112 CHAPTER 2.

MATHEMATICAL INDUCTION

The point is this: when talking with a friend, and verbalizing some ideas,
it might be okay to write “1, 2, 3, . . . , 100” and ensure that whoever is listening
knows exactly what you mean. In general, though, we can’t assume that a reader
would just naturally intuit whatever we were trying to convey; we should be as
explicit and rigorous as possible.
It may seem to you now like we’re nit-picking, but the larger point is that
there is a mathematical way of making this argument more precise, so that it
constitutes a completely valid proof. Everything we have done so far is useful in
guiding our intuition, but we will have to do a little more work to be sure our
arguments are completely convincing. There are a few other concepts required
to make this type of argument rigorous, in general, and we will investigate those
in the next chapter and return to this subject immediately after that. However,
in the meantime, let’s examine one more example to practice this intuitive
argument style and recognizing when induction is an applicable technique.

2.2.2 Lines On The Plane

Take a clean sheet of paper and a pen and a ruler. How many regions are on
your sheet? Just one, right? Draw a line all the way across the paper. Now
there are two regions. Draw another straight line that intersects your first line.
How many regions are there? You should count four in total. Draw a third line
that intersects both of the first two, but not at the point where the first two
intersect. (That is, there should be three intersection points, in total.) How
many regions are there? Can you predict the answer before counting? What
happens when there are 4 lines? Or 5? Or 100? How do we approach this
puzzle and, ultimately, solve it? Let’s give a more formal statement to be sure
we’re thinking the same way:
Consider n lines on an infinite plane (two-dimensional surface) such that no
two lines are parallel and no more than two lines intersect at one point. How
many distinct regions do the lines create?
We can draw a few examples by hand when n is small (up to, say, n = 5
is reasonable), and let’s use this to guide our intuition into making a general
argument for an arbitrary value of n. (Notice that this strategy is very similar to
what we did in the previous puzzle: identify a pattern with small cases, identify
the relevant components of those cases that can generalize, then abstract to an
aribtrary case.) Specifically, we want to attempt to identify how the number
of regions in one drawing depends on the number of regions in a drawing with
fewer lines. What happens when we draw a new line? Can we determine how
this changes the already existing regions? Can we somehow count how many
regions this creates? Do some investigation of this puzzle on your own before
reading on. If you figure out some results, compare your work to the steps we
follow below.
Let’s start with a small case, say n = 2. We know one line divides a plane
into 2 regions; what happens when we add a second line? We know we get 4
regions, because we can just look and count them:
2.2. EXAMPLES AND DISCUSSION 113

2
1
3
4

However, we are only looking at one specific case of two intersecting lines. How
do we know that we will always find four regions, no matter how we draw those
two lines appropriately? That is, can we describe how this happens in a way
that somehow incorporates the fact that the number of liens is n = 2? Think
about it!
Here’s our approach. Notice that each of the already existing regions is split
into two when we add a second line, and that this is true no matter how you
choose to draw the lines; as long as we make sure the two lines aren’t parallel,
they will always behave this way. That is, if we take one line that splits the
plane into two regions,

2
1

then adding a new line will split each of those existing regions in two. This adds
two new regions to the whole plane, giving four regions in total:

2 n2 split
Regio
1
3
it
n 1 spl 4
Regio

What about when n = 3? In this case, we want to think about adding a

third line to a diagram with two lines and four regions. We want to make
an argument that doesn’t depend on a particular arrangement of the lines, so
eventually the only facts we should use are that no lines are parallel and any
point of intersection only lies on two lines (not three or more). For now, though,
114 CHAPTER 2. MATHEMATICAL INDUCTION

it helps to look at a particular arrangement of lines so that we are talking about

the same diagram; we can use our observations of this specific diagram to guide
our general argument. Let’s start with a two-line diagram, on the left below,
and add a third line, but let’s choose the third line so that all of the intersection
points are “nearby” or within the scope of the diagram, so that we don’t have
to rescale the picture:

2
5 Region 2 split 7
Region 1 split
6 Region 3 split
1
3
4

We certainly have 7 regions now, but we made the third line a separate color
so that we can identify where the “new” regions appear: one region (the lower
quadrant, Region 4) remains unchanged, but the three other regions are split in
two and each of those “splits” adds 1 to our count (where there was 1 region,
now there are 2). What if we had placed the line differently?

6 2
Region 2
split
Region
1 split
5 3
1
4
7 Region
4 split

The same phenomenon occurs, where one quadrant remains untouched but
the other three are split in two. (How do we know there aren’t any other
regions not depicted within the scale of our diagram? This is not as easy a
question to answer as you might think at first blush, and it’s worth thinking
about.) Experiment with other arrangements of the three lines and try to
convince yourself that this always happens; also, think about why this is the
case and how we could explain that this must happen. Before giving a general
explanation, though, let’s examine another small case.
When n = 4, we start with three lines and 7 regions and add a fourth line
that is not parallel to any of the existing lines and doesn’t pass through any
existing intersection points. Again, we will want to make an argument that isn’t
tied to a specific arrangement of the lines, but looking at the following specific
diagram will help guide our intuition into making that argument:
2.2. EXAMPLES AND DISCUSSION 115

5 9 2
7
11
6
1 8 3

4
10

Notice that three of the original regions remain unchanged (Regions 3 and
5 and 7), and the other four are split in two. Do you notice a pattern here?
It seems like for every n we’ve examined, adding the n-th line leaves exactly
n − 1 regions unchanged while the rest are split in two. Let’s try to explain why
this happens. Remember that we’re trying to identify how many regions appear
when we draw n lines, so let’s assign that value a “name” so we can refer to it;
let’s say R(n) represents the number of regions created by drawing n lines on
the plane so that no two lines are parallel and no intersection point belongs to
more than two lines. In these examples we’ve considered for small values of n,
we’ve looked at what changes when we add a new line; that is, we’ve figured out
what R(n) is by already knowing R(n − 1). Let’s try to adapt our observations
so that they apply to any arbitrary value of n.
Assume that we know R(n) already. (Why can we do this? Do we know any
particular value of R(n) for sure, for some specific n? Which? How?) Say we
have an arbitrary diagram of n lines on the plane that satisfy the two conditions
given in the puzzle statement above. How many regions do these lines create?
Yes, exactly R(n). Now, what happens when we add the (n + 1)-th line? What
can we say for sure about this line and how it alters the diagram? Well, the
only information we really have is that (a) this new line is not parallel to any of
the existing n lines and (b) this new line does not intersect any of the already
existing intersection points. Now, condition (a) tells us that this new line must
intersect all of the exisiting n lines; parallel lines don’t intersect, and non-parallel
lines must intersect somewhere. Thus, we must create n new intersection points
on the diagram. Can any of those intersection points coincide with any existing
intersection points? No! This is precisely what condition (b) tells us. These
two pieces of information together tell us that, no matter how we draw this new
line, as long as it satisfies the requirements of the puzzle, we must be able to
identify n “special” points along that line. Those special points are precisely
the points where the new line intersects an existing line.
We’d now like to take these special points and use them to identify new
regions in the diagram. Look back to the cases we examined above: identify
the new intersection points and see if you can associate them with new regions.
Perhaps it would help to label those intersections with a large dot and mark
the new regions with an X to make them all stand out. We’ll show you one
example below, where n = 4. What do you notice? Can you use these dots to
116 CHAPTER 2. MATHEMATICAL INDUCTION

help identify how many new regions are created with the addition of that n-th
line? Think about this for a minute and then read on.

Exactly! Between any two of the new intersection points, we have a line
segment that splits a region in two! All that remains is to identify how many
new such segments we’ve created. Since each one corresponds to exactly one
existing region split in two, this will tell us exactly how many new regions
we’ve created. We’ve already figured out that this (n + 1)-th line creates n new
intersection points. Think about how these points are arranged on the line. Any
two “consecutive” points create a segment, but the extreme points also create
infinite segments (that coninue past those extreme points forever). How many
are there in total? Exactly n + 1. (Look at the diagram above, for n = 3. We
see that there are 3 new intersection points and 4 new segments, with two of
them being infinite rays.) This means there are n + 1 line segments that split
regions in two, so we have created exactly n + 1 new regions! This allows us to
say that
R(n + 1) = R(n) + n + 1
Phew, what an observation! It took a bit of playing around with examples
and making some geometric arguments, but here we are. We’ve identified some
inductive structure to this puzzle; we’ve found how one case depends on another
one. That is, we’ve found how R(n+1) depends on R(n). This hasn’t completely
solved the puzzle, but we are now much closer. All that remains is to replace
R(n) with a similar expression, and continually do this until we reach a value
we know, R(1) = 2. Observe:
R(n + 1) = R(n)
 + n+1
= R(n
 −1) + n + n+1
=  − 2) + (n − 1) +
R(n n + n+1
 

..
.
= R(2)

+ 3
 + ··· + n + n+1
= R(1) + 2 + 3 + ··· + n + n+1
Since we know R(1) = 2, we can say
n+1 n+1
!
X X
R(n + 1) = 2 + (2 + 3 + · · · + n + (n + 1)) = 2 + k −1=1+ k
k=1 k=1
2.2. EXAMPLES AND DISCUSSION 117

and this is a sum we have investigated before! (Also notice that we had to
subract 1 because of the missing first term of the sum in parentheses.) Recall
Pn
that k=1 k = n(n+1)
2 , and to represent the sum we have in the equation above,
we just replace n with n + 1. Therefore,
(n + 1)(n + 2)
R(n + 1) = 1 +
2
One final simplification we would like to make is to replace n+1 with n through-
out the equation, because it makes more sense to have an expression for R(n)
(For what values of n is this valid?)

n(n + 1)
R(n) = 1 +
2
Finally, we have arrived at an answer to the originally-posed puzzle! In so doing,
we employed an inductive technique: we explained how one “fact”, namely the
value of R(n + 1), depends on the value of a “previous fact”, namely R(n),
and used these iterative dependencies to work backwards until we reached a
particular, known value, namely R(1).
We want to point out, again, that the derivation we followed and the obser-
vations we made in this section have guided our intuition into an answer, but
this has not rigorously proven anything. The issue is with the “· · · ”, which is
not a concrete, “officially” mathematical way of capturing the inductive pro-
cess underlying our technique. Furthermore, our method with the “lines in the
plane” problem had us starting with a diagram of n − 1 lines and building a
new diagram with n lines; is this okay? Why does this actually tell us anything
about an arbitrary diagram of n lines? Do all such diagrams come from a smaller
diagram with one fewer line?
We will, in the next two chapters, learn the necessary tools to fully describe
a rigorous way of doing what we have done so far, and in the chapter after
that, we will employ those tools to make mathematical induction officially
rigorous. For now, though, we want to give a heuristic definition of induction and
continue to examine interesting puzzles and observations that rely on inductive
techniques. Practicing with these types of puzzles–learning when to recognize
an inductive process, how to work with it, how to use that structure to solve
a problem, and so on–will be extremely helpful in the future, and we have no
need to delve into technical mathematical detail. (At least, not just yet!)

2.2.3 Questions & Exercises

Remind Yourself
Answering the following questions briefly, either out loud or in writing. These
are all based on the section you just read, so if you can’t recall a specific defi-
nition or concept or example, go back and reread that part. Making sure you
can confidently answer these before moving on will help your understanding and
memory!
118 CHAPTER 2. MATHEMATICAL INDUCTION

(1) What properties characterize an inductive process?

n
X n(n + 1)
(2) How did we prove that k= is correct? How was our method
2
k=1
inductive? (Reread Section 1.4.2 if you forget!)
(3) Why can we take the sum formula mentioned in the previous question and
“replace” n with n+1 and know that it still holds true? Can we also replace
n with n − 1?
(4) Work through the algebraic steps to obtain our final expression for the sum
of the first n squares; that is, verify that
1 1 n(n + 1) 1
(n + 1)3 − (n + 1) − = n(n + 1)(2n + 1)
3 3 2 6

(5) Try to recall the argument that adding the (n + 1)-th line on the plane
created exactly n + 1 new regions. Can you work through the argument for
a friend and convince him/her that it is valid?
(6) To find the sum of the first n squares, why couldn’t we just square the
formula for the sum of the first n numbers? Why is that wrong?

Try It
Try answering the following short-answer questions. They require you to actu-
ally write something down, or describe something out loud (to a friend/class-
mate, perhaps). The goal is to get you to practice working with new concepts,
definitions, and notation. They are meant to be easy, though; making sure you
can work through them will help you!
(1) Draw 5 lines on the plane (that satisfy the two conditions of the puzzle)
and verify that there are 16 regions. Can you also verify that 6 lines yield
22 regions?
(2) Come up with another description of a sequence that goes 1, 2, 3, 4, . . . , 100
that is not simply all of the numbers from 1 to 100. (Recall the example we
gave: all the numbers from 1 to 100 with no “i” in their English spelling.)
(3) Come up with an algebraic expression that relates (n + 1)4 to n4 , like we
did with cubes.
(Challenge: Can you come up with a geometric interpretation for the
expression you just derived?)
(4) Challenge: Let’s bump the “lines in the plane” puzzle up one dimension!
Think about having n planes in three-dimensional space. How many regions
are created? Assume that no two planes are parallel, and no three of them
intersect in one line. (Think about how these two conditions are directly
analogous to the specified conditions for the “lines” puzzle.)
2.3. DEFINING INDUCTION 119

2.3 Defining Induction

To properly motivate the forthcoming definition of mathematical induction
as a proof technique, we want to emphasize that the above examples used some
intuitive notions of the structure of the puzzle to develop a “solution”, where
we use quotation marks around solution to indicate that we haven’t officially
proven it yet. In that sense, we ask the following question: What if we had been
given the formula that we derived and asked to verify it? What if we had not
gone through any intuitive steps to derive the formula and someone just told
us that it is correct? How could we check their claim? The reason we ask this
is because we are really facing that situation now, except the person telling us
the formula is . . . the very same intuitive argument we discovered above!
Pretend you have a skeptical friend who says, “Hey, I heard about this
formula for the sum of the first n natural numbers squared. Somebody told me
that they add up to 61 n(n + 1)(2n + 1). I checked the first two natural numbers,
and it worked, so it’s gotta be right. Pass it on!” Being a logical thinker, but
also a good friend, you nod along and say, “I did hear that, but let’s make sure
it’s correct for every number.” How would you proceed? Your friend is right
that the first few values “work out” nicely:
1
12 = 1 = (1)(2)(3)
6
1
12 + 22 = 5 = (2)(3)(5)
6
1
12 + 22 + 32 = 14 = (3)(4)(7)
6
1
12 + 22 + 32 + 42 = 30 = (4)(5)(9)
6
and so on. We could even check, by hand, a large value of n, if we wanted to:
1
12 + 22 + 32 + 42 + 52 + 62 + 72 + 82 + 92 + 102 = 385 = (10)(11)(21)
6
Remember, though, that this formula is claimed to be valid for any value of
n. Checking individual results by hand would take forever, because there are
an infinite number of natural numbers. No matter how many individual values
of n we check, there will always be larger values, and how do we know that
the formula doesn’t break down for some large value? We need a far more
efficient procedure, mathematically and temporally speaking, to somehow verify
the formula for all values of n in just a few steps. We have an idea in mind,
of course (it’s the upcoming rigorous version of mathematical induction), and
here we will explain how the procedure works, in a broad sense.

2.3.1 The Domino Analogy

Pretend that we have a set of dominoes. This is a special set of dominoes because
we have an infinite number of them (!) and we can imagine anything we want
120 CHAPTER 2. MATHEMATICAL INDUCTION

written on them, instead of the standard array of dots. Let’s also pretend that
they are set up in an infinite line along an infinite tabletop, and we are viewing
the dominos from the side and we can see a label under each one so that we
know where we are in the line:

n=1 n=2 n=3 n=4 n=5

For this particular example, to verify the formula

n
X 1
k2 = n(n + 1)(2n + 1)
6
k=1

we will imagine a particular “fact” written on each domino. Specifically, we will

imagine that the 1st domino has the expression
1
X 1
k2 = (1)(2)(3)
6
k=1

written on it, and the 2nd domino has the expression

2
X 1
k2 = (2)(3)(5)
6
k=1

written on it. In general, we imagine that the n-th domino in the infinite line
has the following “fact” written on it:
n
X 1
k2 = n(n + 1)(2n + 1)
6
k=1

Since we’re dealing with dominos that are meant to fall into each other and
knock each other over, let’s pretend that whenever a domino falls, that means the
corresponding “fact” written on it is a true statement. This is how we will relate
our physical interpretation of the dominos to the mathematical interpretation
of the validity of the formula we derived.
We did check the sum for n = 1 by hand: 12 = 61 (1)(2)(3). Thus, the fact
written on the first domino is a true statement, so we know that the first domino
will, indeed, fall over. We also checked the sum for n = 2 by hand, so we know
that the second domino will fall over:
2.3. DEFINING INDUCTION 121

n=1 n=2 n=3 n=4 n=5

However, continuing like this brings us back to the same problem as before: we
don’t want to check every individual domino to make sure it falls. We would
really like to encapsulate our physical notion of the line of dominos—namely,
that when a domino falls it will topple into the next one and knock that over, and
so on—and somehow relate the “facts” that are written on adjacent dominos.
Let’s look at this situation for the first two dominos. Knowing that Domino
1 falls, can we guarantee that Domino 2 falls without rewriting all of the terms
of the sum? How are the statements written on the two dominos related? Each
statement is a sum of squared natural numbers, and the one on the second
domino has exactly one more term. Thus, knowing already that Domino 1 has
fallen, we can use the true statement written on Domino 1 to verify the truth
of the statement written on Domino 2:

2
X 1
k 2 = 12 + 22 = 1 + 22 = 5 = (2)(3)(5)
6
k=1

Now, this may seem a little silly because the only “work” we have saved is not
having to “do the arithmetic” to write 12 = 1. Let’s use this procedure on a
case with larger numbers so we can more convincingly illustrate the benefit of
this method. Let’s assume that Domino 10 has fallen. (In case you are worried
about this assumption, we wrote the full sum a few paragraphs ago and you can
verify it there.) This means we know that

10
X 1
k2 = (10)(11)(21) = 385
6
k=1

is a true statement. Let’s use this to verify the statement written on Domino
11, which is
11
X 1
k2 = (11)(12)(23)
6
k=1

The sum written on Domino 11 has 11 terms, and the first 10 are exactly the
sum written on Domino 10! Since we know something about that sum, let’s just
separate that 11th term from the sum and apply our knowledge of the other
122 CHAPTER 2. MATHEMATICAL INDUCTION

terms:
11
X
k 2 = 12 + 22 + · · · + 102 + 112

k=1
10
!
X
2
= k + 112
k=1
= 385 + 121
= 506
1 1
= 3036 = (11)(12)(23)
6 6
Look at all of the effort we saved! Why bother reading the first 10 terms of the
sum if we know something about them already?
Now, imagine if we could do this procedure for all values of n, simultaneously!
That is, imagine that we could prove that any time Domino n falls, we are
guaranteed that Domino (n + 1) falls. What would this tell us? Well, think
about the infinite line of dominos again. We know Domino 1 will fall, because
we checked that value by hand. Then, because we verified the “Domino n knocks
over Domino (n + 1)” step for all values of n, we know Domino 1 will fall into
Domino 2, which in turn falls into Domino 3, which in turns falls into Domino
4, which . . . The entire line of dominos will fall! In essence, we could collapse
the whole line of dominos falling down into just two steps:
(a) Make sure the first domino topples;
(b) Make sure every domino knocks over the one immediately after it in line.
With only these two steps, we can guarantee every domino falls and, therefore,
prove that all of the facts written on them are true. This will prove that the
formula we derived is valid for every natural number n.
We have already accomplished step (a), so now we have to complete step
(b). We have done this for specific cases in the previous paragraphs (Domino
1 topples Domino 2, and Domino 10 topples Domino 11), so let’s try to follow
along with the steps of those cases and generalize to an arbitrary value of n. We
assume, for some specific but arbitrary value of n, that Domino n falls, which
tells us that the equation
n
X 1
k2 = n(n + 1)(2n + 1)
6
k=1

is a true statement. Now, we want to relate this to the statement written on

Domino (n + 1) and apply the knowledge given in the equation above. Let’s do
what we did before and write a sum of n + 1 terms as a sum of n terms plus the
last term:
n+1 n
!
X X
2 2 2 2 2
k = 1 + 2 + · · · + n + (n + 1) = k + (n + 1)2
2

k=1 k=1
2.3. DEFINING INDUCTION 123

Next, we can apply our assumption that Domino n has fallen (which tells us
that the fact written on it is true) and write
n+1
X 1
k2 = n(n + 1)(2n + 1) + (n + 1)2
6
k=1

Is this the same as the fact written on Domino (n + 1)? Let’s look at what that
is, first, and then compare. The “fact” on Domino (n + 1) is similar to the fact
on Domino n, except everywhere we see “n” we replace it with “n + 1”:
n+1
X 1 1
k2 = (n + 1)((n + 1) + 1)(2(n + 1) + 1) = (n + 1)(n + 2)(2n + 3)
6 6
k=1

It is not clear yet whether the expression we have derived thus far is actually
equal to this. We could attempt to simplify the expression we’ve derived and
factor it to make it “look like” this new expression, but it might be easier to just
expand both expressions and compare all the terms. (This is motivated by the
general idea that expanding a factored polynomial is far easier than recognizing
a polynomial can be factored.) For the first expression, we get

1 1
n(n + 1)(2n + 1) + (n + 1)2 = n(2n2 + 3n + 1) + (n2 + 2n + 1)
6 6
1 1 1
= n3 + n2 + n + n2 + 2n + 1
3 2 6
1 3 3 2 13
= n + n + n+1
3 2 6
and for the second expression, we get
1 1
(n + 1)(n + 2)(2n + 3) = (n + 1)(2n2 + 7n + 6)
6 6
1 3
(2n + 7n2 + 6n) + (2n2 + 7n + 6)

=
6
1 3 13
= n3 + n2 + n + 1
3 2 6
Look at that; they’re identical! Also, notice how much easier this was than try-
ing to rearrange one of the expressions and “morph” it into the other. We proved
they were identical by manipulating them both and finding the same expression,
ultimately. Now, let’s look back and assess what we have accomplished:

1. We likened proving the validity of the formula

n
X 1
k2 = n(n + 1)(2n + 1)
6
k=1

for all values of n to knocking over an infinite line of dominos.

124 CHAPTER 2. MATHEMATICAL INDUCTION

2. We verified that Domino 1 will fall by checking the formula corresponding

to that case by hand.
3. We proved that Domino n will fall into Domino (n+1) and knock it over by
assuming the fact written on Domino n is true and using that knowledge
to show that the fact written on Domino (n + 1) must also be true.
4. This guarantees that all dominos will fall, so the formula is true for all
values of n!
Are you convinced by this technique? Do you think we’ve rigorously proven
that the formula is valid for all natural numbers n? What if there were a value
of n for which the formula didn’t hold? What would that mean in terms of our
domino scheme?
Remember that this domino analogy is a good intuitive guide for how in-
duction works, but it is not built on mathematically rigorous foundations. That
will be the goal of the next couple of chapters. For now, let’s revisit the other
example we’ve examined in this section: lines in the plane. Again, the use of
ellipses in our derivation of the formula R(n) is troublesome and we want to
avoid it. Let’s try to follow along with the domino scheme in the context of this
puzzle.
Imagine that we have defined the expression R(n) to represent the number
of distinct regions in the plane created by n lines, where no two lines are parallel
and no three intersect at one point. Also, imagine that on Domino n we have
written the “fact” that “R(n) = 1 + n(n+1) 2 ”. Can we follow the same steps as
above and verify that all the dominos will fall?
First, we need to check that Domino 1 does, indeed fall. This amounts to
verifying the statement: “R(1) = 1+ 1(2) 2 = 1+1 = 2”. Is this a true statement?
Yes, of course, we saw this before; one line divides the plane into two regions.
Second, we need to prove that Domino n will topple into Domino (n + 1) for
any arbitrary value of n. That is, let’s assume that “R(n) = 1 + n(n+1) 2 ” is a
(n+1)(n+2)
true statement for some value of n and show that “R(n + 1) = 1 + 2 ”
must also be a true statement. How can we do this? Well, let’s follow along
with the argument we used before to relate R(n + 1) to R(n). By considering
the geometric consequences of adding an extra line to any diagram with n lines
(that also fit our rules about the lines) we proved that R(n + 1) = R(n) + n + 1.
Using this knowledge and our assumption about Domino n falling, we can say
that
n(n + 1)
R(n + 1) = R(n) + n + 1 = 1 + +n+1
2
Is this the same expression as what is written on Domino (n + 1)? Again, let’s
simplify both expressions to verify they are the same. We have
n(n + 1) n2 + n 1 3
1+ +n+1=2+n+ = n2 + n + 2
2 2 2 2
and
(n + 1)(n + 2) n2 + 3n + 2 1 3
1+ =1+ = n2 + n + 2
2 2 2 2
2.3. DEFINING INDUCTION 125

Look at that; they’re identical! Thus, we have shown that Domino n is guaran-
teed to fall into Domino (n + 1), for any value of n. Accordingly, we can declare
that all dominos will fall!
Think about the differences between what we have done with this “domino
technique” and what we did before to derive the expressions we just proved.
Did we use any ellipses in this section? Why is it better to prove a formula
this way? Could we have used the domino induction technique to derive the
formulas themselves?

2.3.2 Other Analogies

The Domino Analogy is quite popular, but it’s not the only description of how
induction works. Depending on what you read, or who you talk to, you might
learn of a different analogy, or some other kind of description altogether. Here,
we’ll describe a couple that we’ve heard of before. It will help solidify your
understanding of induction (at least as far as we’ve developed it) to think about
how these analogies are all equivalent, fundamentally.

Mojo the Magical, Mathematical Monkey

Imagine an infinite ladder, heading straight upwards into the sky. There are
infinitely-many rungs on this ladder, numbered in order: 1, 2, 3, and so on
going upwards. Our friend Mojo happens to be standing next to this ladder.
He is an intelligent monkey, very interested in mathematics but also a little bit
magical, because he can actually climb up this infinite ladder!
If Mojo makes it to a certain rung on the ladder, that means the fact cor-
responding to that number is True. How can we make sure he climbs up the
entire ladder? It would be inefficient to check each rung individually. Imagine
that: we would have to stand on the ground and make sure he got to Rung 1,
then we would have to look up a bit and make sure he got to Rung 2, and then
Rung 3, and so on . . . Instead, let’s just confirm two details with Mojo before he
starts climbing. Is he going to start climbing? That is, is he going to make it to
Rung 1? If so, great! Also, are the rungs close enough together so that he can
always reach the next one, no matter where he is? If so, even greater! These
are exactly like the conditions established in our Domino Analogy. To ensure
that Mojo gets to every rung, we just need to know he gets to the first one and
that he can always get to the next one.

Doug the Induction Duck

Meet Doug. He’s a duck. He also loves bread, and he’s going to go searching
through everyone’s yards to find more bread. These yards are all along Induction
Street in Math Town, where the houses are numbered 1, 2, 3, and so on, all in
a row.
Doug starts in the yard of house #1, looking for bread. He doesn’t find any,
so he’s still hungry. Where else can he look? The house next door, #2, has a
126 CHAPTER 2. MATHEMATICAL INDUCTION

backyard, too! Doug heads that way, his tummy rumbling. He doesn’t find any
bread there, either, so he has to keep looking. He already knows house #1 has
no bread, so the only place to go is next door to house #3. We think you see
where this is going . . .
If we were keeping track of Doug’s progress, we might wonder whether he
eventually gets to every yard. Let’s say we also knew ahead of time that no-
body has any bread. This means that whenever he’s in someone’s yard, he will
definitely go to the next house, still searching for a meal. This means that he
will definitely get to every house! That is, no matter which house we live in, no
matter how large the number on our front door might be, at some point we will
see Doug wandering around our backyard. (Unfortunately, he will go hungry all
this time, though! Poor Doug.)

2.3.3 Summary
Let’s reconsider what we’ve accomplished with the two example puzzles we’ve
seen thus far, and the analogies we’ve given. In our initial consideration of each
puzzle, we identified some aspect of the structure of the puzzle where a “fact”
depended on a “previous fact”. In the case of the cubic numbers, we found a
way to express (n + 1)3 in terms of n3 ; in the case of the lines in the plane,
we described how many regions were added when an extra line was added to
a diagram with n lines. From these observations, we applied this encapsulated
knowledge over and over until we arrived at a “fact” that we knew, for a “small”
value of n (in both cases, here, n = 1). This allowed us to derive a formula or
equation or expression for a general fact that should hold for any value of n.
This work was interesting and essential for deriving these expressions, but
it was not enough to prove the validity of the expressions. In doing the work
described above, we identified the presence of an inductive process and utilized
its structure to derive the expressions in question. This was beneficial in two
ways, really; we actually found the expressions we wanted to prove and, by
recognizing the inductive behavior of the puzzle, we realized that proving the
expressions by mathematical induction would be prudent and efficient.
For the actual “proof by induction”, we followed two main steps. First,
we identified a “starting value” for which we could check the formula/equation
by hand. Second, we assumed that one particular value of n made the cor-
responding formula hold true, and then used this knowledge to show that the
corresponding formula for the value n + 1 must also hold true. Between those
two steps, we could rest assured that “all dominos will fall” and, therefore, the
formulas would hold true for all relevant values of n.

One Concern: What’s at the “top” of the ladder?

You might be worried about something, and we’ll try to anticipate your question
here. (We only bring this up because it’s a not uncommon observation to make.
If you weren’t thinking about this, try to imagine where the idea would come
from.) You might say, “Hey now, I think I see how Mojo is climbing the ladder,
2.3. DEFINING INDUCTION 127

but how can he actually get all the way to the top? It’s an infinite ladder, right?
And he never gets there . . . right?”
In a way, you would be right. Since this magical ladder really does go
on forever, then there is truly no end to it and Mojo will never get “there”.
However, that isn’t the point; we don’t care about any “end” of the ladder (and
not just because there isn’t one). We just need to know that Mojo actually gets
to every possible rung. He doesn’t have to surpass all of them and stand at the
top of the ladder, looking down at where he came from. That wasn’t the goal!
Think of it this way: pretend you have a vested interest in some particular
fact that we’re proving. Let’s say it’s Fact #18,458,789,572,311,000,574,003.
(Some huge number. It doesn’t matter, really.) Its corresponding rung is
waaayyyyyy up there on the ladder, and all you care about is whether or not
Mojo gets there on his journey. Does he? . . . You bet he does! It might take a
long time (how many steps will it take?), but in this magical world of monkeys
and ladders, who cares about time anyway! You know that he’ll eventually
get there, and that makes you happy. Now, just imagine that for each fact,
there’s somebody out there in that magical world that cares about only that
fact. Surely, everyone will be happy with the knowledge that Mojo will get to
their rung on his journey. Nobody cares about whether he gets to the top; that
isn’t their concern. Meanwhile, out here in our regular, non-magical world, we
are extremely happy with the fact that everyone in that world will eventually be
happy. That entire infinite process of ladder-climbing was condensed into just
two steps, and with only those two steps, we can rest assured that every rung
on that ladder will be touched. Every numbered fact is true.
Think about this in terms of the Domino Analogy, as well. Do we care
whether or not there is some “ending point” of the line of dominoes, so that
they all fall into a wall somewhere? Of course not; the line goes on forever.
Every domino will eventually fall over, and we don’t even care how “long” that
takes. Likewise, we know Doug will get to everyone’s yard; we don’t care “when”
he gets to any individual yard, just that he gets to all of them.

2.3.4 Questions & Exercises

Remind Yourself
Answering the following questions briefly, either out loud or in writing. These
are all based on the section you just read, so if you can’t recall a specific defi-
nition or concept or example, go back and reread that part. Making sure you
can confidently answer these before moving on will help your understanding and
memory!

(1) How are the Domino, Mojo, and Doug analogies all equivalent? Can you
come up with some “function” that describes their relationship, that con-
verts one analogy into another?

(2) Find a friend who hasn’t studied mathematical induction before, and try to
describe it. Do you find yourself using one of the analogies? Was it helpful?
128 CHAPTER 2. MATHEMATICAL INDUCTION

(3) Why is it the case that our work with the cubes didn’t prove the summation
formula? Why did we still need to go through all that work?

(4) Think about the Domino Analogy. Is it a problem that the line of dominoes
goes on forever? Does this mean that there are some dominoes that will
never fall down? Try to describe what this means in terms of the analogy.

Try It

Try answering the following short-answer questions. They require you to actu-
ally write something down, or describe something out loud (to a friend/class-
mate, perhaps). The goal is to get you to practice working with new concepts,
definitions, and notation. They are meant to be easy, though; making sure you
can work through them will help you!

(1) Work through the inductive steps to prove the formula

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

(2) Work through the inductive steps to prove the formula

n
X
(2k − 1) = n2
k=1

(3) Work through the inductive steps to prove the formula

n  2
X n(n + 1)
k3 =
2
k=1

(4) Suppose we have a series of facts that are indexed by natural numbers. Let’s
use the expression “P (n)” to represent the n-th fact.

(a) If we want to prove every instance is True, for every natural number n,
how can we do this?
(b) What if we want to prove that only every even value of n makes a True
statement? Can we do this? Can you come up with a modification of
one of the analogies we gave that would describe your method?
(c) What if we want to prove that only every value of n greater than or
equal to 4 makes a True statement? Can we do this? Can you come up
with a modification of one of the analogies we gave that would describe
your method?
2.4. TWO MORE (DIFFERENT) EXAMPLES 129

2.4 Two More (Different) Examples

This short section serves a few purposes. For one, we don’t want you to get the
idea, right away, that induction is all about proving a numerical formula with
numbers and polynomials. Induction is so much more useful than that! One
of the following examples, in particular, will be about proving some abstract
property is true for any “size” of the given situation. You will see how it still
falls under the umbrella of “induction”, but you will also notice how it is dif-
ferent from the previous examples. Furthermore, these examples will illustrate
that sometimes we need to know “more information” to knock over some domi-
noes. In the previous examples, we only needed to know that Domino n fell to
guarantee that Domino n + 1 will fall. Here, though, we might have to know
about several previous dominoes. After these two examples, we will summarize
how this differs from the domino definition given above, and preview a broader
definition of the technique of induction, as it applies to these examples.

2.4.1 Dominos and Tilings

This next example is a little more complicated than the first two. We will still
end up proving a certain numerical formula, but the problem is decidedly more
visual than just manipulating algebraic expressions. Furthermore, we’ll notice
an interesting “kink” in the starting steps, where we have to solve a couple of
“small cases” before being able to generalize our approach. This will be our first
consideration of how the technique of induction can be generalized and adapted
to other situations.
The question we want to answer is nicely stated as follows:

Given a 2 × n array of squares, how many different ways can we tile

the array with dominoes? A tiling must have every square covered
by one–and only one–domino.

For example, the following are proper tilings

whereas the following are not proper tilings

130 CHAPTER 2. MATHEMATICAL INDUCTION

As before, let’s examine the first few cases–where n = 1, 2, 3, and so on–and

see if we notice any patterns. Try working with the problem yourself, before
reading on, even!
When n = 1, we have an array that is exactly the shape of one domino, so
surely there is only one way to do this. Let’s use the notation T (n) to represent
the number of tilings on a 2 × n array. Thus, T (1) = 1.

When n = 2, we have a 2 × 2 array. Since the orientation of the array matters,

we have each of the following distinct tilings. Thus, T (2) = 2.

What about when n = 3? Again, we can enumerate these tilings by hand and
be sure that we aren’t missing any. We see that T (3) = 3.

Okay, one more case, when n = 4. We see that T (4) = 5.

2.4. TWO MORE (DIFFERENT) EXAMPLES 131

Can we start to find a pattern now? Writing out larger arrays will just
be tiresome! Let’s think about how we could have used the fact that T (1) =
1 to deduce something about T (2) . . . Well, wait a minute . . . We couldn’t,
right? There was something fundamentally different about those two cases.
Specifically, because dominoes are 2 × 1 in size, the fact that we only added one
row to the array didn’t help us.

Alright, let’s consider n = 3, then. Could we use the fact that T (2) = 2 at
all? In this case, yes! Knowing there were two tilings of the 2 × 2 array, we
could immediately build two tilings of the 2 × 3 array without much thought.
Specifically, we can just append a vertical domino to each of those previous
tilings. But we know now that T (3) = 3. Where did the third tiling come
from? Look at that tiling again and how it compares to the other two. In that
third tiling, the dominoes on the right side are horizontal, as opposed to the
vertical one in the other two tilings. If we remove those two parallel, horizontal
dominoes, we are left with precisely the situation when n = 1. Put another way,
we can build a tiling of a 2 × 3 array by appending a square of two horizontal
dominoes to the right side. In total, then, we have described all of the tilings of
a 2 × 3 board in terms of boards of smaller sizes, namely 2 × 2 and 2 × 1:

T (2) = 2

T (1) = 1

T (3) = 3 = 2 + 1 = T (2) + T (1)

Now you might see how the pattern develops! Let’s show you what happens
when n = 4, how we can construct all of the tillings that make up T (4) by
appending a vertical domino to each of the tilings that make up T (3), or by
appending two horizontal dominoes to each of the tilings that make up T (2):
132 CHAPTER 2. MATHEMATICAL INDUCTION

T (3) = 3

T (2) = 2

T (4) = 5 = 3 + 2 = T (3) + T (2)

Notice, as well, that no tiling for the 2 × 4 array was produced twice in this
way. (Think carefully about why this is true. How can we characterize the two
types of tilings in a way that will guarantee they don’t coincide at all?) With
this information, we can immediately conclude that T (4) = T (3) + T (2).
Furthermore, we can generalize this argument; nothing was special about
n = 4, right? For any particular n, we can just consider all possible tilings, and
look at what happens on the far right-hand side of the array: either we have
one vertical domino (which means the tiling came from a 2 × (n − 1) array) or
two horizontal dominoes (which means the tiling came from a 2 × (n − 2) array).
With confidence in this argument, we can conclude that
T (n) = T (n − 1) + T (n − 2)
for all of the values of n for which this expression makes sense. What values are
those? Remember that we had to identify T (1) and T (2) separately; this argu-
ment doesn’t apply to those values. Accordingly, we have to add the restriction
n ≥ 3 for the equation above to hold true.
With this information, we can then easily figure out T (n) for any value of
n, given enough time. We could write a computer program fairly easily, even.
It was this inductive argument, though—the pattern that we noticed and our
thorough description of why it occurs—that allowed us to make the conclusion
in the first place. In this case, too, it just so happens that the value of every
term, T (n), depends on the value of two previous terms, T (n − 1) and T (n − 2).
This did not happen in our previous examples in this chapter, and it hints at
something deeper going on here. Do you see how our previous definition of
induction, and the domino analogy, doesn’t exactly apply here anymore? How
might you try to amend our analogy to explain this kind of situation? Think
2.4. TWO MORE (DIFFERENT) EXAMPLES 133

about these issues for a bit and then read on. We’ll talk about them more
in-depth after the next example.
By the way, did you notice something interesting about our solution to this
example? Do you know any other sequences of numbers that behave similarly?
Think about it . . .

2.4.2 Winning Strategies

This example will be our first induction puzzle that doesn’t prove a numerical
formula! It might seem strange to think about that, but it’s true, as you’ll see.
This is actually more common in mathematics than you might think, too: a
problem or mathematical object might have some underlying inductive structure
without depending on something algebraic or arithmetic.
In fact, we will be discussing a game. It’s a game in the usual sense—there
are rules to be followed by two players and there is a clear winner and loser—but
it’s also a game in the mathematical sense, where we can formulate the rules and
playing situations using mathematical notation and discuss formal strategies in
an abstract way. We can even solve the game. This is very different than say,
the game of baseball.
Let’s discuss the rules for this game, which we shall call “Takeaway”, for
now. There are two players, called P1 and P2 . The player P1 goes first every
time. The players start with two piles of stones in the middle of a table, each pile
containing exactly n stones, where n is some natural number. (To distinguish
the different versions of the game, we will say the players are “playing Tn ” when
there are n stones per pile.) On each player’s move, they are allowed to remove
any number of stones from either pile. It is illegal, though, to remove stones
from both piles at once. The player who removes the final stone from the piles
is the winner.
Try playing Takeaway with some friends. Use pennies or candies or penny
candy as stones. Try it for different values of n. Try switching roles so you
are P1 and then P2 . Try to come up with a winning strategy, a method of
playing that maximizes your chances of winning. Try to make a conjecture for
what happens for different values of n. Who is “supposed” to win? Can you
prove your claim? Seriously, play around with this game and attempt to prove
something before reading on for our analysis thereof. You might be surprised
by what you can accomplish!
134 CHAPTER 2. MATHEMATICAL INDUCTION

As with the other examples, let’s use some small values of n to figure out
what’s really going on, then try to generalize. When n = 1, this game is rather
silly. P1 must empty one pile of its only stone, then P2 gets the only remaining
stone in the other pile. Thus, P2 wins. (Notice that it doesn’t matter which
of the two piles P1 picks from, P2 will always get the other one. We might say
that P1 picks the pile on the left “without loss of generality” because it doesn’t
matter either way; the situations are equivalent, so we might as well say it’s the
left pile to have something concrete to say. We will explore this idea of “without
loss of generality” later on when we discuss mathematical logic, too.)

P1’s turn P2’s turn P2 wins!

When n = 2, we now have a few cases that might appear. Think about
P1 ’s possible moves. Again, they might act on either the left or right pile, but
because they’re ultimately identical and we can switch the two piles, let’s just
say (without loss of generality) that P1 removes some stones from the left pile.
How many? It could be one or two stones. Let’s examine each case separately.

P2 ’s turn
P2 wins!

P1 ’s turn
P2 ’s turn
???

If P1 removes both stones, how should P2 react? Duh, they should take the
other pile, so P1 probably shouldn’t have made that move in the first place.
However, P1 might not be thinking straight or something and, besides, we need
to consider all possible situations here to fully analyze this game. Thus, in this
case (the top line in the above diagram) P2 wins. Okay, that’s the easy situation.
What if P1 removes just one stone from the left pile (the bottom line above)?
How should P2 react? We now have some options:

• If P2 removes the other stone from the left pile . . . well, P1 takes the other
pile and P1 wins.

• If P2 removes both stones from the right pile . . . well, P1 takes the last
stone from the left pile and P1 wins.
2.4. TWO MORE (DIFFERENT) EXAMPLES 135

• However, if P2 removes just one stone from the right pile . . .

P1’s turn P2’s turn

Now we have exactly the same situation presented by T1 , which we already

analyzed! It is, again, P1 ’s move first, so we know what will happen: P2 wins
no matter what. If you are player P2 , this is obviously the best move: you win
no matter how P1 responds!
Stepping back for a second, let’s think about what this has shown: no matter
what P1 does first (remove one or two stones from either pile), there is some
possible response that P2 can make that will guarantee a win for P2 , regardless
of P1 ’s subsequent response. Wow, P2 is sitting pretty! Let’s see if this happens
for other values of n.
When n = 3, we will again assume (without loss of generality) that player
P1 acted on the left pile. They could remove one, two, or three stones:

• If P1 removes all three, P2 responds by taking the other pile completely

and wins.

• If P2 removes two stones . . . well, what should player P2 do?

Finishing off that left pile is stupid (because P1 can take the whole right pile
and win), and pulling the entire right pile is similarly stupid (because P1 can
take the whole left pile and win), so something in between is required. Now,
if P2 removes just one stone from the right pile, notice that P1 can respond
with the same action; this leaves exactly one stone in both piles, but the roles
reversed. With P2 going first in such a situation, they are now bound to lose,
per our previous analysis. Bad move, P2 !

P1 wins!

P1’s turn P2’s turn P1’s turn

Let’s try again. If P2 removes two stones from the right pile instead . . . look
at that! We now have exactly one stone in each pile, with P1 up first, so we
know P1 is going to lose. P2 strikes again!
136 CHAPTER 2. MATHEMATICAL INDUCTION

P2 wins!

P1’s turn P2’s turn

Think about the case where n = 4 for a minute, and you’ll find the exact
same analysis occurring. You’ll another possibility to consider: player P1 can
remove one, or two, or three, or four stones from the left pile. Whatever they
do, though, you’ll find that P2 can just mimic that action on the other pile,
reducing the whole game to a previous, smaller version of the game, where P2
was shown to be guaranteed a win! It looks like P2 is in the driver’s seat the
whole time, since they can respond to whatever P1 does, making an identical
move on the other pile. No matter what P1 does, there is always a response for
P2 that means they win, regardless of P1 ’s subsequent moves. In this sense, we
say “P2 has a winning strategy”. There is a clear and describable method for
P2 to assess the game situation and choose a specific move to guarantee a win.
How might we prove this? How does this even fit into this chapter on
induction? It might be hard to see, at the moment. What are we really even
proving here? What are the dominoes or rungs in our analogy for this problem?
In wrapping your brain around this example, you should hopefully realize the
following: induction is not about algebraic formulas all the time; induction
represents some kind of “building-up” structure, where larger situations depend
on smaller ones; we have to prove some initial fact, and then argue how an
arbitrary, larger fact can be reduced so that it depends on a previous fact. This
is really what the dominoes analogy is meant to accomplish. It just so happens
that this analogy is particularly illustrative for certain induction problems (but
not all) and is visualizable and memorable. It does not perfectly apply to all
situations, though.
Read back through these four examples from this chapter and think about
how they are similar and how they are different. Try to come up with a more
precise, mathematical description of mathematical induction using some better
terminology, perhaps of your own invention. (By this, we mean something
better than our intuitive analogy. You’d be surprised at how well you might be
able to describe induction without really knowing what you “ought” to say, and
you’ll actually learn a lot, in the process!) In due time, we will have a rigorous
statement to make, and prove, about mathematical induction and its various
forms. In the meantime, we need to take a trip through some other areas of
mathematics to build up the necessary language, notation, and knowledge to
come back and tackle this problem. Before we go, though, we should mention
a few useful applications of mathematical induction.
2.5. APPLICATIONS 137

2.4.3 Questions & Exercises

Remind Yourself
Answering the following questions briefly, either out loud or in writing. These
are all based on the section you just read, so if you can’t recall a specific defi-
nition or concept or example, go back and reread that part. Making sure you
can confidently answer these before moving on will help your understanding and
memory!
(1) How are these two examples inductive? In what ways are they similar to
the previous examples, with the cubes and lines? In what ways are they
different?
(2) With the domino tilings, how many previous values did we need to know to
compute T (n)?
(3) What is the difference between writing T (n) = T (n − 1) + T (n − 2) and
T (n + 2) = T (n + 1) + T (n)?
(4) What is the winning strategy in the Takeaway game? Try playing with a
friend who doesn’t know the game, and use that strategy as player P2 . How
frustrated do they get every time you win? Do they start to catch on?

Try It
Try answering the following short-answer questions. They require you to actu-
ally write something down, or describe something out loud (to a friend/class-
mate, perhaps). The goal is to get you to practice working with new concepts,
definitions, and notation. They are meant to be easy, though; making sure you
can work through them will help you!
(1) What is T (5)? Can you draw all of those tilings?
(2) Work through the possibilities for takeaway with two piles of 4 stones. Can
you make sure that player P2 always has a winning move?
(3) Challenge: What happens if you play Takeaway with three piles of equal
sizes? Can you find a winning strategy for either player? Try playing with
a friend and see what happens!
(4) Look up the Fibonacci numbers. How are they related to the sequence of
numbers T (n) we found in the domino tiling example?

2.5 Applications
2.5.1 Recursive Programming
The concepts behind mathematical induction are employed heavily in computer
Pn
science, as well. Think back to how we first derived the formula for k=1 k 2 .
138 CHAPTER 2. MATHEMATICAL INDUCTION

Once we had a way to represent a cubic number in terms of a smaller cube and
some leftover terms, we repeated this subsitution process over and over until
we arrived at the “simplest” case, namely, the one that we first observed when
starting the problem: 23 = 1+3+3+1. Recursive programming takes advantage
of this technique: to solve a “large” problem, identify how the problem depends
on “smaller” cases, and reduce the problem until one reaches a simple, known
case.
A classical example of this type of technique is seen in writing code to com-
pute the factorial function, n!, which is defined as the product of the first n
natural numbers:
n! = 1 · 2 · 3 · · · (n − 1) · n
This is a simple definition that we, as humans, intuitively understand, but
telling a computer how to perform this product doesn’t work quite the same
way. (Try it! How do you say “and just keep going until you reach n” in
computer code?) A more efficient way to program the function, and one that
models the mathematically inductive definition, in fact, is to have one program
recursively call itself until it reaches that “simple” case. With the factorial
function, that case is 1! = 1. For any other value of n, we can simply apply the
knowledge that
n! = (n − 1)! · n
over and over to compute n!. Consider the following pseudocode that represents
this idea:
factorial(n):

if n = 1
return 1
else
return n * factorial(n-1)
end

We know that 1! = 1, so if the program is asked to compute that, the correct

value is returned right away. For any larger value of n, the program refers to
itself and says, “Go back and compute (n−1)! for me, then I’ll add a factor of n
at the end, and we’ll know the answer.” To compute (n − 1)!, the program asks,
again, if the input is 1; if not, it calls itself and says, “Go back and compute
(n−2)! for me, then I’ll add a factor of n−1 at the end.” This process continues
until the program returns 1! = 1. From there, it knows how to find 2! = 1 × 2,
and then 3! = 2! × 3, and so on, until n! = (n − 1)! × n.
Another example involving recursive programming arises with the Fibonacci
numbers. You may have seen this sequence of numbers before in a mathematics
course. (In fact, we even mentioned them in the last section, with the domino
tilings!) You also might have heard about how they appear in nature in some in-
teresting and strange ways. (The sequence was first “discovered” by the Italian
mathematician Leonardo of Pisa while studying the growth of rabbit popula-
tions.) The first two numbers in the sequence are specified to be 1, and any
2.5. APPLICATIONS 139

number in the sequence is defined as the sum of the previous two. That is, if
we say F (n) represents the n-th Fibonacci number, then

F (1) = 1 and F (2) = 1 and F (n) = F (n − 1) + F (n − 2) for every n ≥ 3

Now, what is F (5)? Or F (100)? Or F (10000)? This can be handled quite easily
by a recursive program. The idea is the same: if the program refers to either one
of the “simple cases”, i.e. F (1) or F (2), then it will know to return the correct
value of 1 immediately. Otherwise, it will call itself to compute the previous two
numbers and then add those together. Look at the pseudocode below and think
about how it works. What would happen if we used this program to compute
F (10)? How would it figure out the answer?

Fibonacci(n):

if n = 1 or n = 2
return 1
else
return Fibonacci(n-1) + Fibonacci(n-2)
end

This follows the same idea as the factorial program above (let the program
call itself to compute values for “smaller” cases of the function until we reach
a known value) but there’s something a little deeper going on here. If we
were to input n = 10 into the program, it would recognize that it does not
know the output value yet, and it will call itself to compute Fibonacci(9) and
Fibonacci(8). In each of those calls to the program, it would again recognize
the value is as yet unknown. Thus, it would call upon itself again to compute
Fibonacci(8) and Fibonacci(7), but also Fibonacci(7) and Fibonacci(6).
That’s right, the program calls itself multiple times with the same input value.
To compute F (9), we need to know F (8) and F (7), but meanwhile, to compute
F (8), we also need to know F (7) and F (6). In this way, we end up calling the
program Fibonacci many times.
Try to compare the programs Fibonacci and factorial, especially in re-
gards to the inductive processes we have been investigating in this chapter. Do
they use similar ideas? How do they relate to the “domino” analogy of math-
ematical induction that we outlined? Think of the “fact” written on Domino
n as being the computation of the correct value of n! or F (n). How does the
analogy work in each case? Will all the dominos fall? Keep these questions in
mind as you read on. There is some very powerful mathematics underlying all
of these ideas.

2.5.2 The Tower of Hanoi

Let’s take a short break and play a game. Well, it’s not exactly a break because
this is, in a sense, an inductive game, so it’s completely relevant. But it is a
140 CHAPTER 2. MATHEMATICAL INDUCTION

game, nevertheless! The Tower of Hanoi is a very popular puzzle, partly be-
cause it involves such simple equipment and rules. Solving it is another matter,
though!
Imagine that we have three vertical rods and three disks of three different
sizes (colored blue, green, and red) stacked upon each other like so:

The goal is to move all three disks to another rod (either the middle or the right
one, it doesn’t matter) by following these rules:
1. A single move consists of moving one (and only one) disk from the top of
the stack on any rod and moving it to the top of the stack on another rod.
2. A disk cannot be placed on top of a smaller disk.
That’s it! Two simple rules, but a difficult game to play. Try modeling the
game with a few coins or playing cards or whatever you have handy. (You can
even buy Tower of Hanoi sets at some games stores.) Can you solve it? How
many moves did it take you? Is your solution the “best” one? Why or why not?
We mentioned that this is an inductive game, so let’s explore that idea
now. We want to consider how many moves it takes to solve the puzzle (where
one move accounts for moving one disk from one rod to another) and, more
specifically, identify the smallest possible number of moves it would take to
solve the puzzle. To solve the puzzle with three disks, we could keep moving
the smallest disk back and forth between two rods and generate 100 moves, if
we wanted to, and then solve it, but that’s certainly not the best way to do
it, right? Let’s say we found a way to solve the puzzle in a certain number of
moves; how could we show that the number of moves we used is the smallest
possible number of moves?
To address this question, we want to break down the method of solving the
puzzle recursively. In doing so, we are actually going to answer a far more
general question: What is the smallest number of moves required to solve the
Tower of Hanoi puzzle with n disks on 3 rods? We posed the puzzle above with
just 3 disks to give you a concrete version to think about and work with, but
we can answer this more general question by thinking carefully. To make sure
we are on the same page, we will show you how we solved the version with 3
disks:
2.5. APPLICATIONS 141

Start

Move 1

Move 2

Move 3

Move 4

Move 5

Move 6

Move 7
142 CHAPTER 2. MATHEMATICAL INDUCTION

Notice that the largest disk is essentially “irrelevant” for most of the solution.
Since we are allowed to place any other disk on top of it, all we need to do is
“uncover” that disk by moving the other disks onto a different rod, move the
largest disk to the only empty rod, then move the other disks on top of the large
one. In essence, we perform the same procedure (shifting the two smaller disks
from one rod to another) twice and, in between those, we move the large disk
from one rod to another. If the largest disk hadn’t been there at all, what we
actually did was solve the version of the puzzle with 2 disks, but twice! (Think
carefully about this and make sure you see why this is true. Follow along with
the moves in the diagrams above and pretend the large, blue disk isn’t there.)
This shows that the way to solve the 3-disk puzzle involves two iterations of
solving the 2-disk puzzle, with one extra move in between (moving the largest
disk). This indicates a recursive procedure to solve the puzzle, in general. To
optimally solve the n-disk puzzle, we would simply follow the procedure to
optimally solve the (n − 1)-disk puzzle, use one move to shift the largest, n-th
disk, then solve the (n − 1)-disk puzzle again.
Now that we have some insight into how to optimally solve the puzzle, let’s
identify how many moves that procedure requires. Recognizing that solving this
puzzle uses a recursive algorithm, we realize that proving anything about the
optimal solution will require induction. Accordingly, we would need to identify
a “starting point” for our line of dominos, and it should correspond to the
“smallest” or “simplest” version of the puzzle. For the Tower of Hanoi, this is
the 1-disk puzzle. Of course, this is hardly a “puzzle” because we can solve it in
one move, by simply shifting the only disk from one rod to any other rod. If we
let M (n) represent the number of moves required to optimally solve the n-disk
puzzle, then we’ve just identified M (1) = 1. To identify M (2), we can use our
observation from the previous paragraph and say that
M (2) = M (1) + 1
|{z} + M (1) =1+1+1=3
| {z } | {z } | {z }
solve 2-disk solve 1-disk shift largest disk solve 1-disk

and then it must be that

M (3) = M (2) + 1 + M (2) = 3 + 1 + 3 = 7
and
M (4) = M (3) + 1 + M (3) = 7 + 1 + 7 = 15
and so on. Do you notice a pattern yet? Each of these numbers is one less
than a power of 2, and specifically, we notice that M (n) = 2n − 1, for each of
the cases we have seen thus far. It’s important to point out that observing this
pattern doesn’t prove the pattern; just because it works for the first 4 cases
does not mean the trend will continue, but that’s exactly what an induction
proof would accomplish. Also, recognizing that pattern and “observing” that
M (n) = 2n − 1 is a non-trivial matter, itself. We happened to know the answer
and had no problem identifying the formula for you. You should probably try,
on your own, to “solve” the following relationship
M (n) = 2M (n − 1) + 1 and M (1) = 1
2.5. APPLICATIONS 143

and see if you can derive the formula M (n) = 2n − 1. The reason such a formula
is nicer than the above relationship is that, now, M (n) depends only on n, and
not on previous terms (like M (n−1), for example). This relationship and others
like it are known as recurrence relations, and they can be rather difficult to solve,
in general!
We know how to solve this one, though, and it yields M (n) = 2n − 1. We
will leave it to you to verify this. You can do so by checking a few values in the
equation above, but we all know that isn’t a proof. Try working through the
inductive steps to actually prove it! We have already done most of the work, but
it will be up to you to arrange everything carefully and clearly. Remember that
you should identify what the “fact” on each domino is, ensure that Domino 1
falls, and then make a general argument about Domino n toppling into Domino
(n + 1). Try to write that proof. Do the details make sense to you? Try showing
your proof to a friend and see if they understand it. Did you need to tell them
anything else or guide them through it? Think about the best way to explain
your method and steps so that the written version suffices and you don’t have
to add any verbal explanations.

2.5.3 Questions & Exercises

Remind Yourself
Answering the following questions briefly, either out loud or in writing. These
are all based on the section you just read, so if you can’t recall a specific defi-
nition or concept or example, go back and reread that part. Making sure you
can confidently answer these before moving on will help your understanding and
memory!

(1) How is a recursive program inductive?

(2) What is the inductive structure of the Tower of Hanoi? Where did we solve
the 2-disk puzzle while solving the 3-disk puzzle?

Try It
Try answering the following short-answer questions. They require you to actu-
ally write something down, or describe something out loud (to a friend/class-
mate, perhaps). The goal is to get you to practice working with new concepts,
definitions, and notation. They are meant to be easy, though; making sure you
can work through them will help you!

(1) Follow the steps of the psuedocode factorial to compute 5!.

(2) Follow the steps of the psuedocode Fibonacci to compute F (5).

(3) Solve the Tower of Hanoi puzzle with 4 disks. Make sure that you can do
it in the optimal number of moves, 24 − 1 = 15.
144 CHAPTER 2. MATHEMATICAL INDUCTION

2.6 Summary
We have now seen some examples of inductive arguments. We realized that
some of the puzzles we were solving used similar argument styles, and explored
several examples to get a flavor for the different issues that might come up in
such arguments. Specfically, we saw how inductive arguments are not always
about proving a summation formula or an equation: inductive arguments can
apply to any situation where a fact depends on a “previous instance” of that fact.
This led us into developing an analogy for how induction works, mathematically
speaking. We are comfortbal with thinking of induction in terms of the “Domino
Analogy” for now, but one of our main goals in moving forwards is rigorously
stating and proving a principle of induction. For now, let’s get lots of practice
working with these kinds of arguments. This is what this chapter’s exercises are
meant to achieve. Later on, once we’ve formalized induction, we’ll be better off
for it, and we’ll have a thorough understanding of the concept!

2.7 Chapter Exercises

Here are some problems to get you comfortable working with inductive-style
arguments. We aren’t looking for fully rigorous proofs here, just a good de-
scription of what is going on and a write-up of your steps. We’ll come back
to some of these later and rigorously prove them, once we’ve established the
Principle of Mathematical Induction (PMI) and a corresponding proof strategy.
Problem 2.7.1. Prove the following summation formula holds for every natural
number, and for n = 0, as well:
n
X
2i = 2n+1 − 1
i=0

Follow-up question: use this result to state how many games are required to
determine a winner in a single-elimination bracket tournament with 2n teams.
(For example, the NCAA March Madness Tournament uses this format, with
n = 6.)
Problem 2.7.2. Prove that 3n ≥ 2n+1 for every natural number n that is greater
than or equal to 2.
Problem 2.7.3. For which natural numbers n do the following inequalities hold
true? State a claim and then prove it.

1. 2n ≥ (n + 1)2

2. 2n ≥ n!

3. 3n+1 > n4

4. n3 + (n + 1)3 > (n + 2)3

2.7. CHAPTER EXERCISES 145

Problem 2.7.4. The December 31 Game: Two players take turns naming
dates from a calendar. On each turn, a player can increase the month or date
but not both. The starting position is January 1, and the winner is the person
who says December 31. Determine a winning strategy for the first player.
For example, a sequence of moves that yields Player 1 winning is as follows:
(1) January 10, (2) March 10, (1) August 10, (2) August 25, (1)
August 28, (2) November 28, (1) November 30, (2) December 30,
(1) December 31
By winning strategy we mean a method of play that Player 1 follows that guar-
antees a win, no matter what Player 2 does.
Problem 2.7.5. Find and prove a formula for the sum of a geometric series,
which is a series of the form
n−1
X
qi
i=0
for some real number q and some natural number n. (Hint: be careful when
q = 1.)
Problem 2.7.6. Write a sentence that depends on n such that the sentence is
true for all values of n from 1 to 99 (inclusive), but such that the sentence is
false when n = 100.
Problem 2.7.7. What is wrong with the following “spoof” of the claim that
an = 1 for every n?
“Spoof ”: Let a be a nonzero real number. Notice that a0 = 1. Also, notice
that we can inductively write
an 1
an+1 = an · a = an · n−1
=1· =1
a 1
“”

Problem 2.7.8. In a futuristic society, there are only two different denominations
of currency: a coin worth 3 Brendans, and a coin worth 8 Brendans. There is
also a nation-wide law that says shopkeepers can only charge prices that can be
paid in exact change using these two coins.

What are the legal costs that a shopkeeper could charge you for a cup of coffee?
Hint: Try a bunch of small values and see what happens.
Problem 2.7.9. Consider a chessboard of size 2n × 2n , for some arbitrary nat-
ural number n. Remove any square from the board. Is it possible to tile the
remaining squares with L-shaped triominoes?
If your answer is Yes, prove it.

If your answer is No, provide a counterexample argument. (That is, find an n

such that no possible way of tiling the board will work, and show why this is
the case.)
146 CHAPTER 2. MATHEMATICAL INDUCTION

Problem 2.7.10. Consider an n × n grid of squares. How many sub-squares, of

any size, exist within this grid? For example, when n = 2, the answer is 5: there
are 4 1 × 1 squares and 1 2 × 2 square. Find a formula for your answer and try
to prove it is correct.
Problem 2.7.11. Prove that, in a line of at least 2 people, if the 1st person is a
woman and the last person is a man, then somewhere in the line there is a man
standing immediately behind a woman.
Problem 2.7.12. Prove that n3 − n is a multiple of 3, for every natural number
n.
Problem 2.7.13. A binary n-tuple is an ordered string of 0s and 1s, with n
total numbers in the string. Provide an inductive argument to explain why there
are 2n possible binary n-tuples.
Problem 2.7.14. Recall that the Fibonacci Numbers are defined by setting
f0 = 0 and f1 = 1 and then, for every n ≥ 2, setting fn = fn−1 + fn−2 . This
produces the sequence 0, 1, 1, 2, 3, 5, 8, 13, 21, 34, . . .
You might not know that the Fibonacci Numbers also have a closed form;
that is, there is a specific formula that defines them, in addition to the usual
recursive definition given above. Here it is:
" √ !n √ !n #
1 1+ 5 1− 5
fn = √ −
5 2 2

Prove that this formula is correct for all values of n ≥ 0.

Problem 2.7.15. Again, considering the Fibonacci Numbers, fn , prove the fol-
lowing:
n
X
1. fi = fn+2 − 1
i=0
n
X
2. fi2 = fn · fn+1
i=0

3. fn−1 · fn+1 − fn2 = (−1)n

4. fm+n = fn · fn+1 + fm−1 · fn
5. fn2 + fn+1
2
= f2n+1
Problem 2.7.16. Try to provide an inductive argument that explains why every
natural number n ≥ 2 can be written as a product of prime numbers. Can you
also show that this product is unique? That is, can you also explain why there
is exactly one way to factor a natural number into primes?
Problem 2.7.17. Prove that
n
X
k · k! = 1 · 1! + 2 · 2! + 3 · 3! + · · · + n · n! = (n + 1)! − 1
k=1
2.7. CHAPTER EXERCISES 147

Problem 2.7.18. What is wrong with the following “spoof” that all pens have
the same color.
“Spoof ”: Consider a group of pens with size 1. Since there is only 1 pen, it
certainly has the same color as itself.
Assume that any group of n pens has only one color represented inside the
group. (Note: we explained why this assumption is valid for n = 1 already,
so we can make this assumption.) Take any group of n + 1 pens. Line them
up on a table and number them from 1 to n + 1, left to right. Look at the
first n of them, i.e. look at pens 1, 2, 3, . . . , n. This is a group of n pens so, by
assumption, there is only one color represented in the group. (We don’t know
what color that is yet.) Then, look at the last n of the pens; i.e. look at pens
2, 3, . . . , n + 1. This is also a group of n pens so, by assumption, there is only
one color represented in this group, too. Now, pen #2 happens to belong to
both of these groups. Thus, whatever color pen #2 is, that is also the color of
every pen in both groups. Thus, all n + 1 pens have the same color.
By induction, this shows that any group of pens, of any size, has only one
color represented. Looking at the finite collection of pens in the world, then, we
should only find one color. “”

Problem 2.7.19. ? This problem is extremely difficult to analyze, and is taken

from the blog of the famous mathematician Terence Tao (link here).
There is an island upon which a tribe resides. The tribe consists of 1000
people, with various eye colours. Yet, their religion forbids them to know their
own eye color, or even to discuss the topic; thus, each resident can (and does)
see the eye colors of all other residents, but has no way of discovering his or her
own (there are no reflective surfaces). If a tribesperson does discover his or her
own eye color, then their religion compels them to commit ritual suicide at noon
the following day in the village square for all to witness. All the tribespeople
are highly logical and devout, and they all know that each other is also highly
logical and devout (and they all know that they all know that each other is
highly logical and devout, and so forth).
(For the purposes of this logic puzzle, ”highly logical” means that any conclu-
sion that can logically deduced from the information and observations available
to an islander, will automatically be known to that islander.)
Of the 1000 islanders, it turns out that 100 of them have blue eyes and 900
of them have brown eyes, although the islanders are not initially aware of these
statistics (each of them can of course only see 999 of the 1000 tribespeople).
One day, a blue-eyed foreigner visits to the island and wins the complete
trust of the tribe.
One evening, he addresses the entire tribe to thank them for their hospitality.
However, not knowing the customs, the foreigner makes the mistake of men-
tioning eye color in his address, remarking how unusual it is to see another
blue-eyed person like myself in this region of the world.
What effect, if anything, does this faux pas have on the tribe?
148 CHAPTER 2. MATHEMATICAL INDUCTION

2.8 Lookahead
In this chapter, we have introduced you to the concept of mathematical in-
duction. We looked at a few examples of puzzles where an inductive process
guided our solution, and then we described how a proof by induction would fol-
low to rigorously verify that solution. With the mathematical techniques and
concepts we have at hand thus far, we had to rely on a non-technical analogy
to describe this process to you. Thinking of an infinite line of dominos with
“facts” written on them knocking into each other is a perfectly reasonable in-
terpretation of this process, but it fails to represent the full mathematical extent
of induction. In a way, it’s like having a friend describe to you how to swing
a golf club, even though you’ve never played golf before. Sure, they can pro-
vide you with some mental imagery of what a swing “feels like”, but without
getting out there and practicing yourself, how will you truly understand the
mechanics of the golf swing? How will you learn how to adapt your swing, or
tell the differences between using a driver and a five iron and a sand wedge? By
investigating the underlying mechanics and practicing with those concepts, we
hope to gain a better understanding of mathematical induction so that, in the
future, we can use it appropriately, identify situations where it would be useful,
and, eventually, learn how to adapt it to other situations. Of course, it will help
to have that domino analogy in mind to guide our intuition, but we should also
remember that it is not rigorous mathematics. It also doesn’t perfectly describe
the other examples we discussed, where a falling domino depended on not only
the one immediately behind it, but several others before it.
In the next chapter, we will explore some relevant concepts needed to rig-
orously state and prove mathematical induction as a proof technique. Specif-
ically, we will study some ideas of mathematical logic and investigate how to
break down complicated mathematical statements and theorems into their con-
stituent parts, and also how to build interesting and complex statements out
of basic building blocks. Along the way, we will introduce some new notation
and shorthand that will let us condense some of the wordy statements we make
into concise (and precise) mathematical language. With that in hand, we will
explore some more fundamental proof strategies, that we will then apply to ev-
erything else we do in this course, including the induction technique, itself! We
will also study some of the ideas of set theory, a branch of mathematics that
forms the foundation for all other branches. This will be extremely useful for
organizing our ideas in the future, but it will also help us define the natural
numbers in a rigorous manner. With some concepts and knowledge from these
two branches of mathematics under our collective belts, we will be able to build
mathematical induction on a solid foundation and continue to use it properly.