Books Number Theory

Mathematical Induction

Mathematical Induction

The template for a proof by mathematical induction:

  1. Define a statement 𝑃(𝑛)𝑃(𝑛) that depends on the positive integer 𝑛𝑛.
  2. Base case. Prove 𝑃(1)𝑃(1).
  3. Inductive step. Assuming 𝑃(𝑘)𝑃(𝑘) for some 𝑘≥1𝑘\geq 1, prove 𝑃(𝑘+1)𝑃(𝑘+1).
Theorem 1.3The principle of mathematical induction

A set of positive integers that contains the integer 11, and that has the property that, if it contains 𝑘𝑘, then it also contains 𝑘+1𝑘+1, must be set of all the positive integers.

In symbols: Let 𝑆⊆Z+𝑆\subseteq {\mathbb{Z}}^{+} with 1∈𝑆1\in 𝑆. If 𝑆𝑆 has the property that

𝑘∈Z+,𝑘∈𝑆⟹𝑘+1∈𝑆,𝑘\in {\mathbb{Z}}^{+},\quad 𝑘\in 𝑆\quad ⟹\quad 𝑘+1\in 𝑆,

then 𝑆=Z+𝑆={\mathbb{Z}}^{+}.

Back to the template. The theory behind this proof technique is that we define

𝑆={𝑛∈Z+:𝑃(𝑛)  holds}𝑆=\{𝑛\in {\mathbb{Z}}^{+}:𝑃(𝑛)\, \text{ holds}\}

and use the principle to conclude that 𝑆=Z+𝑆={\mathbb{Z}}^{+}. Equivalently, 𝑃(𝑛)𝑃(𝑛) holds for all 𝑛∈Z+𝑛\in {\mathbb{Z}}^{+}.

The proof will mainly make use of the well-ordering principle.

Proof

We prove by contradiction. Assume 𝑆≠Z+𝑆\neq {\mathbb{Z}}^{+}. Equivalently, the complement Z+∖𝑆{\mathbb{Z}}^{+}∖𝑆 is nonempty. Since Z+∖𝑆⊆Z+{\mathbb{Z}}^{+}∖𝑆\subseteq {\mathbb{Z}}^{+} and Z+∖𝑆≠∅{\mathbb{Z}}^{+}∖𝑆\neq \emptyset , the well-ordering principle Axiom 1.1 gives a least element 𝑛∈Z+∖𝑆𝑛\in {\mathbb{Z}}^{+}∖𝑆.

Then 𝑛>1𝑛>1, because 1∈𝑆1\in 𝑆, and hence 𝑛−1∈Z+𝑛-1\in {\mathbb{Z}}^{+}. The property of 𝑆𝑆 forces 𝑛−1∉𝑆𝑛-1\notin 𝑆: otherwise 𝑛∈𝑆𝑛\in 𝑆, which is absurd. So 𝑛−1∈Z+∖𝑆𝑛-1\in {\mathbb{Z}}^{+}∖𝑆, contradicting that 𝑛𝑛 is the least element of Z+∖𝑆{\mathbb{Z}}^{+}∖𝑆. Therefore the assumption is false, and 𝑆=Z+𝑆={\mathbb{Z}}^{+}.

Exercise 1.1

Use mathematical induction to prove

∑𝑗=1𝑛𝑗2=𝑛(𝑛+1)(2𝑛+1)6\sum _{𝑗=1}^{𝑛}{𝑗}^{2}=\frac{𝑛(𝑛+1)(2𝑛+1)}{6}
Proof

Let 𝑃(𝑛)𝑃(𝑛) be the statement that the equation holds for 𝑛𝑛.

Base case. When 𝑛=1𝑛=1, the LHS is 12=1{1}^{2}=1 and the RHS is 1⋅(1+1)⋅(2⋅1+1)/6=11\cdot (1+1)\cdot (2\cdot 1+1)/6=1. Thus, 𝑃(1)𝑃(1) holds.

Inductive step. Assume 𝑃(𝑘)𝑃(𝑘). Consider 𝑛=𝑘+1𝑛=𝑘+1. On the LHS, we have

∑𝑗=1𝑘+1𝑗2=∑𝑗=1𝑘𝑗2+(𝑘+1)2Apply  𝑃(𝑘)  to ∑𝑗=1𝑘𝑗2=𝑘(𝑘+1)(2𝑘+1)6+(𝑘+1)2=𝑘+16⋅[𝑘(2𝑘+1)+(𝑘+1)]=𝑘+16⋅(2𝑘2+7𝑘+6)=𝑘+16⋅(𝑘+2)(2𝑘+3)=(𝑘+1)(𝑘+2)(2(𝑘+1)+1)6.\begin{aligned}\sum _{𝑗=1}^{𝑘+1}{𝑗}^{2} & = & & \sum _{𝑗=1}^{𝑘}{𝑗}^{2}+{(𝑘+1)}^{2} \\ & & & \text{Apply }\, 𝑃(𝑘)\, \text{ to }\sum _{𝑗=1}^{𝑘}{𝑗}^{2} \\ & = & & \frac{𝑘(𝑘+1)(2𝑘+1)}{6}+{(𝑘+1)}^{2} \\ & = & & \frac{𝑘+1}{6}\cdot [𝑘(2𝑘+1)+(𝑘+1)] \\ & = & & \frac{𝑘+1}{6}\cdot (2{𝑘}^{2}+7𝑘+6) \\ & = & & \frac{𝑘+1}{6}\cdot (𝑘+2)(2𝑘+3) \\ & = & & \frac{(𝑘+1)(𝑘+2)(2(𝑘+1)+1)}{6}.\end{aligned}

Note that the RHS is exactly the expression of the RHS of the equation to prove when 𝑛=𝑘+1𝑛=𝑘+1. Therefore, we have shown that 𝑃(𝑘+1)𝑃(𝑘+1) also holds.

By mathematical induction, we claim that 𝑃(𝑛)𝑃(𝑛) holds for all 𝑛∈Z+𝑛\in {\mathbb{Z}}^{+}.

There is a variant we call the second principle of mathematical induction. It is often useful when we feel that 𝑃(𝑘)𝑃(𝑘) alone is not enough, and we need 𝑃(1),…,𝑃(𝑘)𝑃(1),\ldots ,𝑃(𝑘) to prove 𝑃(𝑘+1)𝑃(𝑘+1). The usual template is then:

  1. Define a statement 𝑃(𝑛)𝑃(𝑛) that depends on the positive integer 𝑛𝑛.
  2. Base case. Prove 𝑃(1)𝑃(1).
  3. Inductive step. Assuming 𝑃(1),…,𝑃(𝑘)𝑃(1),\ldots ,𝑃(𝑘) all hold for some 𝑘≥1𝑘\geq 1, prove 𝑃(𝑘+1)𝑃(𝑘+1).

Of course, we can always reduce the second template to the first by altering the statement itself, but that is a more roundabout way to think.

Theorem 1.4The second principle of mathematical induction

A set of positive integers that contains the integer 11, and for every positive integer 𝑘𝑘, if it contains 1,…,𝑘1,\ldots ,𝑘, then it also contains 𝑘+1𝑘+1, must be set of all the positive integers.

In symbols: Let 𝑆⊆Z+𝑆\subseteq {\mathbb{Z}}^{+} with 1∈𝑆1\in 𝑆. If 𝑆𝑆 has the property that

𝑘∈Z+,{1,…,𝑘}⊆𝑆⟹𝑘+1∈𝑆,𝑘\in {\mathbb{Z}}^{+},\quad \{1,\ldots ,𝑘\}\subseteq 𝑆\quad ⟹\quad 𝑘+1\in 𝑆,

then 𝑆=Z+𝑆={\mathbb{Z}}^{+}.

The proof is almost the same as that of the first principle of mathematical induction, except for a tiny difference in how we argue the contradiction.

Proof

We prove by contradiction. Assume 𝑆≠Z+𝑆\neq {\mathbb{Z}}^{+}. Equivalently, Z+∖𝑆≠∅{\mathbb{Z}}^{+}∖𝑆\neq \emptyset . Since Z+∖𝑆⊆Z+{\mathbb{Z}}^{+}∖𝑆\subseteq {\mathbb{Z}}^{+} and Z+∖𝑆≠∅{\mathbb{Z}}^{+}∖𝑆\neq \emptyset , the well-ordering principle gives a least element 𝑛∈Z+∖𝑆𝑛\in {\mathbb{Z}}^{+}∖𝑆.

Now, 𝑛≥2𝑛\geq 2. Since 𝑛∉𝑆𝑛\notin 𝑆, there must exist 𝑚∈{1,…,𝑛−1}𝑚\in \{1,\ldots ,𝑛-1\} such that 𝑚∉𝑆𝑚\notin 𝑆. Otherwise, by the property of 𝑆𝑆, it would mean that 𝑛∈𝑆𝑛\in 𝑆, which is not the case. So, 𝑚∈Z+∖𝑆𝑚\in {\mathbb{Z}}^{+}∖𝑆, contradicting that 𝑛𝑛 the least element of Z+∖𝑆{\mathbb{Z}}^{+}∖𝑆. Therefore the assumption is false, and 𝑆=Z+𝑆={\mathbb{Z}}^{+}.

We can also prove Theorem 1.4 by mathematical induction since we have already proved the first principle Theorem 1.3.

Proof

Let 𝑃(𝑛)𝑃(𝑛) be the statement that 𝑆𝑆 contains all positive integers less than or equal to 𝑛𝑛, i.e. 𝑚∈𝑆𝑚\in 𝑆 for all 𝑚≤𝑛𝑚\leq 𝑛.

Base case. 𝑛=1𝑛=1. This is the given condition. There is nothing to show.

Inductive step. Assume 𝑃(𝑘)𝑃(𝑘), which states that {1,…,𝑘}⊆𝑆\{1,\ldots ,𝑘\}\subseteq 𝑆. By the property of 𝑆𝑆, we have 𝑘+1∈𝑆𝑘+1\in 𝑆, which proves 𝑃(𝑘+1)𝑃(𝑘+1).

Therefore, 𝑃(𝑛)𝑃(𝑛) holds for all 𝑛∈Z+𝑛\in {\mathbb{Z}}^{+}. This is equivalent to say that 𝑆𝑆 contains all positive integers.

Recurrence Relations

A recurrence relation is an equation according to which the 𝑛𝑛-th term 𝑎𝑛{𝑎}_{𝑛} of a sequence {𝑎𝑛}\{{𝑎}_{𝑛}\} is equal to some expression in the preceding terms.

If each 𝑎𝑛{𝑎}_{𝑛} (except 𝑎1{𝑎}_{1}) is determined only by the immediatel peceding term, we can write

𝑎𝑛=φ(𝑛,𝑎𝑛−1),𝑛≥2.{𝑎}_{𝑛}=\varphi (𝑛,{𝑎}_{𝑛-1}),\quad 𝑛\geq 2.

The reason why φ\varphi also depends on 𝑛𝑛 (as its first parameter) is that the expression of each pair of consecutive terms may differ. Then by the principle of mathematical induction Theorem 1.3, we may conclude that 𝑎𝑛{𝑎}_{𝑛} is well-defined for each 𝑛∈Z+𝑛\in {\mathbb{Z}}^{+}, and hence the sequence {𝑎𝑛}\{{𝑎}_{𝑛}\} is defined.

The recurrence relation of the arithmetic sequence 𝑎𝑛=𝑎+(𝑛−1)𝑑{𝑎}_{𝑛}=𝑎+(𝑛-1)𝑑 is

𝑎𝑛=𝑎𝑛−1+𝑑,𝑛≥2.{𝑎}_{𝑛}={𝑎}_{𝑛-1}+𝑑,\quad 𝑛\geq 2.

And the recurrence relation of the geometric sequence 𝑎𝑛=𝑎𝑟𝑛−1{𝑎}_{𝑛}=𝑎{𝑟}^{𝑛-1} is

𝑎𝑛=𝑎𝑛−1𝑟,𝑛≥2.{𝑎}_{𝑛}={𝑎}_{𝑛-1}𝑟,\quad 𝑛\geq 2.

In the preceding two examples, φ(𝑛,𝑎𝑛−1)\varphi (𝑛,{𝑎}_{𝑛-1}) only depends on 𝑎𝑛−1{𝑎}_{𝑛-1}.

Consider the sequence of factorials 𝑎𝑛=𝑛!{𝑎}_{𝑛}=𝑛!. Its recurrence relation is

𝑎𝑛=𝑛𝑎𝑛−1,𝑛≥2.{𝑎}_{𝑛}=𝑛{𝑎}_{𝑛-1},\quad 𝑛\geq 2.

Here φ(𝑛,𝑎𝑛−1)=𝑛𝑎𝑛−1\varphi (𝑛,{𝑎}_{𝑛-1})=𝑛{𝑎}_{𝑛-1} does not only depend on 𝑎𝑛−1{𝑎}_{𝑛-1}.

More generally, the 𝑛𝑛-th term 𝑎𝑛{𝑎}_{𝑛} may depend on its preceding 𝑘𝑘 terms, i.e.,

𝑎𝑛=φ(𝑛,𝑎𝑛−1,𝑎𝑛−2,…,𝑎𝑛−𝑘),𝑛≥𝑘+1.{𝑎}_{𝑛}=\varphi (𝑛,{𝑎}_{𝑛-1},{𝑎}_{𝑛-2},\ldots ,{𝑎}_{𝑛-𝑘}),\quad 𝑛\geq 𝑘+1.

The classical example is the Fibonacci sequence

𝑎𝑛=𝑎𝑛−1+𝑎𝑛−2,𝑛≥3{𝑎}_{𝑛}={𝑎}_{𝑛-1}+{𝑎}_{𝑛-2},\quad 𝑛\geq 3

with 𝑎1=𝑎2=1{𝑎}_{1}={𝑎}_{2}=1.

By the second principle of mathematical induction Theorem 1.4, we can show that 𝑎𝑛{𝑎}_{𝑛} is defined for all positive integer numbers 𝑛𝑛.

Exercises

Consider the 𝑛𝑛-th partial sum of the harmonic series, 𝐻𝑛=∑𝑗=1𝑛1𝑗{𝐻}_{𝑛}={\sum }_{𝑗=1}^{𝑛}\frac{1}{𝑗}.

Exercise 1.2

Show by mathematical induction that

𝐻2𝑛≥1+𝑛2.{𝐻}_{{2}^{𝑛}}\geq 1+\frac{𝑛}{2}.
Proof

Let 𝑃(𝑛)𝑃(𝑛) be the statement that the inequality to show holds for 𝑛𝑛.

Base case. 𝑛=1𝑛=1. LHS is 𝐻2=1+1/2{𝐻}_{2}=1+1/2 and RHS is 1+1/21+1/2. In this case LHS equals to RHS. Hence, 𝑃(1)𝑃(1) holds.

Inductive step. Assume 𝑃(𝑘)𝑃(𝑘). Let 𝑛=𝑘+1𝑛=𝑘+1. We have

𝐻2𝑛=𝐻2𝑘+1=∑𝑗=12𝑘1𝑗⏟≥1+𝑘2  by  𝑃(𝑘)+∑𝑗=2𝑘+12𝑘+11𝑗≥1+𝑘2+∑𝑗=2𝑘+12𝑘+11𝑗Each  𝑗≤2𝑘+1≥1+𝑘2+∑𝑗=2𝑘+12𝑘+112𝑘+1≥1+𝑘2+12𝑘+1⋅2𝑘≥1+𝑘2+12≥1+𝑘+12.\begin{aligned}{𝐻}_{{2}^{𝑛}}={𝐻}_{{2}^{𝑘+1}} & = & & \underbrace{\sum _{𝑗=1}^{{2}^{𝑘}}\frac{1}{𝑗}}_{\geq 1+\frac{𝑘}{2}\, \text{ by }\, 𝑃(𝑘)}+\sum _{𝑗={2}^{𝑘}+1}^{{2}^{𝑘+1}}\frac{1}{𝑗} \\ & \geq & & 1+\frac{𝑘}{2}+\sum _{𝑗={2}^{𝑘}+1}^{{2}^{𝑘+1}}\frac{1}{𝑗} \\ & & & \text{Each }\, 𝑗\leq {2}^{𝑘+1} \\ & \geq & & 1+\frac{𝑘}{2}+\sum _{𝑗={2}^{𝑘}+1}^{{2}^{𝑘+1}}\frac{1}{{2}^{𝑘+1}} \\ & \geq & & 1+\frac{𝑘}{2}+\frac{1}{{2}^{𝑘+1}}\cdot {2}^{𝑘} \\ & \geq & & 1+\frac{𝑘}{2}+\frac{1}{2} \\ & \geq & & 1+\frac{𝑘+1}{2}.\end{aligned}

This proves 𝑃(𝑘+1)𝑃(𝑘+1).