An Introduction to Induction

Principles
Patricia H ILL
School of Computing
University of Leeds, United Kingdom

Roberto B AGNARA
Department of Mathematics
University of Parma, Italy

Copyright c 2004–2013 Patricia Hill. Distributed under the terms of the GNU Free Documentation License 1.2 1
 M ATHEMATICAL I NDUCTION
Let P (n) be a property of natural numbers n = 0, 1, 2, . . . The
principle of mathematical induction says that to show P (n) holds for
all natural numbers, it is sufficient to show:

• The base case: P (0) holds.

• The step case: If P (m) holds then so does P (m + 1), for any
natural number m.

P (m) is called inductive hypothesis.

Formally,


P (0) ∧ ∀m ∈ N : P (m) =⇒ P (m + 1) =⇒ ∀n ∈ N : P (n).
| {z }
ind. hyp.

M ATHEMATICAL I NDUCTION 2
 M ATHEMATICAL I NDUCTION : E XAMPLE 1


Let us show that P (n) ≡ 0 + 2 + 4 + · · · + 2n = n(n + 1) :
Pm

• Inductive hypothesis: P (m) ≡ i=0 2i = m(m + 1) .


• Base case, m = 0: P (0) ≡ 0 = 0(0 + 1) .

• Step case, m ≥ 0: Assume P (m) holds.

m+1
X
2i = 0 + 2 + · · · + 2m + 2(m + 1)
i=0
= (0 + 2 + . . . + 2m) + 2(m + 1) [by rearranging]
= m(m + 1) + 2(m + 1) [by ind. hyp. P (m)]
= (m + 1)(m + 2) [by rearranging]

so that P (m + 1) holds.

M ATHEMATICAL I NDUCTION : E XAMPLE 1 3

 E XERCISES
Prove the following identities by mathematical induction:

1 + 3 + 5 + · · · + (2n + 1) = (n + 1)2 ;

1 1 1 1 n+1
+ + + ···+ = ;
1·2 2·3 3·4 (n + 1)(n + 2) n+2

2 1 − r n+1
n
1 + r + r + ···+ r = .
1−r

E XERCISES 4
 M ATHEMATICAL I NDUCTION : E XAMPLE 2
Let Q(n) be the property that, for all symbols a, b, s1 , . . . , sn , we have
(as1 · · · sn = s1 · · · sn b) =⇒ a = b. We show that ∀n ∈ N : Q(n).
• Inductive hypothesis:


Q(m) ≡ (as1 · · · sm = s1 · · · sm b) =⇒ a = b .
• Base case, m = 0: Q(0) ≡ (a = b =⇒ a = b).
• Step case, m ≥ 0: Suppose

as1 · · · sm sm+1 = s1 · · · sm sm+1 b;

then b = sm+1 and therefore

as1 · · · sm = s1 · · · sm sm+1 = s1 · · · sm b.

By the inductive hypothesis, a = b, and thus Q(m + 1) holds.

M ATHEMATICAL I NDUCTION : E XAMPLE 2 5

 C OMPLETE M ATHEMATICAL I NDUCTION
Let P (n) be a property of natural numbers n = 0, 1, 2, . . . The
principle of complete mathematical induction says that to show P (n)
holds for all natural numbers, it is sufficient to show:
• The base case: P (0) holds.
• The step case: If P (k) holds for each k = 0, 1, . . . m then so
does P (m + 1), for any natural number m.
Vm
The conjunction k=0 P (k) is called inductive hypothesis.
Formally,
 ^
m  !
P (0)∧ ∀m ∈ N : P (k) =⇒ P (m+1) =⇒ ∀n ∈ N : P (n).
k=0
| {z }
ind. hyp.

C OMPLETE M ATHEMATICAL I NDUCTION 6

 S TRUCTURAL I NDUCTION
The principle of structural induction is:

In order to show that a property is true of all expressions, it

suffices to show it is true of all atomic expressions and is
preserved by all methods of forming the expressions.

S TRUCTURAL I NDUCTION 7
 S TRUCTURAL I NDUCTION : E XERCISE
Definition:

1. “tt” and “ff” are boolean expressions;

2. If b is a boolean expression, “(not b)” is a boolean expression;

3. If b0 and b1 are boolean expressions, “(b0 and b1 )” is a boolean

expression;

4. If b0 and b1 are boolean expressions, “(b0 or b1 )” is a boolean

expression.

5. Nothing is a boolean expression unless it is constructed by rules

1–4.

Prove using structural induction that in every boolean expression the

number of ‘(’ equals the number of ‘)’.

S TRUCTURAL I NDUCTION : E XERCISE 8

 W ELL -F OUNDED I NDUCTION : I NTRODUCTION
Mathematical and structural induction are instances of well-founded
induction.

• Mathematical induction relies on the fact that for any number n

every descending sequence n > n − 1 > n − 2 > · · · is finite.

• Structural induction relies on the fact that for any expression a0

every subexpression sequence of the form

a0 ⊐ a1 ⊐ a2 ⊐ · · ·

is finite. (a′ ⊏ a denotes a′ 6= a is a subexpression of a.)

These rely on a well-founded relation.

W ELL -F OUNDED I NDUCTION : I NTRODUCTION 9

 W ELL -F OUNDED R ELATIONS
A well-founded relation is a binary relation ‘≺’ on a set A such that
there are no infinite “descending chains”

a0 ≻ a1 ≻ · · · ≻ ai ≻ · · ·

• When a ≺ b, element a is called a predecessor of b.

• An equivalent definition of well-founded relation is:

Every non-empty subset S of A has an element with no
predecessors in S:

∀S ⊆ A : (S 6= ∅ =⇒ ∃m ∈ S . ∀s ∈ S : s ⊀ m)

W ELL -F OUNDED R ELATIONS 10

 W ELL -F OUNDED R ELATIONS ( CONT.)
• If ‘≺’ is well-founded, then the transitive closure ‘≺+ ’ is
well-founded.

• We write ‘’ for the reflexive closure of ‘≺’.

a  b ⇐⇒ a = b ∨ a ≺ b

Exercise: Show that ‘’ is not a well-founded relation.

W ELL -F OUNDED R ELATIONS ( CONT.) 11

 PARTIAL O RDERING
Suppose R is a binary relation on a set A. R is:
1. Reflexive: if, for all a ∈ A, a R a;
2. Transitive: if, for all a1 , a2 , a3 ∈ A :

(a1 R a2 ∧ a2 R a3 ) =⇒ a1 R a3 ;

3. Antisymmetric: if, for all a1 , a2 ∈ A :

(a1 R a2 ∧ a2 R a1 ) =⇒ a1 = a2 .

A partial ordering is a binary relation that is reflexive, transitive and

antisymmetric.
A total ordering on a set A is a partial ordering  on A such that, for
all a1 , a2 ∈ A, either a1  a2 or a2  a1 .

PARTIAL O RDERING 12
 W ELL -F OUNDED R ELATIONS AND PARTIAL O RDERINGS
Exercise: Let ‘≺’ a well-founded relation over a set A.

• Show that ‘≺+ ’, the transitive closure of ‘≺’, is well-founded.

• Show that ‘≺⋆ ’, the reflexive and transitive closure of ‘≺’, is a

partial ordering.

W ELL -F OUNDED R ELATIONS AND PARTIAL O RDERINGS 13

 W ELL -F OUNDED R ELATIONS : E XAMPLES
• Let A = N be the set of natural numbers and ‘≺’ be the relation
m ≺ n if and only if n = m + 1.
• Let A = N be the set of natural numbers and ‘≺’ be the relation
‘<’.
• Let A = Aexp be the set of all arithmetic expressions in IMP. Let
b ⊏ a be defined to hold if and only if b is an immediate
subexpression of a.
Then ‘⊏’ is a well-founded relation.
• Let A be the set of all finite character strings.
Let t ⋖ s be defined to hold if and only if t is a proper substring of
s.
Then ‘⋖’ is a well-founded relation.

W ELL -F OUNDED R ELATIONS : E XAMPLES 14

 W ELL -F OUNDED I NDUCTION
The principle of well-founded induction is:

Let ≺ be a well-founded relation on a set A. Let P be a

property. Then P (a) holds for all a in A if and only if


∀a ∈ A : ∀b ≺ a : P (b) =⇒ P (a) .

So, to prove that a property holds of all elements of a well-founded

set (i.e., a set with a well-founded ordering) we just have to prove
that if it holds for all predecessors of any a ∈ A in this ordering, then
it holds for a.

W ELL -F OUNDED I NDUCTION 15

 D EFINITIONS BY I NDUCTION
Consider the following definition: for all a ∈ Aexp we define:



 1, if a ≡ m;


 if a ≡ x;
1,


def
length(a) = length(a0 ) + length(a1 ), if a ≡ (a0 + a1 );





 length(a0 ) + length(a1 ), if a ≡ (a0 − a1 );


length(a ) + length(a ), if a ≡ (a ∗ a ).
0 1 0 1

Definitions of this form are often called inductive or recursive.

D EFINITIONS BY I NDUCTION 16