CS1231S Chapter 8

Induction and recursion

8.1 Mathematical Induction

Principle 8.1.1 (Mathematical Induction (MI)). Let m ∈ Z. To prove that ∀n ∈ Z>m P (n)
is true, where each P (n) is a proposition, it suffices to:
(base step) show that P (m) is true; and


(induction step) show that ∀k ∈ Z>m P (k) ⇒ P (k + 1) is true.
Justification. The two steps ensure the following are true:

P (m) by the base step;

P (m) ⇒ P (m + 1) by the induction step with k = m;
P (m + 1) ⇒ P (m + 2) by the induction step with k = m + 1;
P (m + 2) ⇒ P (m + 3) by the induction step with k = m + 2;
..
.

We deduce that P (m), P (m + 1), P (m + 2), . . . are all true by a series of modus ponens.
Terminology 8.1.2. In the induction step, we assume we have k ∈ Z>m such that P (k) is
true, and then show P (k + 1) using this assumption. In this process, the assumption that
P (k) is true is called the induction hypothesis.
1
Example 8.1.3. 1 + 2 + · · · + n = 2 n(n + 1) for all n ∈ Z>1 .
1
Proof. 1. For each n ∈ Z>1 , let P (n) be the proposition “ 1 + 2 + · · · + n = 2 n(n + 1) ”.
2. (Base step) P (1) is true because 1 = 21 × 1 × (1 + 1).
3. (Induction step)
3.1. Let k ∈ Z>1 such that P (k) is true, i.e., such that
1
1 + 2 + ··· + k = k(k + 1).
2
3.2. Then 1 + 2 + · · · + k + (k + 1)
1
3.3. = k(k + 1) + (k + 1) by the induction hypothesis P (k);
2
k  k+2
3.4. = + 1 (k + 1) = (k + 1)
2 2
1
3.5. = (k + 1)((k + 1) + 1).
2
3.6. So P (k + 1) is true.

24
 Q
2k
2k+1

Figure 8.1: Covering a checkerboard with L-trominos

4. Hence ∀n ∈ Z>1 P (n) is true by MI.

Terminology 8.1.4. We call the proof above an induction on n because n is the active
variable in it.
Example 8.1.5. n! > 2n for all n ∈ Z>4 , where n! = n × (n − 1) × · · · × 1.
Proof. 1. For each n ∈ Z>4 , let P (n) be the proposition “ n! > 2n ”.
2. (Base step) P (4) is true because 4! = 24 > 16 = 24 .
3. (Induction step)
3.1. Let k ∈ Z>4 such that P (k) is true, i.e., such that

k! > 2k .

3.2. Then (k + 1)! = (k + 1) × k! by the definition of !;

3.3. > (k + 1) × 2k by the induction hypothesis P (k);
3.4. > 2 × 2k as k + 1 > 4 + 1 > 2;
3.5. = 2k+1 .
3.6. So P (k + 1) is true.
4. Hence ∀n ∈ Z>4 P (n) is true by MI.
Example 8.1.6. An L-tromino is the following L-shape formed by three squares of the
checkerboard:

For all n ∈ Z>1 , if one square is removed from a 2n × 2n checkerboard, then the remaining
squares can be covered by L-trominos.
Proof. 1. For each n ∈ Z>1 , let P (n) be the proposition
if one square is removed from a 2n × 2n checkerboard, then the remaining
squares can be covered by L-trominos.
2. (Base step) P (1) is true because such a board itself is an L-tromino.
3. (Induction step)
3.1. Let k ∈ Z>1 such that P (k) is true.
3.2. 3.2.1. Let B be a 2k+1 × 2k+1 checkerboard with one square removed.
3.2.2. Divide B into four 2k × 2k quadrants.
3.2.3. Let Q be the quadrant containing the removed square.
3.2.4. Remove one L-tromino from the centre of B in a way such that each quad-
rant other than Q has one square removed.
3.2.5. We are left with four 2k × 2k checkerboards, each with one square removed.
3.2.6. By the induction hypothesis, each quadrant can be covered by L-trominos.
3.2.7. Hence B can be covered by L-trominos.
3.3. This shows P (k + 1) is true.
4. Hence ∀n ∈ Z>1 P (n) is true by MI.
Example 8.1.7. All participants in this Zoom meeting have the same birthday.  8a

25
Proof. 1. For each n ∈ Z>1 , let P (n) be the proposition
if a Zoom meeting has exactly n participants, then all its participants have the
same birthday.
2. (Base step) P (1) is true because if a Zoom meeting has exactly 1 participant, then clearly
all its participants have the same birthday.
3. (Induction step)
3.1. Let k ∈ Z>1 such that P (k) is true.
3.2. 3.2.1. Suppose a Zoom meeting has exactly k + 1 participants.
3.2.2. Pick two different participants a, b in the meeting.
3.2.3. Ask a to leave the meeting.
3.2.4. Since there are k people left in the meeting, by the induction hypothesis,
all the remaining participants have the same birthday, including b.
3.2.5. Tell a to join the meeting again, and then ask b to leave the meeting.
3.2.6. Since there are k people left in the meeting, by the induction hypothesis,
all the remaining participants have the same birthday, including a.
3.2.7. The participants who stayed in the meeting throughout have the same birth-
day as both a and b.
3.2.8. So a and b have the same birthday.
3.3. This shows P (k + 1) is true.
4. Hence ∀n ∈ Z>1 P (n) is true by MI. 

8.2 Strong Mathematical Induction

Principle 8.2.1 (Strong Mathematical Induction (Strong MI)). To prove that ∀n ∈ Z>m P (n)
is true, where each P (n) is a proposition and m ∈ Z, it suffices to choose some ` ∈ Z>0 and:
(base step) show that P (m), P (m + 1), . . . , P (m + ` − 1) are true;
(induction step) show that


∀k ∈ Z>0 P (m) ∧ P (m + 1) ∧ · · · ∧ P (m + ` − 1 + k) ⇒ P (m + ` + k)

is true.
Justification. The two steps ensure the following are true:

P (m) ∧ P (m + 1) ∧ · · · ∧ P (m + ` − 1)
by the base step;
P (m) ∧ P (m + 1) ∧ · · · ∧ P (m + ` − 1) ⇒ P (m + `)
by the induction step with k = 0;
P (m) ∧ P (m + 1) ∧ · · · ∧ P (m + ` − 1) ∧ P (m + `) ⇒ P (m + ` + 1)
by the induction step with k = 1;
P (m) ∧ P (m + 1) ∧ · · · ∧ P (m + ` − 1) ∧ P (m + `) ∧ P (m + ` + 1) ⇒ P (m + ` + 2)
by the induction step with k = 2;
..
.

We deduce that P (m), P (m + 1), P (m + 2), P (m + 3), . . . are all true by a series of modus
ponens.
Definition 8.2.2. The Fibonacci sequence F0 , F1 , F2 , . . . is defined by setting

F0 = 0 and F1 = 1 and Fn+2 = Fn+1 + Fn

for each n ∈ Z>0 .

26
 0
1
2
3
4
5

Figure 8.2: Rabbits

Example 8.2.3. F2 = 1 + 0 = 1, F3 = 1 + 1 = 2, F4 = 2 + 1 = 3, F5 = 3 + 2 = 5, ....

Example 8.2.4. • Initially, there is one pair of newly born matched rabbits.
• Each newly born rabbit takes one month to mature.
• Each mature pair of matched rabbits produces one pair of matched rabbits per month.

Let rn denote the number of pairs of rabbits after n months. Then for every n ∈ Z>0 ,

r0 = 1 and r1 = 1 and rn+2 = rn+1 + rn ,

where the rn+1 comes from the rabbits already present after (n + 1) months, and the rn
comes from the rabbits born after (n + 1) months.
Observation 8.2.5. rn = Fn+1 for every n ∈ Z>0 .
Example 8.2.6. Fn+1 6 (7/4)n for every n ∈ Z>0 .
Proof. 1. For each n ∈ Z>0 , let P (n) be the proposition “ Fn+1 6 (7/4)n ”.
2. (Base step) P (0) and P (1) are true because

F0+1 = 1 6 1 = (7/4)0 and F1+1 = 1 + 0 = 1 6 7/4 = (7/4)1 .

3. (Induction step)
3.1. Let k ∈ Z>0 such that P (0), P (1), . . . , P (k + 1) are true.
3.2. Then F(k+2)+1 = Fk+3
3.3. = Fk+2 + Fk+1 by the definition of Fk+3 ;
3.4. 6 (7/4)k+1 + (7/4)k as P (k) and P (k + 1) are true;
3.5. = (7/4)k (7/4 + 1)
3.6. < (7/4)k (7/4)2 as 7/4 + 1 = 11/4 < 49/16 = (7/4)2 ;
k+2
3.7. = (7/4) .
3.8. So P (k + 2) is true.
4. Hence ∀n ∈ Z>0 P (n) is true by Strong MI.
Remark 8.2.7. Given the same P (n), Strong MI is more likely to succeed than usual MI,
but the proof may be more cumbersome when written.
Remark 8.2.8. When ` = 0 in Principle 8.2.1 (Strong MI), the base step is empty. Thus to
prove that ∀n ∈ Z>m P (n) is true, where each P (n) is a proposition and m ∈ Z, it suffices
to show only


∀k ∈ Z>0 P (m) ∧ P (m + 1) ∧ · · · ∧ P (m + k − 1) ⇒ P (m + k) .

(The conjunction of no formula is by convention always true.)

Example 8.2.9. (1) S = {x ∈ Z>0 : 0 < x < 5} has smallest element 1.

27
 _5
4 x
3
2 x/2
1
^0
S S0

Figure 8.3: A difference between Z>0 and Q>0

(2) S 0 = {x ∈ Q>0 : 0 < x < 5} has no smallest element because if x ∈ S 0 , then x/2 ∈ S 0
and x/2 < x.
Theorem 8.2.10 (Well-Ordering Principle). Every nonempty subset of Z>m , where m ∈ Z,
has a smallest element.
Proof. We prove this by Principle 8.2.1 (Strong MI) with ` = 0.
1. Let m ∈ Z and S ⊆ Z>m with no smallest element.
2. For each n ∈ Z>m , let P (n) be the proposition “ n 6∈ S ”.
3. (Induction step)
3.1. Let k ∈ Z>0 such that P (m), P (m + 1), . . . , P (m + k − 1) are true, i.e., that m, m +
1, . . . , m + k − 1 6∈ S.
3.2. 3.2.1. Suppose m + k ∈ S.
3.2.2. Then m + k is the smallest element of S by the induction hypothesis as
S ⊆ Z>m .
3.2.3. This contradicts our assumption that S has no smallest element on line 1.
3.3. So m + k 6∈ S.
3.4. Thus P (m + k) is true.
4. Hence ∀n ∈ Z>m P (n) is true by Strong MI.
5. This implies S = ∅ as S ⊆ Z>m .

8.3 Recursively defined sequences

Terminology 8.3.1. A sequence a0 , a1 , a2 , . . . is said to be recursively defined if the defini-
tion of an involves a0 , a1 , . . . , an−1 for all but finitely many n ∈ Z>0 .
Example 8.3.2. (1) Define 0!, 1!, 2!, . . . by setting, for each n ∈ Z>0 ,

0! = 1 and (n + 1)! = (n + 1) × n!.

Then 1! = 1 × 1 = 1, 2! = 2 × 1 = 2, 3! = 3 × 2 = 6, 4! = 4 × 6 = 24, ....

(2) The Fibonacci sequence F0 , F1 , F2 , . . . was defined in Definition 8.2.2 by setting, for
each n ∈ Z>0 ,
F0 = 0 and F1 = 1 and Fn+2 = Fn+1 + Fn .
Then F2 = 1 + 0 = 1, F3 = 1 + 1 = 2, F4 = 2 + 1 = 3, F5 = 3 + 2 = 5, ....
(3) Fix r ∈ [0, 4] and p0 ∈ [0, 1]. Define p1 , p2 , . . . by setting, for each n ∈ Z>0 ,

pn+1 = r(pn − pn 2 ).

If r = 3 and p0 = 1/2, then

 1  1 2  3  3  3 2  9
p1 = 3 − = , p2 = 3 − = , ....
2 2 4 4 4 16

28
 (4) Fix a0 ∈ Z+ . Define a1 , a2 , a3 , . . . by setting, for each n ∈ Z>0 ,
(
an /2, if an is even;
an+1 =
3an + 1, if an is odd.

If a0 = 1, then a1 = 3 × 1 + 1 = 4, a2 = 4/2 = 2, a3 = 2/2 = 1, . . ..

Exercise 8.3.3. Let a1 = 1 and an+1 = an + (n + 1) for all n ∈ Z>1 . Find a general formula  8b
for an in terms of n that does not involve a0 , a1 , . . . , an−1 .
Proposition 8.3.4. There is a unique sequence a0 , a1 , a2 , . . . satisfying, for each n ∈ Z>0 ,

a0 = 0 and a1 = 1 and an+2 = an+1 + an .

Proof (optional material). For the purpose of this proof, let us call a sequence b0 , b1 , . . . , bn−1
a partial sequence if for all i ∈ Z>0 with i < n,

0,
 if i = 0;
bi = 1, if i = 1;

bi−1 + bi−2 , if i > 2.


1. First, we claim that there is a partial sequence of length n for every n ∈ Z>0 .
1.1. For each n ∈ Z>0 , let P (n) be the proposition

“there is a partial sequence of length n”.

1.2. (Base step) P (0) is true because the empty sequence is trivially a partial sequence
of length 0.
1.3. (Induction step)
1.3.1. Let k ∈ Z>0 such that P (k) is true.
1.3.2. This gives a partial sequence b0 , b1 , . . . , bk−1 of length k.
1.3.3. Define 
0,
 if k = 0;
bk = 1, if k = 1;

bk−1 + bk−2 , if k > 2.


1.3.4. Then b0 , b1 , . . . , bk is a partial sequence of length k + 1 by the choice of bk

and because b0 , b1 , . . . , bk−1 is a partial sequence.
1.3.5. So P (k + 1) is true.
1.4. Hence ∀n ∈ Z>0 P (n) is true by MI.
2. If b0 , b1 , . . . , bm−1 and c0 , c1 , . . . , cn−1 are partial sequences with m 6 n, then

b0 = 0 = c 0 ,
b1 = 1 = c 1 ,
b2 = b1 + b0 = c1 + c0 = c2 ,
b3 = b2 + b1 = c2 + c1 = c3 ,
..
.
bm−1 = bm−2 + bm−3 = cm−2 + cm−3 = cm−1 .

3. For each n ∈ Z>0 , define an to be the nth element of any partial sequence of length at
least n.
4. Then the sequence a0 , a1 , a2 , . . . is well defined by lines 1 and 2.
5. This sequence a0 , a1 , a2 , . . . is what we want because it agrees with all the partial se-
quences, and the conditions in the definition of partial sequences match with the required
conditions.

29
6. Let b0 , b1 , b2 , . . . be a sequence satisfying, for each n ∈ Z>0 ,

b0 = 0 and b1 = 1 and bn+2 = bn+1 + bn .

7. We show that an = bn for all n ∈ Z>0 .

7.1. Let n ∈ Z>0 .
7.2. Note that a0 , a1 , . . . , an and b0 , b1 , . . . , bn are partial sequences.
7.3. So an = bn by line 2.

8.4 Recursively defined sets

Theorem 8.4.1. Z>0 is the unique set with the following properties.
(1) 0 ∈ Z>0 . (base clause)
(2) If x ∈ Z>0 , then x + 1 ∈ Z>0 . (recursion clause)
(3) Membership for Z>0 can always be demonstrated by (finitely many) successive appli-
cations of the clauses above. (minimality clause)
Example 8.4.2. 0 ∈ Z>0 by (1).
∴ 1 ∈ Z>0 by (2) and the previous line.
∴ 2 ∈ Z>0 by (2) and the previous line.

Remark 8.4.3. (1) and (2) are true when Z>0 is changed to Q, but (3) is not. So (1) and
(2) are not enough to uniquely determine Z>0 .
Terminology 8.4.4. Theorem 8.4.1 gives a recursive definition of Z>0 .
Rough idea 8.4.5. A recursive definition of a set S consists of three types of clauses.
(base clause) Specify that certain elements, called founders, are in S: if c is a founder,
then c ∈ S.
(recursion clause) Specify certain functions, called constructors, under which the set S is
closed: if f is a constructor and x ∈ S, then f (x) ∈ S.
(minimality clause) Membership for S can always be demonstrated by (finitely many)
successive applications of the clauses above.
In words, the members of S are precisely those objects that can be obtained from the founders
by successively applying the constructors.
Rough idea 8.4.6 (structural induction). Let S be a recursively defined set. To prove that
∀x ∈ S P (x) is true, where each P (x) is a proposition, it suffices to:
(base step) show that P (c) is true for every founder c;


(induction step) show that ∀x ∈ S P (x) ⇒ P (f (x)) is true for every constructor f .
In words, if all the founders satisfy a property P , and P is preserved by all constructors, then
all elements of S satisfy P .
Example 8.4.7. The set 2Z of all even integers can be defined recursively as follows.
(1) 0 ∈ S. (base clause)
(2) If x ∈ S, then x − 2 ∈ 2Z and x + 2 ∈ 2Z. (recursion clause)
(3) Membership for 2Z can always be demonstrated by (finitely many) successive applica-
tions of clauses above. (minimality clause)

30
Theorem 8.4.8 (Structural induction over 2Z). To prove that ∀n ∈ 2Z P (n) is true, where
each P (n) is a proposition, it suffices to:
(base step) show that P (0) is true; and


(induction step) show that ∀x ∈ 2Z P (x) ⇒ P (x − 2) ∧ P (x + 2) is true.

Question 8.4.9. Define a set S recursively as follows.  8c

(1) 1 ∈ S. (base clause)
(2) If x ∈ S, then 2x ∈ S and 3x ∈ S. (recursion clause)

(3) Membership for S can always be demonstrated by (finitely many) successive applica-
tions of clauses above. (minimality clause)
Which of the numbers 9, 10, 11, 12, 13 are in S? Which are not?

31