Principle of Mathematical Induction1

• Formal Setting.
Mathematical Induction is used to prove theorems of the form ∀n ∈ N : P (n) where P is some
predicate with the natural numbers as the domain of discourse. Formally, it is stated as follows
 
P (1) ∧ (∀k ∈ N : P (k) ⇒ P (k + 1)) ⇒ (∀n ∈ N : P (n)) (PMI)

In plain English, it asserts that to prove the statement “P (n) is true for all n ∈ N.”, it suffices to prove

– The Base Case:(often easy) Prove that P (1) is true; and

– The Inductive Case:(the meat!) For any natural number k, if P (k) is true, then prove that
P (k + 1) is true.

In the inductive case, the assumption that “P (k) is true” is called the Induction Hypothesis.

• Arithmetic Series
Pn
Theorem 1. For all positive integers n, i=1 i = n(n + 1)/2

The predicate P (n) takes the value true if ni=1 i = n(n + 1)/2 and false otherwise. Theorem 1
P
asserts that P (n) is true for all natural numbers.

Proof. To prove ∀n ∈ N : P (n), the principle of mathematical induction (or simple induction, hence-
forth) asks us to check/prove the following.
Base Case: Let us verify that P (1) is true. Indeed, 1i=1 i = 1 and 1(1+1)
P
2 = 1, and thus P (1) is
true.
Inductive Case: Fix any natural number k. The induction hypothesis is that P (k) is true. We need to
prove P (k + 1) is true.
P (k) is true implies
k
X k(k + 1)
i= (Induction Hypothesis)
2
i=1

To prove P (k + 1) is true, that is, we need to show

k+1
X (k + 1)(k + 2)
i= (Need to Show)
2
i=1
1
Lecture notes by Deeparnab Chakrabarty. Last modified : 18th Jan, 2023
These have not gone through scrutiny and may contain errors. If you find any, or have any other comments, please email me at
[email protected]. Highly appreciated!

1
 We establish this by noting that the LHS of (Need to Show) is
k+1 k  
X X k(k + 1) k (k + 1)(k + 2)
i= i + (k + 1) = + (k + 1) = (k + 1) · +1 =
2 2 2
i=1 i=1
where in the second inequality we have used the (Induction Hypothesis). Thus, we have established
(Need to Show), and thus ∀n ∈ N : P (n) follows from induction.

Pn i an+1 −1
Exercise: Using induction, prove i=0 a = a−1 for any integer a > 1 and non-negative
integer n.

• A Divisibility Fact. We now prove the following fact by induction.

Theorem 2. For all n ∈ N, 3 divides n3 − n.

Proof. Let P (n) be the predicate representing the truth value of the statement given in the theorem
for a fixed natural number n. We proceed to prove ∀n ∈ N : P (n) by induction.
Base Case: Let us verify P (1). We need to verify that 3 divides 13 − 1 = 0. Indeed, 3 times 0 is 0.
Inductive Case: Let us now assume for a fixed k ∈ N that P (k) is true. That is, 3 divides k 3 − k. We
need to show P (k + 1) is true, that is, 3 divides (k + 1)3 − (k + 1). To do so, we expand (k + 1)3 , to
get
(k + 1)3 − (k + 1) = (k 3 + 3k 2 + 3k + 1) − (k + 1) = (k 3 − k) + 3(k 2 + k)
3(k 2 + k) is divisible by 3, and by the induction hypothesis (that is, P (k) is true), k 3 − k is divisible
by 3. Therefore, (k + 1)3 − (k + 1) is divisible by 3. That is, P (k + 1) is true. By the principle of
mathematical induction, P (n) is true for all n ∈ N.

Exercise: Does 4 divide n4 − n for all non-negative integers n? Mimic the above proof.

• Another Divisibility Fact. We now prove the following fact by induction.

Theorem 3. For all n ∈ N, 7 divides 32n − 2n .

Proof. Let P (n) be the predicate representing the truth value of the statement given in the theorem
for a fixed natural number n. We proceed to prove ∀n ∈ N : P (n) by induction.
Base Case: Let us verify P (1). We need to verify that 7 divides 32 −21 = 7. Indeed it does. Therefore
P (1) is true.
Inductive Case: Let us now assume for a fixed k ∈ N that P (k) is true. That is, 7 divides 32k − 2k .
We need to show P (k + 1) is true, that is, 7 divides 32(k+1) − 2(k+1) . Indeed observe,
32(k+1) − 2(k+1) = 32 · 32k − 2 · 2k
= 9 · 32k − 2 · 2k
= 7 · 32k + 2 · 32k − 2 · 2k (1)
 
= 7 · 32k + 2 · 32k − 2k

2
 7 divides 32k − 2k , by the induction hypothesis. 7 clearly divides 7 · 32k . Therefore, 7 divides
32(k+1) − 2k+1 . That is, P (k + 1) is true. By the principle of mathematical induction, P (n) is true
for all n ∈ N.
This proof was (slightly) tricky. Line (6) is where the trick was where we split the 9 as 7 + 2. Why did
we do that? Well, we knew something about 32k −2k , but when we expanded out we got 9·32k −2·2k .
If the “coefficients” of 32k and 2k were same we would be done (but it isn’t), so we just went ahead
and did that. It seems like a happy accident, but that is by design: the person who devised this theorem
(in this case, me) probably worked backwards to come up with the statement.

Exercise: Can you come up with statements like above? Can you guess which number will
always divide 43n − 32n for all natural numbers n? After guessing, can you prove that guess
using induction.

Remark: Sometimes, the induction principle may look as follows: (a) The base case may involve
proving P (1), P (2), . . . , P (c) for some finite c, and (b) The inductive case may be possible only
for numbers k ≥ c. Note this is also perfectly OK to establish ∀n : P (n). We will see such an
example in class and problem sets.

• A Geometry Fact. We now look at an example of a proof by induction where the predicate P (n)
is itself a statement in predicate logic. The statement we want to prove is one from high-school
geometry which you may have seen before. Below, an n-gon is a polygon on the plane with n vertices
and edges.

Theorem 4. The angles of any convex n-gon for n ≥ 3, measure in radians, adds up to (n − 2)π.

Before we begin, we need to recall/define what a convex n-gon is. For our purposes, it would suffice
to say a n-gon is convex if joining any two vertices which are two-apart by a line segment splits the
vertices into two opposite sides of it. More precisely, if we number the vertices in counter-clockwise
order as p1 , p2 , . . . , pn , then if we join pi and pi+2 by a line segment, the vertex pi+1 is on one side,
and all the other vertices are on the other side of this line segment. By “side”, we mean that if we
extend the line segment to an infinite line, it partitions the plane into two sides (or halves). See
Figure 1 for an illustration. Clear? It’s a bit of jargon, but let’s move on to the point I want to make.

Proof. We want to proceed by induction. But let’s be clear what the predicate is. Remember, we want
the predicate P (n) to depend only on a natural number n. What should it be? It is this. P (n) is true
iff every convex polygon on n vertices has its angles adding up to (n − 2)π. That is, P (n) itself is a
statement of the form ∀ · · · . More formally, here it is
n−1
X
P (n) = ∀ convex polygon (p1 , . . . , pn ) : ∠pn p1 p2 + ∠pi−1 pi pi+1 + ∠pn−1 pn p1 = (n − 2)π
i=2
| {z }
sum of angles of the n-gon

We want to prove ∀n ∈ N, n ≥ 3 : P (n) is true.

3
 𝑝𝑛 𝑝𝑛−1

𝑝1 𝑝𝑛−2

𝑝2

𝑝3

𝑝4

Figure 1: A Convex n-gon. Note that joining pn−1 and p1 splits the n-gon into a triangles on one side and a
convex (n − 1)-gon on the other.

Base Case. We start with n = 3. The statement P (3) is true if every triangle (all triangles are convex)
has its angles sum up to π. This is a fact which is true in Euclidean geometry. We will not prove this
here, but we will just bring to attention that this statement P (3) is itself a statement about all triangles.

Inductive Case. We fix a k ≥ 3 and assume P (k) is true. That is, we assume
k−1
X
∀ convex polygon (p1 , . . . , pk ) : ∠pk p1 p2 + ∠pi−1 pi pi+1 + ∠pk−1 pk p1 = (k − 2)π (IH)
i=2

We now need to show P (k + 1) is true. That is, we want to show for every convex (k + 1)-gon, its
sum of angles is (k − 1)π. This is a “for all” kind of statement. And to this end, we fix an arbitrary
(k + 1)-gon (p1 , . . . , pk , pk+1 ) and just prove that this polygon’s angles sum to (k − 1)π. Then we
would be done by induction, and the remainder of the proof establishes this.
The main idea is this. Connect the line segment p1 pk . Since (p1 , . . . , pk+1 ) is convex, it splits this
polygon into two convex polygons: the triangle (pk , pk+1 , p1 ) and the convex polygon (p1 , . . . , pk ).
Furthermore, two original angles split up thus:

∠pk+1 p1 p2 = ∠pk+1 p1 pk + ∠pk p1 p2 (2)

and
∠pk−1 pk pk+1 = ∠pk−1 pk p1 + ∠p1 pk pk+1 (3)
Refer to Figure 1 again (where you should mentally replace n by k + 1).

4
Now, note that the sum of the angles of (p1 , . . . , pk+1 ), let’s call it A, is
k
X
A = ∠pk+1 p1 p2 + ∠pi−1 pi pi+1 + ∠pk pk+1 p1
i=2
k−1
!
X
= ∠pk+1 p1 p2 + ∠pi−1 pi pi+1 + ∠pk−1 pk pk+1 + ∠pk pk+1 p1
i=2
k−1
X  
= ∠pk+1 p1 p2 + ∠pi−1 pi pi+1 + ∠pk−1 pk p1 + ∠p1 pk pk+1 + ∠pk pk+1 p1 (4)
i=2
  k−1
X  
= ∠pk+1 p1 pk + ∠pk p1 p2 + ∠pi−1 pi pi+1 + ∠pk−1 pk p1 + ∠p1 pk pk+1 + ∠pk pk+1 p1
i=2
(5)

where we applied (2) to get (4) and (3) to get (5). Next, we rearrange to get
k−1
!
X  
A = ∠pk p1 p2 + ∠pi−1 pi pi+1 + ∠pk−1 pk p1 + pk+1 p1 pk + ∠p1 pk pk+1 + ∠pk pk+1 p1
i=2

Since (p1 , . . . , pk ) is some convex polygon on k-vertices, we get that the first parenthesized expression
is (k − 2)π by (IH). This is because (IH) holds for every convex polygon on k vertices. The second
parenthesized term above adds up to π because it is the sum of angles of the triangle (p1 , pk , pk+1 ).
Therefore, A = (k − 2)π + π = (k − 1)π, which is what we wanted to show.

5
Answers to Exercises.
Pn i an+1 −1
• Exercise: Using induction, prove i=0 a = a−1 for any integer a > 1 and non-negative
integer n.

n+1
Proof. Fix any real a > 1. Let P (n) be the predicate which takes the value true if ni=0 ai = a a−1−1 .
P
We need to prove ∀n ∈ N ∪ {0} : P (n). We proceed by inductions.
Base Case. We need to prove P (0) is true. That is, 0i=0 ai = a−1a−1
P
. Indeed, both LHS and RHS are
1.
k+1
Inductive Case. Fix k ≥ 0 and suppose P (k) is true. That is, ki=0 ai = a a−1−1 . We need to prove
P
P (k + 1) is true.
Now note,
k+1 k
ak+2 − ak+1 + ak+1 − 1



X
i k+1
X
i k+1 ak+1 − 1
a =a + a |{z}
= a + =
a − 1 |{z} a−1
i=0 i=0 P (k) algebra

ak+2 −1
And now we see that the RHS is a−1 , thereby establishing P (k + 1). And thus, we have proved
the statement by induction.

• Exercise: Does 4 divide n4 − n for all non-negative integers n? Mimic the above proof.

Actually, 4 does not divide all n4 − n. Rather than giving you a counterexample, let me actually take
you down a “proof”, which will fail and thus give us a counter example.
“Proof” Let P (n) be the predicate representing the truth value of the statement given in the theorem
for a fixed natural number n. We proceed to prove ∀n ∈ N : P (n) by induction.
Base Case: Let us verify P (1). We need to verify that 4 divides 14 − 1 = 0. Indeed, 4 times 0 is 0.
Inductive Case: Let us now assume for a fixed k ∈ N that P (k) is true. That is, 4 divides k 4 − k. We
need to show P (k + 1) is true, that is, 4 divides (k + 1)4 − (k + 1). To do so, we expand (k + 1)4 , to
get

(k + 1)4 − (k + 1) = (k 4 + 4k 3 + 6k 2 + 4k + 1) − (k + 1) = (k 4 − k) + 4(k 3 + k) + 6k 2

And now we see our problem. To assert 4 divides (k + 1)4 − (k + 1), we see that 4 must divide 6k 2 .
This is because 4 does divide k 4 − k (by induction hypothesis) and 4 divides 4(k 3 + k). But does 4
divide 6k 2 always? No! Not when k = 1. And so, it suggests for k = 1, P (k + 1) may not be true.
That is P (2) may not be true.
Indeed, 4 does not divide 24 − 2 = 14. Ta da!
• Exercise: Can you come up with statements like above? Can you guess which number will
always divide 43n − 32n for all natural numbers n? After guessing, can you prove that guess
using induction.

Did you guess? It’s 55 = 43 − 32 .

6
Proof. Let P (n) be the predicate representing the truth value of the statement given in the theorem
for a fixed natural number n. We proceed to prove ∀n ∈ N : P (n) by induction.
Base Case: Let us verify P (1). We need to verify that 55 divides 43 − 32 = 55. Indeed it does.
Therefore P (1) is true.
Inductive Case: Let us now assume for a fixed k ∈ N that P (k) is true. That is, 55 divides 43k − 32k .
We need to show P (k + 1) is true, that is, 55 divides 43(k+1) − 32(k+1) . Indeed observe,

43(k+1) − 32(k+1) = 43 · 43k − 32 · 32k

= 64 · 43k − 9 · 32k
= 64 · 43k − 64 · 32k + 64 · 32k − 9 · 32k (6)
 
= 64 · 43k − 32k + 55 · 32k (7)

55 divides 43k − 32k , by the induction hypothesis. 55 clearly divides 55 · 32k . Therefore, 55 divides
43(k+1) − 32(k+1) . That is, P (k + 1) is true. By the principle of mathematical induction, P (n) is true
for all n ∈ N.