation, which is the process of calculating a derivative.
In this particular exercise, we are given that u and v are functions of x, and we are being
dn
asked to prove a formula for the nth derivative of their product, dx n (uv). This exercise will

be far easier to write up (and easier to follow) if we use sigma notation and let f = u and
g = v, where f 0 and g 0 denote the first derivatives of u and v, respectively, while f (n) and
g (n) denote the nth derivatives of u and v, respectively. Adopting this notation, we can
reformulate the expression for Leibniz’s Theorem, giving (24.6). The rest of the proof that
follows after the change in notation is very similar to the proof of (24.3) in that the logic
behind the reindexing of sums is the same as well as several other algebraic steps.

Exercise 24.10 (ix). Prove that


n n
!2
X X
r3 = r .
r=1 r=1

Solve: Notice that, by virtue of Example 24.4 on page 168 of [5], the problem may be restated
as follows:
Prove that n  2
X
3 n(n + 1)
r =
r=1
2
holds for all n ≥ 1, n ∈ N.
Using this modified notation, for any integer n ≥ 1, let S(n) denote the statement
n  2
X
3 n(n + 1)
S(n) : r = .
r=1
2

Base step (n = 1): The statement S(1) says that (1)3 = (1)2 which is true because 1 = 1.
Inductive step S(k) → S(k + 1): Fix some k ≥ 1, k ∈ N. Assume that
k  2
X
3 k(k + 1)
S(k) : r =
r=1
2
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 16

holds. To be proved is that


k+1  2
(k + 1)(k + 2)
X
3
S(k + 1) : r =
r=1
2

follows. Beginning with the left side of S(k + 1),


k+1
X k
X
r3 = r3 + (k + 1)3 (evaluate sum for i = k + 1)
r=1 r=1
 2
k(k + 1)
= + (k + 1)3 (by S(k))
2

(k + 1)2  2 (k + 1)2
 

= k + 4(k + 1) factor out
4 4

(k + 1)2
= [(k + 2)(k + 2)] (factor quadratic polynomial)
4

(k + 1)2 (k + 2)2
= (multiply and rearrange)
4
 2
(k + 1)(k + 2)
= , (rearrange)
2

one arrives at the right side of S(k+1), thereby showing that S(k+1) is also true, completing
the inductive step.
Conclusion: By mathematical induction, it is proved that for all n ≥ 1, n ∈ N, the
statement S(n) is true.
Understand and Plan: Recalling that ni=1 i = n(n+1)
P
2
allows us to transform this exercise
2
into a routine induction problem. Regardless, the step where (k+1)
4
is factored out is an
important one. If we do not factor this out and, instead, choose to expand (k + 1)3 , the
problem becomes much more difficult than it needs to be.

n

Exercise 24.10 (x). Show that r
is an integer for all 0 ≤ r ≤ n.
Solve: For any integer n ≥ 0, let S(n) denote the statement
 
n
S(n) : ∈ Z | 0 ≤ r ≤ n.
r
0

Base step (n = 0): The statement S(0) says that
0
 r ∈ Z, 0 ≤ r ≤ 0 which is true because
0 ≤ r ≤ 0 means that r = 0 and consequently 0 = 1 ∈ Z.
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 17

Inductive step S(k) → S(k + 1): Fix some k ≥ 0, k ∈ N ∪ {0}. Assume that
 
k
S(k) : ∈Z | 0≤r≤k (24.7)
r
holds. To be proved is that
 
k+1
S(k + 1) : ∈Z | 0≤r ≤k+1
r
follows. Begin by using Pascal’s Identity to show that S(k + 1) follows, i. e. that k+1

r
is an
integer, given the restrictions:
     
k+1 k k
= + (Pascal’s Identity)
r r−1 r
   
k k
=⇒ ∈Z + ∈Z (by S(k))
r−1 r
    
k k
=⇒ ∈Z + ∈Z ∈Z (sum of two integers is an integer)
r−1 r
 
k+1
=⇒ ∈Z (express as desired)
r
k+1

Thus, one arrives at what was to be proved, namely that r
∈ Z | 0 ≤ r ≤ k + 1 follows
from the inductive hypothesis, S(k).
Conclusion: By mathematical induction, it is proved that for all n ≥ 0, n ∈ N ∪ {0}, the
statement S(n) is true.
Understand and Plan: The key to solving this problem lies in how the inductive hypothesis
is being applied. You really have to know what the inductive hypothesis is communicating in
order to be able to use it effectively. What exactly is the inductive hypothesis communicating
here? By noting the restriction 0 ≤ r ≤ k in (24.7), we can see that S(k) is essentially saying
that any term in the set M, where
       
k k k k
M= , ,..., , , (24.8)
0 1 r−1 r
is an integer. Why is this observation important? Well, when Pascal’s Identity is used in
the proof, we get that r = r−1 + r . By (24.7), it is clear that kr ∈ Z, but what can
k+1 k k
 
k

we deduce about r−1 ? If we cannot deduce anything about this, then we are in trouble,
but any issue that might have arisen concerning this is avoided by noticing just what the
k

inductive hypothesis is saying in (24.8). We see that r−1 ∈ Z by (24.8), allowing us to
continue with the proof.
The lesson to be learned from this exercise is that if an induction problem is giving you
trouble, then it may help to sit back and think about what the inductive hypothesis is really
saying. The importance of this kind of reasoning is highlighted in Exercise 24 (xvi).
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 18

Exercise 24.10 (xi). Let X be a finite set with n elements. Show that X has 2n distinct
subsets.
Solve: For n ≥ 0, n ∈ N ∪ {0} let S(n) denote the statement

S(n) : If X is a finite set with n elements, then X has 2n distinct subsets. (24.9)

Base step (n = 0): The statement S(0) says that if X is a finite set with 0 elements, then
X has 20 = 1 distinct subsets. This is true because a set with 0 elements is the empty set, ∅,
and ∅ ⊆ ∅ by Example 1.12 on page 7 of [5]∗∗ ; that is, if X = ∅, then X has 20 = 1 distinct
subset and that subset is itself, i. e. ∅ ⊆ ∅.
Inductive step S(k) → S(k + 1): Fix some k ≥ 0, k ∈ N ∪ {0}. Assume that

S(k) : If X is a finite set with k elements, then X has 2k distinct subsets

holds. To be proved is that

S(k + 1) : If X is a finite set with k + 1 elements, then X has 2k+1 distinct subsets

follows.
Solution 1 †† : Start by letting L be a set with k+1 elements, where L = {a1 , a2 , a3 , . . . , ak , `}.
It is possible to write L = M ∪ {`} and M = L \ {`}. In this way,

L = M ∪ {`} = {a1 , a2 , a3 , . . . , ak , `} (24.10)

and
M = L \ {`} = {a1 , a2 , a3 , . . . , ak }.
By the inductive hypothesis, the set M has 2k distinct subsets. Let N ⊆ M for some arbitrary
set N . Notice that if N ⊆ M , then N1 ⊆ M ⊆ L and N2 ⊆ N2 ∪ {`} ⊆ L necessarily follow,
where N1 and N2 denote distinct arbitrary sets N ; that is, for each subset N of M there are
exactly two subsets of L, namely N1 and N2 ∪ {`}. Such subsets constitute all of the subsets
of L and are all distinct. Since there are 2k distinct subsets of M and there are two subsets
of L for each subset of M , there are 2 · 2k = 2k+1 distinct subsets of L in total, completing
the inductive step.
Conclusion: By mathematical induction, it is proved that for all n ≥ 0, n ∈ N ∪ {0}, the
statement S(n) is true.
Solution 2: Start by letting L be a set with k + 1 elements, where L = {a1 , a2 , a3 , . . . , ak , `}.
∗∗
Note that the empty set has no subsets other than itself.
††
The difference between the first and second solutions is that Solution 1 splits subsets of (24.10) into
2k groups of two subsets (2 · 2k = 2k+1 ) whereas Solution 2 splits subsets of (24.10) into two groups of 2k
subsets (2k + 2k = 2k+1 ).
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 19

Consider the set, P(L), of all subsets of L, including the empty set and L itself‡‡ . Partition
P(L) into two groups, where the first group contains subsets of L which do not have ` in
them, and the second group contains subsets of L which do have ` in them. Consequently:

Partitions of P(L)
First Group Second Group
∅ {`}
{a1 } {a1 , `}
{a1 , a2 } {a1 , a2 , `}
{a1 , a2 , . . . , ak } {a1 , a2 , . . . , ak , `}

The first group is made up of all of the subsets from the k-element set {a1 , a2 , a3 , . . . , ak }.
Thus, by the inductive hypothesis, the first group has 2k distinct entries. By construction,
i. e. by the nature of how the second group is constructed in relation to the first, the num-
ber of entries in the second group must be the same as the number of entries in the first
group. Thus, the second group must also have 2k distinct entries. Because the collection
of all subsets of L has been partitioned into the two groups described above, there must be
2k + 2k = 2k+1 distinct subsets of L, completing the inductive step.
Conclusion: By mathematical induction, it is proved that for all n ≥ 0, n ∈ N ∪ {0}, the
statement S(n) is true.
Understand and Plan: This is a pretty tough problem, for it is far more abstract than for-
mulaic/mechanical for the exercises we have done so far. Two solutions are given, where
it may be argued that Solution 1 is more abstract and harder to understand than Solution
2, but both solutions are equally valid. Try to understand both proofs if you can. If, after
looking over both solutions, you still cannot quite grasp why (24.9) is true, then consider
writing out what happens when you specify a small value for n. For example, what happens
when X is a 4-element set?
For the sake of clarity, the case for when X = {1, 2, 3, `} will be worked out using the
approach employed in Solution 2 (that is, we will partition X into two groups and so on and
so forth). Seeing this worked out example may facilitate one’s progress in understanding the
general proof§§ .
Start by letting A be a set with 4 elements, where A = {1, 2, 3, `}. Consider the set,
P(A), of all subsets of A, including the empty set and A itself. This is given by
n
P(A) = ∅, {1}, {2}, {3}, {`}, {1, 2}, {1, 3}, {1, `}, {2, 3}, {2, `}, {3, `},
o
{1, 2, 3}, {1, 2, `}, {1, 3, `}, {2, 3, `}, {1, 2, 3, `} .
‡‡
Such a set is called a power set, and, in this case, the power set of L is written as P(L). As an example,
if S = {1, 2, 3}, then P(S) = {∅, {1}, {2}, {3}, {1, 2}, {1, 3}, {2, 3}, {1, 2, 3}}. Notice that there are three
elements in S in this example and there are 8 = 23 subsets of S or, more importantly, there are 23 = 8
elements in P(S). Thus, one could restate the original problem as follows: Prove that if S = {1, 2, 3, . . . , n},
then P(S), the power set of S, has 2n elements.
§§
Note that if X = {1, 2, 3, `}, then we will be trying to show that there are 24 = 16 subsets of X.
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 20

Partition P(A) into two groups, where the first group contains subsets of A which do not
have ` in them, and the second group contains subsets of A which do have ` in them.
Consequently:

Partitions of P(A)
First Group Second Group
∅ {`}
{1} {1, `}
{2} {2, `}
{3} {3, `}
{1, 2} {1, 2`}
{1, 3} {1, 3, `}
{2, 3} {2, 3, `}
{1, 2, 3} {1, 2, 3, `}

The first group is made up of all of the subsets from the 3-element set {1, 2, 3}. Thus, by
the inductive hypothesis, the first group has 23 = 8 distinct entries, easily seen by the table
above. By construction, the number of entries in the second group must be the same as the
number of entries in the first group. Thus, the second group must also have 23 = 8 distinct
entries, also easily seen by the table above. Because the collection of all subsets of A has
been partitioned into the two groups described above, there must be 23 +23 = 23+1 = 24 = 16
distinct subsets of A, completing the example.

Exercise 24.10 (xii). This exercise will, in effect, generalize the result of Exercise (iii).

(a) Show that xn−1 is divisible by x − 1 for all n ∈ N and where x 6= 1 is a natural number.
xn−1
(b) Find a formula for and use induction to show it holds for all n.
x−1
Solve:
(a) For any integer n ≥ 1 let S(n) denote the statement

S(n) : (x − 1) | (xn − 1) | ∀n ∈ N, x ∈ N \ {1}.

Note that S(n) is a special case of a more general statement: Let P (n) be the statement
that for any polynomials p and q,

P (n) : (p − q) | (pn − q n ) | ∀n ∈ N, p 6= q.

It is clear that S(n) is the case when p = x and q = 1.


Base step (n = 1): The statement P (1) says that (p − q) | (p1 − q 1 ) which is true because
(p − q) | (p − q).
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 21

Inductive step P (k) → P (k + 1): Fix some k ≥ 1, k ∈ N. Assume that


P (k) : (p − q) | (pk − q k ) | ∀k ∈ N, p 6= q
holds. That is, pk − q k = (p − q)`, where ` is some arbitrary polynomial. To be proved is
that
P (k + 1) : (p − q) | (pk+1 − q k+1 )
follows. To see that P (k + 1) follows, show that pk+1 − q k+1 is divisible by p − q:
pk+1 − q k+1 = pk+1 − pk q + pk q − q k+1 (strategically manipulate)
= pk (p − q) + q(pk − q k ) factor out pk and q


= pk (p − q) + q(p − q)` (by P (k))


= (p − q)[pk + q`]. (factor out p − q)
It has been shown that (p − q) | (pk+1 − q k+1 ), thus concluding the proof of the inductive
step.
Conclusion: By the principle of mathematical induction, for every n ∈ N, the stronger
result P (n) holds, and thus S(n) holds as well.
Understand and Plan: This problem is very similar to Exercises 24.10 (iii, iv, vii) in that
we are being asked to show that something is divisible by something else, but this problem
illustrates something in proof-writing that has some practical value—sometimes it is more
useful to prove a stronger result?? than what you are originally asked to prove. By proving
the more general statement P (n) that covers S(n), we make things easier on ourselves not
to mention that a proof for P (n) is more impressive than a proof for S(n).

xn − 1
(b) Expression: ; Try a few n-values and attempt to discover a pattern:
x−1
x1 − 1 x
 −
1
n=1: = =1
x−1 x−1
 

x2 − 1 (x−1)(x

+ 1)
n=2: = =x+1

x−1 (x− 1)
 

x3 − 1  2
(x−1)(x + x + 1)
n=3: = = x2 + x + 1

x−1 (x
 −1)


..
.

xn − 1  n−1
(x−1)(x + · · · + x + 1)
n: = = xn−1 + · · · + x + 1 (Conjecture)

x−1 (x
 −1)


??
See the “Concluding Remarks” section for a more detailed description of what it means for a statement
to be stronger than another statement.
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 22

The sample of n-values tested suggests the formula


n
xn − 1 X
= xn−1 + xn−2 + · · · + x + 1 = xi−1 .
x−1 i=1

Of course, this is only a suggested formula. To prove that the formula is actually valid, the
principle of mathematical induction will be used, as required. Give the statement that is to
be proven:
Prove that n
xn − 1 X i−1
= x
x−1 i=1

holds for all n ∈ N.


Solve: For any integer n ≥ 1, let S(n) denote the statement
n
X xn − 1
S(n) : xi−1 = .
i=1
x−1

x1 −1
Base step (n = 1): The statement S(1) says that x0 = x−1
which is true because 1 = 1.
Inductive step S(k) → S(k + 1): Fix some k ≥ 1, k ∈ N. Assume that
k
X xk − 1
S(k) : xi−1 =
i=1
x−1

holds. To be proved is that


k+1
X xk+1 − 1
S(k + 1) : xi−1 =
i=1
x−1

follows. Beginning with the left side of S(k + 1),


k+1
X k
X
i−1
x = xi−1 + xk (evaluate sum for i = k + 1)
i=1 i=1

xk − 1
= + xk (by S(k))
x−1

xk − 1 + xk+1 − xk
= (common denominator)
x−1

xk+1 − 1
= (simplify)
x−1
one arrives at the right side of S(k + 1), thereby showing S(k + 1) is also true, completing
the inductive step.
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 23

Conclusion: By mathematical induction, it is proved that for all n ≥ 1, n ∈ N, the


statement S(n) is true.
Understand and Plan: This problem is similar to Exercise 24.2 except that we are asked to
prove the formula that we come up with. Once we come up with a good conjecture, this
exercise becomes another routine induction problem.

Exercise 24.10 (xiii). Consider the statement ‘A(n) : 2n < 2n−1 ’. Prove that A(n) is false
for all n ∈ N. Show that the inductive step holds, i. e. that A(k) =⇒ A(k + 1). Note that
this shows that we need the initial case to be true to show the statement holds for all n. This
is like ‘The Moon is made of cheese implies that the Moon is a tasty snack’ from Chapter 7.
That is, ‘A(n) =⇒ A(n + 1)’ can be true even if both A(n) and A(n + 1) are false.
Solve: This exercise really consists of two problems:
(a) Prove that A(n) is false for all n ∈ N.
Proof: To prove that A(n) : 2n < 2n−1 is false for all n ∈ N, it suffices to show that
S(n) : 2n ≥ 2n−1 is true for all n ∈ N. In the spirit of this chapter, it will be shown
by induction that S(n) holds for all n ∈ N; however, there is a much easier and concise
algebraic proof readily available££ . For any integer n ≥ 1, let S(n) denote the statement

S(n) : 2n ≥ 2n−1 .

Base step (n = 1): The statement S(1) says that 21 ≥ 20 which is true because 2 ≥ 1.
Inductive step S(k) → S(k + 1): Fix some k ≥ 1, k ∈ N. Assume that

S(k) : 2k ≥ 2k−1

holds. To be proved is that


S(k + 1) : 2k+1 ≥ 2k
follows. Beginning with the left side of S(k + 1),

2k+1 = 2 · 2k
≥ 2 · 2k−1 (by S(k))
= 2k ,

one arrives at the right side of S(k + 1), thereby showing S(k + 1) is also true, completing
the inductive step
Conclusion: By mathematical induction, it is proved that for all n ≥ 1, the statement
S(n) is true. Since S(n) is true, it necessarily follows that A(n) : 2n < 2n−1 is false.
Understand and Plan: It may seem odd that one should prove a statement to be false for all
n ∈ N, and this may confuse someone that encounters this for the first time. Nonetheless, it
££
Simply manipulate the inequalities: 2n ≥ 2n−1 ⇔ 2 ≥ 1 (by dividing both sides by 2n−1 )
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 24

should be clear that you can prove a statement to be false for all n ∈ N if you can prove that
an opposite statement is true for all n ∈ N. Such reasoning is required of us to effectively
solve this problem. Once we realize this, this exercise also becomes another routine induction
problem.

(b) Show that A(k) =⇒ A(k + 1) holds.


Proof: The key to this problem is realizing that you basically do not have to do any work.
It has been established that A(k) is false for some arbitrary k ∈ N in problem (a) above. In
the implication A(k) =⇒ A(k + 1), since the hypothesis, A(k), is false, it necessarily follows
that the implication is true. Why? Recall the truth table for A =⇒ B given on page 66 of
[5]. Any implication will be true if it has a false hypothesis. In this case, A(k) is the false
hypothesis, meaning that A(k) =⇒ A(k + 1) is true even though A(k) and A(k + 1) are both
false.
Understand and Plan: There is nothing particularly difficult concerning this exercise other
than realizing what it is exactly that you are trying to prove and how to write it up. Splitting
the exercise into two problems effectively addresses everything asked of us.

Exercise 24.10 (xv). Show that the generalized Triangle Inequality,



Xn Xn
|xi | ≥ xi ,


i i

is valid for n ≥ 1, n ∈ N.
Solve: Because it must be shown that the inequality holds for all n ∈ N, the lower bounds
for the summations become i = 1 instead of just i. Consequently, for any integer n ≥ 1, let
S(n) denote the statement
Xn Xn
S(n) : |xi | ≥ xi .


i=1 i=1

Inductive step S(k) → S(k + 1): Fix some k ≥ 1, k ∈ N. Assume that



Xk Xk
S(k) : |xi | ≥ xi


i=1 i=1

holds. To be proved is that


k+1
k+1
X X
S(k + 1) : |xi | ≥ xi


i=1 i=1
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 25

follows. Beginning with the left side of S(k + 1),


k+1
X k
X
|xi | = |xi | + |xk+1 | (evaluate sum for i = k + 1)
i=1 i=1


Xk
≥ xi + |xk+1 | (by S(k))


i=1

!
k
X
≥ xi + xk+1 (by (24.2))


i=1

k+1
X
= xi , (express as desired)


i=1

one arrives at the right side of S(k + 1), thereby showing S(k + 1) is also true, completing
the inductive step.
Conclusion: By mathematical induction, it is proved that for all n ≥ 1, n ∈ N, the
statement S(n) is true.
Understand and Plan: The key to solving this problem is to “handle the summations with
care.” That is, be careful in how you logically progress throughout your proof. Besides
using the inductive hypothesis, it is important to use (24.2), a weaker version of the Triangle
Inequality that was used in Exercise 24.10 (v).

Exercise 24.10 (xvi). Suppose that x1 , x2 , . . . , xn are non-negative real numbers. Prove
that
x 1 + x2 + · · · + xn √
≥ n x1 x2 · · · xn . (24.11)
n
Solve: This problem is known for being notoriously difficult to prove [4], but there are
many ways to prove it, ranging from the long and complicated to the somewhat brief and
more simple. The inequality (24.11) is referred to as the AM-GM inequality (arithmetic
mean and geometric mean, respectively). It must first be said that the following proof is
largely adapted from [6], though the proof given here is slightly modified, far more detailed,
and nearly everything is explained—the proof given by the author of [6] was presented to
Iran’s Olympiad team, and readers of this book are not expected to have such mathematical
maturity just yet.
Explanatory remarks and comments are generally made in the “Understand and Plan”
portion immediately following the solution of a problem, but it is more fitting for this problem
to slightly alter this format in order to make understanding the solution more manageable.
Instead of following up the solution with comments, there will be two versions of the proof
given for this exercise. The first version will show the pure mathematics of the proof with
few explanations while the second version will give explanations for almost everything.
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 26

Try to understand how everything works in the first version of the proof before thumbing
to the second version. If you manage to come up with a brand new proof for (24.11) that is
different from the proofs that appear in the vast literature concerning the inequality, then you
should consider publishing your result. The idea behind presenting two versions of the proof,
one short and sweet and the other long and littered with comments, is that you will likely
not be able to understand Version 1 upon your first reading of it. Thus, the anticipation
is that you will look at Version 1 for an overview of what the flow of the proof will be like
and then you will turn to Version 2 to actually understand why Version 1 is a valid proof.
That is, if you trek through all of the comments in Version 2 and understand all that is
presented there, then Version 1 should not be as complicated as it first seems. Indeed, you
should appreciate just how brief and clever the proof presented in Version 1 really is. That
being said, it is time to start the proof!
Version 1: Suppose that x1 , x2 , . . . , xn are non-negative real numbers. For any integer n ≥ 1,
let S(n) denote the statement
x 1 + x2 + · · · + xn √
S(n) : ≥ n x1 x2 · · · xn | n ∈ N.
n

Base step (n = 1): The statement S(1) says that x11 ≥ 1 x1 , which is true because x1 = x1 .
Base step (n = 2): The statement S(2) says that
x1 + x2 √
≥ x1 x 2 , (24.12)
2
which is true because
ab
a ≤ x ≤ b ←→ a + b ≥ x + | 0 < a ≤ b, x > 0. (24.13)
x
Inductive step S(k) → S(k + 1): Fix some k ≥ 1, k ∈ N. Assume that
x1 + x2 + · · · + xk √
S(k) : ≥ k x1 x2 · · · xk
k
holds. To be proved is that
x1 + x2 + · · · + xk+1 √
S(k + 1) : ≥ k+1
x1 x2 · · · xk+1 (24.14)
k+1
follows. If x1 = x2 = · · · = xk+1 , then the proof is done. If not, let x1 x2 · · · xk+1 = ρk+1 .
Without loss of generality, assume that x1 ≤ xi and xi ≤ x2 for all i; that is, assume that
x1 < ρ < x2 . Beginning with the left side of S(k + 1) [excluding the k + 1 divisor],
x1 x2
x1 + x2 + · · · + xk+1 > ρ + + x3 + · · · + xk + xk+1 (base step for n = 2)
ρ
r 
x1 x 2
≥ρ+k· k
x3 · · · xk+1 (by S(k))
ρ

= (k + 1)ρ,
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 27

one arrives at the right side of S(k + 1) [with a k + 1 multiple], thereby showing that S(k + 1)
is also true, completing the inductive step.
Conclusion: By mathematical induction, it is proved that for all n ≥ 1, n ∈ N, the
statement S(n) is true.

Version 2: Suppose that x1 , x2 , . . . , xn are non-negative real numbers. For any integer n ≥ 1,
let S(n) denote the statement
x 1 + x2 + · · · + xn √
S(n) : ≥ n x1 x2 · · · xn | n ∈ N.
n

Base step (n = 1): The statement S(1) says that x11 ≥ 1 x1 , which is true because x1 = x1 .
Base step (n = 2): This is where things start to get somewhat complicated. If it is your
first time encountering the AM-GM inequality, you are likely to start off with the left side
of S(k + 1) in (24.14),
x1 + x2 + · · · + xk+1
,
k+1
but you will quickly see why this problem is notoriously difficult to prove—how can you use
the inductive hypothesis to prove that S(k) → S(k + 1)? That is, how can you start with
the left side of S(k + 1) and logically, i. e. by the inductive hypothesis, end up with the right
side of S(k + 1)?
The idea behind having a base case of n = 2 is to facilitate the progression of the proof,
as is made evident by Version 1. The statement S(2) [i. e. (24.12)] says that
x1 + x2 √
≥ x1 x 2 .
2
How is it possible to verify this? To verify that (24.12) holds, it suffices to prove that (24.13)
holds:
ab
a ≤ x ≤ b ←→ a + b ≥ x + | 0 < a ≤ b, x > 0
x
Why is it sufficient to show this? It is sufficient because the inequality a + b ≥ x + ab x
is
stronger than
√ the AM-GM inequality when n = 2 [i. e. (24.13) → (24.12)]. To see why,
choose x = ab, which yields
ab √ ab √
a+b≥x+ ⇐⇒ a + b ≥ ab + √ (because x = ab)
x ab

ab + ab
⇐⇒ a + b ≥ √ (common denominator)
ab

a+b ab
⇐⇒ ≥√ (add ab terms, divide by 2)
2 ab

a+b √
⇐⇒ ≥ ab. (simplify)
2
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 28


The reason why a + b ≥ x + ab x
is stronger than x1 +x
2
2
≥ x1 x2 should now be apparent.
The inequality a + b ≥ x + ab | 0 < a ≤ b, x > 0 covers the n = 2 case for the AM-GM
x √ √
inequality when x is chosen to be x = ab or x = x1 x2 (where the latter representation of
x preserves the notation used in (24.11)). If one wants to further preserve the notation used
in (24.11), the inequality
ab
a ≤ x ≤ b ←→ a + b ≥ x + | 0 < a ≤ b, x > 0
x
may be represented as
x1 x2
x1 ≤ x ≤ x2 ←→ x1 + x2 ≥ x + | 0 < x1 ≤ x2 , x > 0. (24.15)
x
Note that (24.15) will be useful when carrying out the inductive step in this version of the
proof. For the time being though, observe that (24.15) is a biconditional (←→) statement¶¶ .
This means that in order to prove (24.15), one has to prove the forward direction (−→),
ab
a ≤ x ≤ b −→ a + b ≥ x + | 0 < a ≤ b, x > 0, (24.16)
x
as well as the backward direction (←−),
ab
a+b≥x+ −→ a ≤ x ≤ b | 0 < a ≤ b, x > 0. (24.17)
x
The order of proof for either direction does not matter (simply choose whatever suits your
fancy), but it may be more advantageous in this situation to prove the backward direction
first and then the forward direction∗∗∗ :
(←−):
ab
a+b≥x+ ⇐⇒ x(a + b) ≥ x2 + ab (multiply through by x)
x

⇐⇒ x2 − x(a + b) + ab ≤ 0 (move all terms to one side)

⇐⇒ x2 − xa − xb + ab ≤ 0 (expand)

⇐⇒ (x − a)(x − b) ≤ 0 (factor)

⇐⇒ a ≤ x ≤ b (since 0 < a ≤ b, x > 0)

The key to making the last deduction, (x − a)(x − b) ≤ 0 ⇐⇒ a ≤ x ≤ b, is to consider when


(x − a)(x − b) ≤ 0 is a valid inequality:
¶¶
As mentioned on page 76 of [5], the biconditional (p ↔ q) is logically equivalent to the conjunction
(p → q) ∧ (q → p). Using this notation, let p = a ≤ x ≤ b and q = a + b ≥ x + ab x . If it can be shown that
the forward direction, (p → q), is true and the backward direction, (q → p), is true, then (p → q) ∧ (q → p)
will logically follow and the truth of (24.13) will be validated.
∗∗∗
When proving inequalities, it is important to keep in mind whatever restrictions apply; in this instance,
the restrictions are 0 < a ≤ b and x > 0.
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 29

Case 1 : x = a and x = b : (x − a)(x − b) = 0 ≤ 0. X

Case 2 : x = a or x = b : (x − a)(x − b) = 0 ≤ 0. X

Case 3 : x > a and x < b (wlog††† ): (x − a)(x − b) < 0‡‡‡ . X

The three cases above indicate that (x − a)(x − b) ≤ 0 is a valid inequality when x = a, x = b
or x > a and x < b. This may be written mathematically as a ≤ x ≤ b in light of the
restriction 0 < a ≤ b given at the beginning of the problem statement. Thus, the backward
direction has been proven. The forward direction must now be proven:
(−→):

a ≤ x ≤ b ⇐⇒ (x − a)(x − b) ≤ 0 (since 0 < a ≤ b, x > 0)

⇐⇒ x2 − ax − bx + ab ≤ 0 (expand)

⇐⇒ x2 − x(a + b) + ab ≤ 0 (factor out x)

⇐⇒ x(a + b) ≥ x2 + ab (rearrange)

ab
⇐⇒ a + b ≥ x + (divide by x throughout)
x
Notice how easy it was to prove the forward direction once the backward direction had been
established—this is because the proof for the backward direction was comprised of if and
only if statements, i. e. biconditionals. Thus, a great deal of the work done on the proof for
the backward direction could be applied to the proof for the forward direction. Indeed, all
that had to be done was reverse our proof for the backward direction. Now that the base
step for n = 2 has been explained and verified, it is time to move on to the inductive step.
Inductive step S(k) → S(k + 1): Fix some k ≥ 1, k ∈ N. Assume that
x1 + x2 + · · · + xk √
S(k) : ≥ k x1 x2 · · · xk
k
holds. Notice that S(k) may be equivalently expressed as

S(k) : x1 + x2 + · · · + xk ≥ k · ( k x1 x2 · · · xk )

To be proved is that

S(k + 1) : x1 + x2 + · · · + xk+1 ≥ (k + 1) · [ k+1 x1 x2 · · · xk+1 ]
†††
The phrase “without loss of generality” is often indicated by writing wlog and is often used when what
you are proving applies elsewhere. That is, in this particular case, it is unnecessary to write out a fourth
case where x > b and x < a because x and b are random variables, meaning that the third case takes care
of the third and fourth cases simultaneously.
‡‡‡
Note that < is stronger than ≤, meaning that (x − a)(x − b) < 0 → (x − a)(x − b) ≤ 0.
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 30

follows. If x1 = x2 = · · · = xk+1 , then the proof is done. If not, then we may assume without
loss of generality (wlog) that x1 ≤ xi and xi ≤ x2 for all i, where 1 ≤ i ≤ k + 1 and i ∈ N.
Furthermore, put x1 x2 · · · xk+1 = ρk+1 and hence x1 < ρ < x2 . Two comments are warranted
concerning these statements before giving the inductive proof:

1. wlog, x1 and x2 are minimum and maximum terms of the set {x1 , x2 , . . . , xk+1 }, re-
spectively. That is, whatever two numbers are the minimum and maximum in the
unordered set of numbers {x1 , x2 , . . . , xk+1 } are being denoted by x1 and x2 , respec-
tively. Also, since it is not the case that x1 = x2 = · · · = xk+1 , if the unordered set
of numbers {x1 , x2 , . . . , xk+1 } is ordered in terms of magnitude, then we will get an
ordered set of the form {x1 , . . . , x2 }, and it should then be apparent that x1 6= x2 and
thus x1 < x2 .

2. By putting x1 x2 · · · xk+1 = ρk+1 , observe that ρ = k+1 x1 x2 · · · xk+1 . What is the
significance of this? Consider the possible values of ρ. The smallest value of ρ would
occur when
q
xk+1
k+1
x 1 x1 · · · x1 =
r
ρ= 1 = x1 ,
k+1 | {z }
k + 1 times

given that x1 is the minimum term. On the other hand, the largest value of ρ would
occur when
q
xk+1
k+1
x 2 x2 · · · x2 =
r
ρ= 2 = x2 ,
k+1 | {z }
k + 1 times

given that x2 is the maximum value. Since x1 6= x2 , the actual value of ρ must lie
somewhere between its smallest and largest possible values, namely x1 < ρ < x2 .

The usefulness of the second comment should now be clear, especially in light of all of the
work done to prove (24.15), where it was established that
x1 x2
x1 ≤ x ≤ x2 ←→ x1 + x2 ≥ x + | 0 < x1 ≤ x2 , x > 0.
x
Since it has been deduced, via the second comment, that x1 < ρ < x2 , it may be useful to
reformulate (24.15) as
x1 x2
x1 < ρ < x2 ←→ x1 + x2 > ρ + | 0 < x1 < x2 , ρ > 0. (24.18)
ρ
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 31

Finally, beginning with the left side of S(k + 1),


x1 x2
x1 + x2 + · · · + xk+1 > ρ + + x3 + · · · + xk+1 (by (24.18))
ρ
r 
x1 x 2
≥ρ+k· k
x3 · · · xk+1 (by S(k))
ρ
r 
x1 x2 · · · xk+1
=ρ+k· k
(rearrange)
ρ
s !
ρk+1
=ρ+k· k
(since x1 x2 · · · xk+1 = ρk+1 )
ρ
p 
k
=ρ+k· ρk (simplify)

=ρ+k·ρ (simplify)

= (k + 1)ρ (factor)
√ √
= (k + 1) · [ k+1 x1 x2 · · · xk+1 ] , (ρ = k+1
x1 x2 · · · xk+1 )

one arrives at the right side of S(k + 1), thereby showing S(k + 1) is also true, completing
the inductive step. One should notice three things about the sequence of deductions that
occurs above:

1. The first and last parts of the deductions indicate that



x1 + x2 + · · · + xk+1 > (k + 1) · [ k+1 x1 x2 · · · xk+1 ] ,

which may clearly be rewritten as


x1 + x2 + · · · + xk+1 √
> k+1
x1 x2 · · · xk+1 ,
k+1
perhaps lending a modicum of excitement to the poor soul that has trekked through
this proof—it is pleasing to obtain the form of what was “originally” to be proved,
even if it is just a trivial symbolic curiosity.

2. The application of the inductive hypothesis that yields


r 
x1 x2 x1 x2
ρ+ + x3 + · · · + xk+1 ≥ ρ + k · k
x3 · · · xk+1
ρ ρ

may be somewhat unclear at first, but it makes complete sense if one truly understands
what the inductive hypothesis is actually saying. Recall the inductive hypothesis:

S(k) : x1 + x2 + · · · + xk ≥ k · ( k x1 x2 · · · xk )
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 32

What S(k) is saying is that for k numbers, all of which must be non-negative real
numbers, the inductive hypothesis may be applied. In the context of the sequence of
deductions, this is how the mental picture might look:

x1 x2 x1 x2
x + x2 + · · · + xk+1 > ρ + + x3 + · · · + xk+1 = ρ + + x3 + · · · + xk+1
|1 {z } ρ |{z} ρ
k+1 numbers | {z } 1 number | {z }
k+1 numbers k numbers

This sort of “mental picture” indicates that the inductive hypothesis may be applied
to
x1 x2
+ x3 + · · · + xk+1 ,
ρ
for this expression has k numbers, all of which are non-negative real numbers.
3. At the beginning of the inductive step, the assumption that it was not the case that
x1 = x2 = · · · = xk+1 allowed for the validation of the conclusion
x1 + x2 + · · · + xk+1 √
> k+1 x1 x2 · · · xk+1 ;
k+1
that is, the equality
x1 + x2 + · · · + xk+1 √
= k+1
x1 x2 · · · xk+1
k+1
holds only when x1 = x2 = · · · = xk+1
Conclusion: By mathematical induction, it is proved that for all n ≥ 1, n ∈ N, the
statement S(n) is true.

24.3 Closing Remarks


The brief subsections that follow are meant to supplement concepts used or referred to in
this chapter that could not be more thoroughly developed within the context of the text in
which they appeared.

24.3.1 Explanation of Mathematical Induction’s Validity


The following definition of mathematical induction is adapted from [3]:

Principle of mathematical induction: For some fixed integer b,


and for each integer n ≥ b, let S(n) be a statement involving n. If
(i) S(b) is true, and
(ii) for any integer k ≥ b, S(k) → S(k + 1),
then for all n ≥ b, the statement S(n) is true.
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 33

Mathematical induction is a valid proof technique because of a fundamental axiom con-


cerning the set of positive integers. The following statement of this axiom is adapted from
[1], wherein it is called the Least Integer Principle, but it is often referred to as the Well-
Ordering Principle or wop. The principle is as follows:
Well-Ordering Principle. Every nonempty set of positive integers contains a least ele-
ment.
The validity of mathematical induction is established by using a proof by contradiction.
This proof will contain several “steps” or “parts.” Before giving all of the steps to the proof
of mathematical induction, it may be useful to reformulate the definition of the proof tech-
nique in terms of the notation that will be used throughout the sequence of steps in the
explanation (for consistency and facilitated understanding):

Principle of mathematical induction: For some fixed integer 1,


and for each integer n ≥ 1, let S(n) be a statement involving n. If
(i) S(1) is true, and
(ii) for any integer k ≥ 1, S(k) → S(k + 1),
then for all n ≥ 1, the statement S(n) is true.

1. Start by supposing that S(1) is true and that the proposition S(k) → S(k + 1) is true
for all positive integers k, i. e. where (i) and (ii) hold as indicated above.

2. The goal is to verify whether or not S(n) is true for all n ≥ 1 if S(1) and S(k) → S(k+1)
are true. The statement of mathematical induction in the box above indicates that
S(n) will logically follow if S(1) and S(k) → S(k + 1) are true, but does S(n) really
follow if (i) and (ii) are true? If yes, then mathematical induction is a valid proof
technique. If not, then it is mere rubbish.

3. We are skeptics, and we think that mathematical induction is a sham (hint: a proof
by contradiction is about to take place). Our skepticism induces us to assume that
there is at least one positive integer for which S(n) is false [keep in mind that we are
assuming that (i) and (ii) are true, albeit we are disputing whether or not S(n) really
follows from their truthfulness]. Surely there is at least one positive integer for which
S(n) is false even though S(1) and S(k) → S(k + 1) are true.

4. Let P denote the set of all positive integers for which S(n) is false. Is this set empty?
We think not—after all, we are assuming that there is at least one positive integer, say
`, for which S(n) is false; that is, the assumption is that S(`) is false, where ` ∈ P.
Are there other positive integers in P? Perhaps, but we cannot say for certain at the
moment. We can, however, declare with certainty that P has a least element by the
well-ordering principle. We let the least element of P be ` without loss of generality.

5. Since S(1) is true, we know that ` 6= 1, and because ` is positive and greater than 1,
we also know that ` − 1 must be a positive integer. Moreover, because ` − 1 is less than
`, it should be clear that ` − 1 cannot be in P [this is because ` is the least element
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 34

in P, meaning that any lesser positive integer cannot be in P]. Notationally, we have
it that ` ∈ P and (` − 1) ∈/ P. What does this mean? Since (` − 1) ∈ / P and ` − 1 is
a positive integer, and because P is the set of all positive integers for which S(n) is
false, it must be that S(` − 1) is true.

6. Finally, recall that we maintained that (ii) was true; that is, S(k) → S(k + 1) is true
for any integer k ≥ 1. Since ` and ` − 1 are both positive integers, we may let k = ` − 1
and k +1 = `. Substituting these values into the implication that we assume to be true,
we get that S(` − 1) → S(`). Do you see the problem now (and hence the conclusion
of the proof by contradiction)?

7. By supposing that (i) and (ii) held while also supposing that S(n) was false for some
positive integer `, we deduced through a series of steps that S(` − 1) → S(`) [by (ii)
where k = `−1 and k +1 = `]. What is wrong with this? Simply consider the following
three assertions that occur within the proof:

(a) S(` − 1) → S(`) [True—by supposition (ii)]


(b) S(`) [False—because of the assumption that S(n) was false for ` ∈ Z+ ]
(c) S(` − 1) [True—by the well-ordering principle]

The logical issue should now be apparent. We know S(` − 1) → S(`) is true by our
original supposition, but this implication cannot be true if S(`) is false. Why? As
addressed on page 66 of [5], an implication of the form p → q is only false when the
hypothesis, p, is true and the conclusion, q, is false. In our own case, since S(` − 1)
is true, the implication S(` − 1) → S(`) is only true when S(`) is also true, thus
contradicting the choice of `. Consequently, S(n) must be true for every positive
integer n.

It may take a few readings before fully understanding why mathematical induction is a
valid proof technique, but that effort spent upfront will pay large dividends in the future.

24.3.2 Reindexing the Sum


As noted in [2], the key to success with sums is an ability to change one summation into
another that is either simpler or closer to some goal. One may achieve this in various ways,
often making use of the associative, distributive, and commutative laws for sums. Regardless,
the ability to effectively reindex a summation is indispensable. Consider two examples that
should bring to light how reindexing summations works:
Example 1: Consider the summation given by
3
X
i = 1 + 2 + 3 = 6. (24.19)
i=1

Suppose we want to represent the sum differently. We may use all sorts of algebraic wizardry
to transform (24.19) into a vast array of different sums. For the sake of this example, let’s
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 35

insist that we represent (24.19) as a different sum with a lower bound of i = 3. How can we
do this? Tinkering with the summation yields
5
X
(i − 2) = (3 − 2) + (4 − 2) + (5 − 2) = 6, (24.20)
i=3

which is, of course, equivalent to (24.19). We could even reindex (24.20) as much as we like.
As long as we make sure that the reindexed sum is equivalent to whatever sum it started off
as, there should not be any problems. Now let’s consider a more complicated example, one
that relates back to (24.3).
Example 2: Consider the sum
k  
X k
xk−r y r+1 . (24.21)
r=0
r
Suppose we want the lower bound to start at r = 1 instead of r = 0. How can we accomplish
this? Reindex the sum as follows:
k   k+1   k+1  
X k k−r r+1 X k k−(r−1) (r−1)+1
X k
x y = x y = xk+1−r y r (24.22)
r=0
r r=1
r − 1 r=1
r − 1

How did that happen? Keep a few things in mind when trying to make such a manipulation
(these points will explain the logic behind (24.22)):
1. By shifting r = 0 up to r = 1, we must shift k to k + 1; otherwise, we will not be
evaluating the sum for the same number of terms. If we changed r = 0 to r = 5, then
we would have to change k to k + 5. The lower and upper bounds simply indicate the
number of terms that are to be evaluated.
2. By changing around the upper and lower bounds, we must check to make sure that we
do not change the actual terms that are being summed; that is, we must realize that
k   k+1  
X k k−r r+1 X k k−r r+1
x y 6= x y .
r=0
r r=1
r

This is because the terms that are being summed, kr xk−r y r+1 , have been constructed


with the upper and lower bounds in mind. Thus, if we decide to change the upper and
lower bounds, then we must appropriately change the terms that are being summed
(taking into consideration the new upper and lower bounds). Since we shifted r = 0
up by 1 to r = 1, this suggests that we replace r in the terms to be summed by r − 1
everywhere. Why? Making such an adjustment guarantees that the terms summed
from r = 0 to k will be the same as those summed from r = 1 to k + 1.
3. Trying the case for k = 2 may make it more apparent why the adjustment guarantees
that the correct terms are being summed:
2         3  
X 2 2−r r+1 2 2 2 2 2 3 X 2
x y = x y+ xy + y = x3−r y r
r=0
r 0 1 2 r=1
r−1
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 36

4. Try your own hand at reindexing some random summations. For example, try rein-
dexing (24.21) so that the lower bound starts at r = 5 instead of r = 0. You should
get that
k   k+5  
X k k−r r+1 X k
x y = xk+5−r y r−4 .
r=0
r r=5
r−5

Knowing how to reindex summations effectively is enormously practical in almost all of


mathematics. Its usefulness in calculus, combinatorics, and discrete mathematics should be
readily apparent.

24.3.3 Strength of Statements


What does it mean for a statement to be stronger than another statement? Recall the
definition of tautology given on page 61 of [5]§§§ . Consider the situation where p ↔ q is not
a tautology but one of p → q or q → p is. If p → q is a tautology but q → p is not, then it is
said that p is a stronger statement than q. What does this mean? It means that the truth
of p necessitates the truth of q, but the truth of q is not necessarily accompanied by the
truth of q. See [7] for an extended discussion concerning the strength of statements, but two
examples should clear up any foggy notions related to the whole topic of statement strength.
Example 1: Which statement is stronger, p or p ∧ q? Confirm your guess with a truth table.
Solution: Construct a truth table:

p q p∧q (p ∧ q) → p p → (p ∧ q)
T T T T T
T F F T F
F T F T T
F F F T T

Since (p ∧ q) → p is a tautology while p → (p ∧ q) is not, p ∧ q is stronger than p. This


means that knowing p ∧ q is true guarantees that p is true, but knowing p is true does not
guarantee that p ∧ q is true.
Example 2: Which statement is stronger, p → r or (p ∨ q) → r?
Solution: Writing p → r as Ψ and (p ∨ q) → r as Ω (in order to make writing the truth
table more manageable), we need truth table values for Ψ → Ω and Ω → Ψ. Now construct
the truth table:
§§§
A tautology is a statement such that the truth table has True for all outputs.
CHAPTER 24. TECHNIQUES OF PROOF IV: INDUCTION 37

p q r Ψ:p→r p∨q Ω : (p ∨ q) → r Ψ→Ω Ω→Ψ


T T T T T T T T
T T F F T F T T
T F T T T T T T
T F F F T F T T
F T T T T T T T
F T F T T F F T
F F T T F T T T
F F F T F T T T

Since Ω → Ψ is a tautology while Ψ → Ω is not, it follows that Ω is stronger than Ψ, i. e.


(p ∨ q) → r is stronger than p → r.
Bibliography

[1] J. R. Durbin. Modern Algebra: An Introduction. John Wiley & Sons, Sixth edition, 2009.

[2] R. L. Graham et al. Concrete Mathematics: A Foundation for Computer Science.


Addison-Wesley, Second edition, 1994.

[3] D. S. Gunderson. Handbook of Mathematical Induction: Theory and Applications. CRC


Press, First edition, 2011.

[4] M. D. Hirschhorn. The AM-GM Inequality. The Mathematical Intelligencer, 29(4):7–7,


2007.

[5] K. Houston. How to Think Like a Mathematician: A Companion to Undergraduate


Mathematics. Cambridge University Press, First edition, 2009.

[6] O. A. S. Karamzadeh. One-Line Proof of the AM-GM Inequality. The Mathematical


Intelligencer, 33(2):3–3, 2011.

[7] R. B. Maddox. A Transition to Abstract Mathematics: Mathematical Thinking and Writ-


ing. Academic Press/Elsevier, Second edition, 2009.

[8] G. Polya. Mathematical Discovery: On Understanding, Learning, and Teaching Problem


Solving. Wiley, Combined edition, 1981.

[9] J. Stewart. Calculus. Thomson Brooks/Cole, Fifth edition, 2003.

38

You might also like