TRƯỜNG ĐẠI HỌC BÁCH KHOA HÀ NỘI

VIỆN CÔNG NGHỆ THÔNG TIN VÀ TRUYỀN THÔNG

Discrete Mathematics

Nguyễn Khánh Phương

Department of Computer Science

School of Information and Communication Technology
E-mail: [email protected]
 PART 1
COMBINATORIAL THEORY
(Lý thuyết tổ hợp)

PART 2
GRAPH THEORY
(Lý thuyết đồ thị) 2
Contents of Part 1
Chapter 0: Sets, Relations
Chapter 1: Counting problem
Chapter 2: Existence problem
Chapter 3: Enumeration problem
Chapter 4: Combinatorial optimization problem

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN 3
 Contents of Part 1: Combinatorial Theory
Chapter 1. Counting problem
• This is the problem aiming to answer the question: “How many ways are there that satisfy
given conditions?” The counting method is usually based on some basic principles and some
results to count simple configurations.
• Counting problems are effectively applied to evaluation tasks such as calculating the
probability of an event, calculating the complexity of an algorithm
Chapter 2. Existence problem
In the counting problem, configuration existence is obvious; in the existence problem, we need to
answer the question: "Is there a combinatorial configuration that satisfies given properties ?”

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
 TRƯỜNG ĐẠI HỌC BÁCH KHOA HÀ NỘI
VIỆN CÔNG NGHỆ THÔNG TIN VÀ TRUYỀN THÔNG

Chapter 2
Existence roblem

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
Content
1. Introduction to existence problems
2. Basic proof methods
3. Dirichlet principle (pigeonhole principle)

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
1. Introduction to existence problems
• In the “Counting problem” chapter, we focused on counting the
combinatorial configurations. In those problems, the existence of the
configurations is obvious, and the main object is to count the number of
elements that satisfy the given properties.
• However, in many combinatorial problems, it is very difficult to point
out the existence of a configuration that satisfies given properties:
– For example, when a player needs to calculate his moves to answer whether there is
a possibility of winning or not?
– A person needs to search for the key to decipher a secret code that he does not know
if this is really the opponent's encrypted message, or just the secret code issued by
the opponent to ensure the safety of real telegrams ...
• In combinatorics, besides the counting problem, there is another very
important problem is considering the existence of combinatorial
configurations satisfying given properties - the problem of existence.

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
The 36 officers problem
• This problem is proposed by Euler, it is described as following:
“You're in command of an army that consists of six regiments, each
containing six officers of six different ranks. Can you arrange the
officers in a 6x6 square so that each row and each column of the square
holds only one officer from each regiment and only one officer from each
rank?

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
The 36 officers problem
• Using:
– A, B, C, D, E, F refer to 6 regiments, respectively
– 1, 2, 3, 4, 5, 6 refer to 6 ranks of officers, respectively.
• This problem could be generalized by replacing 6 by n.
• In the case n = 4, a solution to the problem is:
A1 B2 C3 D4
B4 A3 D2 C1
C2 D1 A4 B3
D3 C4 B1 A2
• In the case n = 5, a solution to the problem is:
A1 B2 C3 D4 E5
C4 D5 E1 A2 B3
E2 A3 B4 C5 D1
B5 C1 D2 E3 A4
D3 E4 A5 B1 C2

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
The 36 officers problem
• As the solution to the problem can be represented by combining two same
size squares array of symbols: one with upper-case letter and the other with
numbers. The problem is also called as Graeco-Latin square (also called an
Euler square)
• Euler struggled to find the solution to the 36 officers problem but was
unsuccessful. However, the solution exists when n =4, 5 and 7. Thus, he
proposed a general hypothesis: The Graeco-Latin squares are impossible if
n = 4k + 2.
• Tarri, in 1901, proves the correct hypothesis for n = 6, by examining all
possible ratings.
• In 1960, three American mathematicians, Boce, Parker, and Srikanda
showed a solution with n = 10 and then showed a method to construct
Graeco-Latin squares for all n = 4k + 2, with k > 1.

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
Four color theorem (Bµi to¸n 4 mµu)
• There are problems that its content could be explained to anyone, but its
solutions anyone can try to find, but it is difficult to find. The 4-color
problem is such a problem.
• The problem can be stated visually as follows: Prove that all maps on the
plane could be colored with 4 colors so that no two adjacent regions
(neighbor regions) are colored by the same color.
• Note that each country is considered a connected region and the two
countries are called neighbors if they share a border of a continuous line.

A is not connected region A and C are not neighbors

Four color theorem (Bµi to¸n 4 mµu)
• The number 4 is not random. It has been shown that all maps are
colored with color numbers greater than 4, but with color numbers
less than 4 it cannot be colored. For example, the map with 4 countries
in the picture below cannot be colored if using less than 4 colors.

A
B
C

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
Four color theorem (Bµi to¸n 4 mµu)
• This problem was first proposed in 1852 by Francis Guthrie, while
trying to color the map of counties of England.
• In 110 years many proofs have been published but have been flawed.
• In 1976, Appel and Haken gave the proof by computer:
K. Appel and W. Hankin, "Every planar map is 4-colorable," Bulletin of the
AMS, Volume 82 (1976), 711-712.

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
Content
1. Introduction to existence problems
2. Basic proof methods
3. Dirichlet principle (pigeonhole principle)

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
2. Basic proof methods
2.1. Direct Proof (Chứng minh trực tiếp)
2.2. Proof by Contradiction (Chứng minh bằng phản chứng)
2.3. Proof by Contrapositive (Chứng minh bằng phản đề)
2.4. Proof by Mathematical Induction (Chứng minh bằng qui nạp
toán học)
2.1. Direct proofs (Chứng minh trực tiếp)
We begin with an example demonstrating the transitivity of divisibility.
Theorem. If a divides b and b divides c then a divides c.
Prove. By using the definition of the divisibility, there exist integers k1
and k2 such that
a = b k1 and b = c k2.
Then
a = b k2 = c k1 k2 .
Let k = k1 k2. We have k as an integer, and a = ck, so by the definition of
divisibility, a divides c.

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
2.1. Direct proofs (Chứng minh trực tiếp)

If P, Then Q
• In most theorems, exercises or tests, you need to prove the form "If P,
Then Q".
• In this example: “if a divides b and b divides c, then a divides c”
– "P" is "If a divides b and b divides c" and "Q" is "a divides c".
• This is the standard state of many theorems.
• The direct proof can be conceived as a series of inferences beginning
with "P" and ending with "Q":
P  ...  Q
Most of proof is direct. When you have to prove, try starting with direct
proof, unless you have a good reason not to.

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
Example 1
Theorem: The sum of two odd integers is an even integer;
Proof:
– Let a and b be odd integers.
– Then a = 2s + 1 and b = 2t + 1 for some integers s and t.
– Then a + b = (2s + 1) + (2t + 1)
= 2(s + t + 1).
– Therefore, a + b is an even integer.
– Finish the proof.

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
Example 2
• A rational number is a number that equals the quotient of two integers.
• Let Q denote the set of rational numbers.
Theorem: The sum of two rational numbers is rational.
• Proof:
– Let r and s be rational numbers.
– Let r = a/b and s = c/d, where a, b, c, d are integers, where b, d > 0
– Then r + s = (ad + bc)/bd.
• We know that ad + bc is an integer.
• We know that bd is an integer.
• We also know that bd  0.
• Therefore, r + s is a rational number.

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
Example 3
• A rational number is a number that equals the quotient of two integers.
• Let Q denote the set of rational numbers.
Theorem: Between every two distinct rationals, there is a rational.
• Proof:
– Let r, s  Q.
– Without loss of generality, we may assume r < s.
– Let t = (r + s)/2.
– Then t  Q. (proof?)
– We must show that r < t < s:
• Since r < s, it follows that
2r < r + s < 2s
• Then divide by 2 to get
r < (r + s)/2 < s
• Therefore, r < t < s.
Example 4
• Prove that if n is an integer and n3+5 is odd, then n is even
Proof:
– n3+5 = 2k+1 for some integer k (definition of odd numbers)
– n3 = 2k-4
–
Umm, so direct proof didn’t work out.

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
2. Basic proof methods
2.1. Direct Proof (Chứng minh trực tiếp)
2.2. Proof by Contradiction (Chứng minh bằng phản chứng)
2.3. Proof by Contrapositive (Chứng minh bằng phản đề)
2.4. Proof by Mathematical Induction (Chứng minh bằng qui nạp
toán học)

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
2.2. Proof by Contradiction
Requirement: Proof the statement P
Proof by contradiction:
• Assume it is false (Assume ¬P)
• Prove that ¬P cannot occur
– (it means a contradiction exists: not satisfying the properties given
in the problem or come to the absurd such as 1 = 0)

Requirement: Proof “If P, Then Q",

Proof by contradiction:
• Assume it is false (Assume that "P and Not Q" are true).
• It thus means a contradiction exists.

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
Example 4
• Prove that if n is an integer and n3+5 is odd, then n is even
Proof:
– n3+5 = 2k+1 for some integer k (definition of odd numbers)
– n3 = 2k-4
–
Umm, so direct proof didn’t work out.

Rephrased: If n3+5 is odd, then n is even

Proof by contradiction:
Assume that n3+5 is odd, and n is odd
• n is odd  n=2k+1 for some integer k (definition of odd numbers)
• n3+5 = (2k+1)3+5 = 8k3+12k2+6k+6 = 2(4k3+6k2+3k+3) is odd ?!!??
– As 2(4k3+6k2+3k+3) is 2 times an integer, it must be even
– Contradiction!
2.2. Proof by Contradiction
Example 5. Given 7 segments with length greater than 10 and less than 100. Prove
that one can always find 3 segments that can be assembled into a triangle.
Solution:
Note that, the necessary and sufficient condition for 3 segments to be assembled
into a triangle is: the sum of the lengths of two smaller segments must be greater
than the length of the largest segment.
Ordering the given 7 segments in ascending order of length, we have:
10 < a1  a2  ...  a7 < 100.
We need to prove that: from the above ordered sequence, one always can find 3
consecutive segments such that the sum of length of the first 2 segments is greater
than that of the last.

Proof by contradiction:
Assume this does not happen (not Q: could not find any 3 consecutive segments
such that the sum of length of the first 2 segments is greater than that of the last).
2.2. Proof by Contradiction
• Proof by contradiction: Assume this does not happen (could not find any 3
consecutive segments such that the sum of length of the first 2 segments is greater
than that of the last). Thus, we have equalities:
a 1 + a2  a 3,
a 2 + a3  a 4,
a 3 + a4  a 5,
a 4 + a5  a 6,
a 5 + a6  a 7.
• As a1, a2 are greater than 10, we have a3 > 20. As a2 > 10 and a3 > 20, we
have a4 > 30, ..., continuing we have a5 > 50, a6 > 80 and a7 > 130.
• The last equalities a7 > 130 is contradiction with the given condition of the
problem: all segments has the length less than 100. Thus, the assumption is
false.
NGUYỄN KHÁNH PHƯƠNG
Bộ môn KHMT – ĐHBK HN
2.2. Proof by Contradiction
Example 6. The vertices of a decagon (a ten-sided polygon or 10-gon) decimal
are arbitrarily numbered by the integers 0, 1, ..., 9. Prove that you can always
find three consecutive vertices whose sum of the numbers is greater than 13.
Solution: Let x1, x2, . . ., x10 be the numbers assigned to the vertices of 1, 2, ...,
10 of the decagon.

Proof by contradiction:
Assume this does not happen (not Q: could not find any three consecutive
vertices to satisfy the assertion of the example). Then we have:
x1 + x2 + x3  13,
x2 + x3 + x4  13,
+ . . . . .
Summing side by side of all inequalities
x9 + x10 + x1  13,
x10 + x1 + x2  13,
3(x1 + x2 + . . . + x10)  130
2.2. Proof by Contradiction
• Summing side by side of all above inequalities, we have
3(x1 + x2 + . . . + x10)  130.
• On the other hand, as
3(x1 + x2 + . . . + x10)
= 3 (0 + 1 + 2 + . . . + 9)
= 135,
• Therefore:
135 = 3(x1 + x2 + . . . + x10)  130
The obtained contradiction proved the claim in the example to be
correct. .

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
2.2. Proof by Contradiction
Example 7. Proof that it is not possible to connect 31 computers into
a network such that each computer is connected to exactly 5 other
computers.

Solution:
Suppose the opposite is finding a way to connect 31 computers so that
each computer is connected to exactly 5 other computers. Then the
number of connected channels is
5 x 31 / 2 = 75.5 ?! (not an integer)
The obtained absurdity proves the claim in the example to be true.

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
2. Basic proof methods
2.1. Direct Proof (Chứng minh trực tiếp)
2.2. Proof by Contradiction (Chứng minh bằng phản chứng)
2.3. Proof by Contrapositive (Chứng minh bằng phản đề)
2.4. Proof by Mathematical Induction (Chứng minh bằng qui nạp
toán học)

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
2.3. Proof by Contrapositive (Chứng minh bằng phản đề)
Proof by contrapositive uses the logically equivalence of two statement
“If P then Q” (P  Q) and “If not Q then not P (¬Q  ¬P):
(P  Q) (¬Q  ¬P)
Example:
The statement “If this is my car, then its color is red”
is equivalent to
“If its color is not red, then it is not my car".
• Thus, to prove “If P, then Q" by using contrapositive proof, we prove
“If not Q then not P”).

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
Example (P  Q) (¬Q  ¬P)
Example 1. If x and y are two integer such that x+y is even, then x and y
are either both even or odd.
Proof. The contrapositive is
“If x and y are neither both even nor odd, then its sum is odd."
• As x and y are neither both even nor odd. Without loss of generality,
we may assume x is even and y is odd. Then we could find integers k
and m such that x = 2k and y = 2m+1. Now we calculate x+y = 2k +
2m + 1 = 2(k+m) + 1, thus x+y is odd.
• Therefore, we have proven the result by contraposition

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
Example (P  Q) (¬Q  ¬P)
Example 2. If n is a positive integer such that n mod 4 is equal to either 2 or 3,
then n is not a perfect square.

Proof. The contrapositive is:

“If n is a perfect square then n mod 4 is equal to either 0 or 1"
• Assume n = k2. There are 4 cases:
– If k mod 4 = 0, then k = 4q, where q is positive integer. Thus, n = k2 = 16
q2 = 4(4 q2) , so n mod 4 = 0.
– If k mod 4 = 1, then k = 4q + 1, where q is positive integer. Thus, n = k2 =
16 q2 + 8 q + 1= 4(4 q2 + 2 q) + 1, so n mod 4 = 1.
– If k mod 4 = 2, then k = 4q + 2, where q is positive integer. Thus, n = k2 =
16 q2 + 16 q + 4 = 4(4 q2 + 4 q + 1), so n mod 4 = 0.
– If k mod 4 = 3, thì k = 4q + 3, where q is positive integer. Thus, n = k2 =
16 q2 + 24 q + 9 = 4(4 q2 + 6 q + 2) + 1, so n mod 4 = 1.
2.3. Proof by Contrapositive (Chứng minh bằng phản đề)
The difference between Contrapositive proof and Contradiction proof ?

We are asked to prove a conditional statement, or a statement of the form "If

P, Then Q”:
• Contradiction proof (Chứng minh bằng phản chứng): Assume “If P and
Not Q is true”, then need to show some sort of fallacy  thus the
assumption is false.
• Contrapositive proof (Chứng minh bằng phản đề): We do not prove “If P,
then Q” directly, instead, we prove “If not Q, then not P”.
The advantage of contrapositive proof is that your goal is clear: prove not
P; while in contradiction proof, your goal is to prove a contradiction (you
need to arrive at some sort of fallacy), but it is not always clear what the
contradiction is going to be at the start.

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
2. Basic proof methods
2.1. Direct Proof (Chứng minh trực tiếp)
2.2. Proof by Contradiction (Chứng minh bằng phản chứng)
2.3. Proof by Contrapositive (Chứng minh bằng phản đề)
2.4. Proof by Mathematical Induction (Chứng minh bằng qui
nạp toán học)

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
2.4. Proof by Mathematical Induction
• This is a very useful proof technique when we have to prove that the
proposition P(n) is true for all natural numbers n  n0.
• Similar to the "domino effect" principle.
Outline of proof by Induction:
• Basic step: Prove the first statement P(n0) is true
• Inductive step: Given any integer n  n0 , prove that P(n)P(n+1)
(Assuming P(n) is true and showing it forces P(n+1) is true)
• Conclusion: P(n) is true n  n0

(The assumption that P(n) is true is called the inductive hypothesis)

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
The “Domino Effect”

 Step #1: Domino #0 falls.

 Step #2: For all nN,
if domino #n falls, then
domino #n+1 also falls. 6
 Conclusion: All dominos 5
must fall! 4
3
Note: This happens even 2
when there are infinitely 1
many dominoes! 0
NGUYỄN KHÁNH PHƯƠNG
Bộ môn KHMT – ĐHBK HN
Outline of proof by Induction:
We need to prove P(n) is true n  n0 .
• Basic step (Bước cơ sở): Prove P(n0) is true.
• Inductive hypothesis (Giả thiết qui nạp): Assume P(n) is true
• Inductive step (Bước chuyển qui nạp): Prove P(n+1) is true.
• Conclusion: It follows by mathematical induction that P(n) is true n  n0

“The First Principle

of Mathematical Induction”
“Nguyên lý qui nạp toán học thứ nhất”
Example 1
Prove that it is always possible to cover a chessboard of size 2n
 2n (n > 1) with T-omino cards.

• Basic step: Prove that it is true when n = 2

Basic step: Chessboard 22 x 22
Basic step: Chessboard 22 x 22
Basic step: Chessboard 22 x 22
Basic step: Chessboard 22 x 22
Inductive step
Assume we can cover chessboard of size 2n  2n. We need to prove
also can cover chessboard of size 2n+1  2n+1.
Actually, divide the chessboard of size 2n+1  2n+1 into 4 parts, each of
size 2n  2n. According to the inductive hypothesis (giả thiết qui nạp),
each of these parts could be covered by T-ominos. Putting them on the
chessboard of size 2n+1  2n+1 , we then get the answer for the problem.
Example 2
• At the end of a volleyball championship consisting of n
participating teams, in which teams take a round trip, the
captains of teams are invited to stand in a row to take pictures.
• P(n): It is always possible to place n captains in a row so that
except for two captains standing at the leftmost and rightmost,
each of them is always next to the captain of the team that won
his team and the captain of the team that lost his team in the
tournament.

lose win win lose

Example 2
Proof. We prove by induction
• Basic step: P(1) is always true.
• Assume P(n-1) is true, we need to prove P(n) is true.
– First, we rank n-1 captains of teams 1, 2,..., n-1. According to the
inductive hypothesis, it is possible to arrange them in the row that
satisfies the given condition of the problem. Without loss of
generality, we may assume this row is:
1 2  ...  n-1
(1 wins 2, 2 wins 3, … n-2 wins n-1)
Example 2
– Now we find the position to place the captain of the nth team. There are 3
cases:
• If team n won team 1, then the row is:
n  1 2  ...  n-1.
• If team n lost team n-1, then the row is:
1 2  ...  n-1  n .
• If team n lost team 1 and won team n-1:
– Let k be the smallest index such that team n won team k.
– Obviously, such k exists.
– The row obtained from the row of n-1 teams is by inserting the
captain of the nth team between the captains of team k-1 and team k.
Example 2
1 2  ... k-1 k  k+1 ... n-1

The row: 1 ... k-1 n k  k+1 ... n-1

Second Principle of Induction – Strong Induction

Occasionally it happens in induction proof that it is difficult to show that P(n)

forces P(n+1) to be true. Instead, you may find that you need to use the fact that
some “lower” statements P(k) (with k < n) force P(n+1) to be true. For these
situations, you can use a slight variant of induction called strong induction.
Strong induction works just like regular induction, except that in Inductive step,
instead of assuming P(n) is true and showing this forces P(n+1) to be true, we
assume that all the statements P(n0), P(n0+1),..,P(n) are true and show this
forces P(n+1) to be true. The idea is that if it always happens that the first n
dominoes falling makes the (n +1)th domino fall, then all the dominoes must
fall. Here is the outline:
We need to prove P(n) is true n  n0 .
• Basic step (Bước cơ sở): Prove P(n0) is true or the first several P(n).
• Inductive hypothesis (Giả thiết qui nạp): Assume any integer k:
 n0  k  n P(k) is true
• Inductive step (Bước chuyển qui nạp): Prove P(n+1) is true.
• Conclusion: It follows by mathematical induction that P(n) is true n  n0
Second Principle of Induction – Strong Induction

Outline of proof by strong Induction:

Assume P is true for all previous cases n0 to n

• P(n0)
n  n0 : ( n0  k  n P(k))  P(n+1)
Conclusion n  n0 : P(n)

• The difference with 1st principle of induction:

– Inductive step use stronger assumption: P(k) is true for all k  n,
not just the case k=n as in the 1st principle of induction.
Outline of proof by strong Induction
We need to prove P(n) is true n  n0.
• Basic step: Prove that P(n0) is true.
• Inductive hypothesis : Assume P(k) is true  n0  k  n.
• Inductive step : Prove P(n+1) is true.
• Conclusion: It follows by mathematical induction that P(n)
is true n  n0.
Example 3
Prove that when natural n ≥ 1, we have 12| n4 – n2

First look at how 1st principle of induction would be problematic:

• Basic step: Need to show 12| n4 – n2 is true for n = 1. This part is easy because it
reduces to 12|0, which is clearly true.
• Inductive step: Assume that 12| n4 – n2 and show this implies 12| (n+1)4 – (n+1)2
– Now 12| n4 – n2 means n4 – n2 = 12a for some a ℤ.
– Thus, we need to use this to try showing that (n+1)4 – (n+1)2 =12b for some b ℤ
(n+1)4 – (n+1)2 = (n4+4n3+6n2+4n+1)-(n2+2n+1)
= (n4-n2) + 4n3+6n2+6n
= 12a + 4n3+6n2+6n
At this point, we are stuck because we can’t factor out of 12. Now let’s see how strong
induction can get us out of this bind.
Example 3: Prove that when natural n ≥ 1, we have 12| n4 – n2
Strong induction involves assuming each of statements P(n0) = S1 , P(n0+1) = S2
,...,P(n)= Sn , is true, and showing that this forces P(n+1) = Sn+1 , to be true.
In particular, if S1 through Sn are true, then certainly Sn−5 is true, provided that 1 ≤ n−5
< n.
The idea is then to show Sn−5 Sn+1 instead of Sn Sn+1
For this to make sense, our basic step must involve checking that S1,S2,S3,S4,S5,S6 are
all true. Once this is established, Sn−5 Sn+1 will imply that the other Sn are all true.
For example, if n = 6, then Sn−5 Sn+1 is S1 S7, so S7 is true;
for n = 7, then Sn−5 Sn+1 is S2 S8, so S8 is true, etc.
Example 3: Prove that when natural n ≥ 1, we have 12| n4 – n2
• Basic step: Need to show 12| n4 – n2 is true for n = 1,2,3,4,5,6
– If n = 1, 12 divides n4 − n2 = 14 −12 = 0.
– If n = 2, 12 divides n4 − n2 = 24 −22 = 12.
– If n = 3, 12 divides n4 − n2 = 34 −32 = 72.
– If n = 4, 12 divides n4 − n2 = 44 −42 = 240.
– If n = 5, 12 divides n4 − n2 = 54 −52 = 600.
– If n = 6, 12 divides n4 − n2 = 64 −62 = 1260.
• Inductive step: Let n ≥ 6 and assume that 12| k4 – k2 for 1 ≤ k ≤ n (that is, assume statements P(1),
P(2),..,P(n) are all true), and show this implies 12| (n+1)4 – (n+1)2 (that is, we must show that
P(n+1) is true)
– Since P(n-5) is true, we have 12| (n-5)4 –(n-5)2 . For simplicity, let’s set m = n-5, so we know
12| m4 –m2 means m4 – m2 = 12a for some a ∈ ℤ.
– Thus, we need to use this to try showing that (n+1)4 – (n+1)2 =12b for some b ∈ ℤ
(n+1)4 – (n+1)2 = (m+6)4 – (m+6)2
= (m4+24m3+216m2+865m+1296) - (m2+12m+36)
= (m4-m2) + 24m3+216m2+852m + 1260
= 12a + 24m3+216m2+852m + 1260
= 12(a + 2m3+18m2+71m + 105)
As (a + 2m3+18m2+71m + 105) is integer, we get 12| (n+1)4 – (n+1)2
This shows by strong induction that 12| n4 – n2 for all natural n ≥ 1
Example 4
For all naturals n ≥ 0, we have 10| n5 – n
Proof: using strong induction
• Basic step: Prove P(0) is true.
When n = 0: we have 05-0=0 và 10|0  statement is true for n = 0
• Inductive hypothesis: Assume P(k) is true  0  k  n.
Assume k ≥ 0, and 10 | k5 – k is true for 0  k  n
As 0  n-1  n, inductive hypothesis implies that P(n-1) is true, so
(n-1)5 – (n-1) = 10c for c ℤ
• Inductive step: Prove P(n+1) is true
10 | (n+1)5 – (n+1)
Example 4

(n+1)5 – (n+1)
= [(n-1)+2]5 – [(n-1)+2]
= (n-1)5 + 10(n-1)4+40(n-1)3+80(n-1)2+80(n-1)+32
-(n-1) -2
= [(n-1)5 – (n-1)] + 10[(n-1)4 + 4(n-1)3 +8(n-1)2+8(n-1)+3]
10c
= 10 [c + (n-1)4 + 4(n-1)3 +8(n-1)2+8(n-1)+3]
integer
 10 | (n+1)5 – (n+1)
Example 4
For all naturals n ≥ 0, we have 10| n5 – n
Proof: using strong induction
• Basic step: Prove P(0) is true.
When n = 0: we have 05-0=0 and 10|0  statement is true for n = 0
When n = 1: we have 15-1=0 and 10|0  statement is true for n = 1
• Inductive hypothesis: Assume P(k) is true  0  k  n.
Asume k ≥ 0, and 10 | k5 – k is true for với 0  k  n
As 0  n-1  n, inductive hypothesis implies that P(n-1) is true, so
(n-1)5 – (n-1) = 10c for c ℤ
• Inductive step: Prove P(n+1) is true
10 | (n+1)5 – (n+1)
Content
1. Introduction to existence problems
2. Basic proof methods
3. Dirichlet principle (pigeonhole principle)

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
3. Dirichlet principle
3.1. Principle statement
3.2. Application examples

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
3.1. Dirichlet principle
If putting more than n objects into n boxes then at least one box has at
least 2 objects (  2 ).

• 7 objects
• 6 boxes

Proof. (Contradiction).
The reverse assumption is that one could not find a box containing  2 objects.
 That means that each box contains  1 object.
 The total number of objects put in n boxes  n
Contrary to the given condition of problem that more than n objects are put in
them.
3.1. Dirichlet principle
The above principle has been successfully applied by the German
mathematician Dirichlet to solving many existence problems in
combinatorics.
It is also presented in the language of pigeons:
“If one put more than n pigeons into n pigeonholes, then at least one hole
has more than one pigeon ( 2).”
So the principle is also known as " Pigeonhole principle ".

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
Example
If putting more than n objects into n boxes then at least one box has at least 2
objects (  2 ).

Example 1. Among 13 people, there are always 2 people born in

the same month as there are only 12 months.
Example 2. In the exam, the test score is assessed by an integer
between 0 and 100. Then at least how many students must take
the test so that it is certainly to exist 2 students get same result ?
Solution. There are 101 different results
 Using Dirichlet principle, the number of students is 102

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
Generalized Pigeonhole Principle (Nguyên lý Dirichlet tổng quát)

Dirichlet principle: If putting more than n objects into n boxes then at

least one box has  2 objects.
When the number of objects putting into k boxes is much larger than the
k, it is obviously that the claim in the principle about the existence of a
box containing at least two objects is too small. In such a case, we use
the following generalized Dirichlet principle:
"If putting n objects into k boxes, one could always find at least one box
containing  n/k objects".
Here the symbol  is the least integer greater than or equal to .
e.g.: 3.14 = 4

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
Generalized Pigeonhole Principle

"If putting n objects into k boxes, one could always find at least one box containing 
n/k objects".
Proof by contradiction.

Assume the claim in principle is not true.

 Each box contains  n/k − 1 < [(n/k)+1] -1 = n/k objects

 There are
There are k boxes < k(n/k) = n objects

The obtained contradiction has proved the principle.

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
Example
Dirichlet principle: "If putting n objects into k boxes, one could always
find at least one box containing  n/k objects".
Example 3. In a group of 100 people, what is the minimum number of people
that were born in the same month ?.
Solution: Putting people born in the same month into one group. There are 12
months. Therefore, according to the Dirichlet principle, there exists at least one
group consisting  100/12 = 9 people

Example 4. There are 5 different types of scholarships. What is the minimum

number of students so that there are at least 6 students getting the same type of
scholarship?
Solution. The minimum number of students needed to ensure that 6 students
getting the same scholarship is the smallest integer n such that n/5 = 6. The
smallest integer is n = 55+1 = 26.
Thus, 26 is the smallest number of student needed to ensure that there are 6
students getting the same type of scholarship
Example
Example 5. What is the minimum number of area codes so that 25 millions
cellphones, each assigned to a unique 10-digit number of the form NXX-NXX-
XXXX where the first 3 numbers is the area code, N representing the numbers
through 2 to 9, and X representing number through 0 to 9
Solution: The number of phone numbers in the form NXX-XXXX is:
8 millions.
According to Dirichlet principle, among 25 millions cellphones, there are
always at least 25/8 = 4 cellphones with the same number. Therefore, we need
at least 4 area codes to ensure that 25 millions cellphones do not have the same
number.

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
Example
Dirichlet principle: "If putting n objects into k boxes, one could always find at least one
box containing  n/k objects".

Example 6.
There are 50 baskets. Every basket contains apples, but not more than 24apples.
Prove that there are at least 3 baskets containing the same amount of apples.
Solution:
Number of Objects ~ Number of baskets  50 baskets
Number of Boxes ~ Number of apples in each basket  24 boxes

1 2 3 24

We put each basket in one of 24 boxes depending on the number of apples in

the basket.
Dirichlet principle: at least one box contains 50/24 = 3 baskets
 There is at least 3 baskets in the same box (same amount of apples)
Example
Example 7. Prove that among 4 arbitrary natural numbers, there are always 2
numbers with difference is divisible by 3.
Solution: When diving a natural number by 3, there could be 3 remainders (0, 1
or 2). There are 4 numbers, so according to Dirichlet principle, there are 2
numbers of them have the same remainders when divided by 3, so we can
write:
n1 = 3k1 + r
n2 = 3k2 + r
where r is the remainder when divided by 3. Then, their difference is
n1 – n2 = (3k1+r) – (3k2+r) = 3(k1-k2)
 n1 – n2 is divisible by 3

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
Example
Example 8. 15 students write a report together. Lan takes 13 mistakes, each of
remaining students takes less than 13 mistakes. Prove that there are always 2
students taking the same number of mistakes.
Proof:

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
Example
Example 9. Set S consists of 6 integers from 1 to 12. Prove that there exists
always 2 nonempty subsets of S such that the sum of elements in these
subsets are equal.

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
Example
Example 10. In a meeting room, there are always 2 people whom are
acquainted with the same number of people present at the meeting.
Solution. Let the number of people in the meeting be n, then the number of
acquaitance of any people in the meeting could only be 0 to n-1.
Obviously, in the room, it could not exists a people having number of
acquaitances is 0 (means do not know anybody) and a people having number
of acquaitances is n-1 (means know everybody in the room). Therefore,
dividing people based on the numeber of acquaitances, one could only divide
n people into n-1 groups.
According to Dirichlet principle, there are at least one group containing at least
2 people, it means one could always find at least 2 people with the same
number of acquaitances
0 1 2 n-1
……….
n groups (n-1) groups
Example
Example 11. In a month of 30 days, a volleyball team plays at least 1 match per day, but does not
play more than 45 matches. Prove that it always could find a period consisting of a certain
number of consecutive days in the month so that the team plays exactly 14 matches during that
period.
Solution: Let aj be the total number of matches played until day j of the team. Then
a1, a2, ..., a30
is the increasing sequence of positive integer numbers and 1  aj  45. Therefore, the sequence
a1+14, a2+14, ..., a30+14
is also the increasing sequence of positive integer numbers and 15  aj +14  59.
• There are 60 positive integer numbers

a1, a2, ..., a30, a1+14, a2+14, ..., a30+14,

where all elements are less than or equal to 59

• Therefore, according to Dirichlet principle, two of these elements are equal. As the numbers
a1, ..., a30 are thirty distinct positive integer numbers and numbers a1+14, ..., a30+14 are thirty
distinct positive integer numbers, thus we could find index i and j such that ai = aj+14. It
means there are exactly 14 matches in the period from day j+1 to day i.
Example
Example 12. Prove that, among n+1 positive integer numbers, each not larger than 2n,
one could always find two numbers, such that one number is divisible by the other
Solution: Let the given numbers be
a1, a2, . . . , an+1 .
Write each number aj of n+1 above numbers in the form:
aj = 2k(j)qj , j = 1, 2, ..., n+1
where k(j) is a non-negative integer, qj is odd.
• Numbers q1, q2, ..., qn+1 are odd integers, each number is not greater than 2n.
• Integers from 1 to 2n, there are only n odd numbers, so according Dirichlet
principle we have: two numbers of (n+1) numbers q1, q2, ..., qn+1 are equal, it means
we could find two indices i and j such that qi = qj = q.
• Then
ai = 2k(i)q, aj = 2k(j)q.
So if k(i) < k(j) then aj is divisible by ai, and if k(i)  k(j) then ai is divisible by aj.
Example
Example 13. On the plane, let’s consider 5 points with integer
coordinates Mi(xi, yi), i=1, 2, ..., 5. Prove that there are always two
points such that the segment connecting them, excluding the two
ends, passes through another point having integer coordinates.
Middle point

Solution. We prove that: always could find 2 points such that the
segment connecting these two points has the middle point with
integer coordinates. According to the parity of two coordinates, given
5 points could be divided into at most 4 groups:
(Even, Even), (Even, Odd), (Odd, Even), (Odd, Odd).

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
Example
• Therefore, according Dirichlet principle, one must find a group of at
least 2 points, such as Mi, Mj. The middle point Gij of the segment
connecting Mi and Mj has the coordinates
Gij = ((xi+xj)/2, (yi+yj)/2).
• As xi and xj as well as yi and yj have the same parity, then the
coordinates of Gij are integers. The claim in the example is proven.
• This claim could be generalized in n-dimensional space: “In n-
dimensional space, let’s consider 2n + 1 points with integer
coordinates. Then one always could find 2 points so that the segment
connecting them, excluding two ends, passes to another point with
integer coordinates”.

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
Example 14
First, we consider some concepts.
Let a1,a2, … an be sequence of real numbers.
• n is the length of the given sequence.
• Subsequence of the given sequence is the sequence with the form ai1, ai2,
…, aim, where 1  i1 < i2 < . . . < im  n
• Sequence is strictly increasing if ai < ai+1 for every i=1,..,n.
• Sequence is strictly decreasing if ai < ai+1 for every i=1,..,n.
Example: Consider the sequence: 1, 5, 6, 2, 3, 9.
– 5, 6, 9 is strictly increasing subsequence
– 6, 3 is strictly decreasing subsequence

NGUYỄN KHÁNH PHƯƠNG

Bộ môn KHMT – ĐHBK HN
Example 14
Theorem: Sequence of n2+1 distinct numbers always contains either strictly increasing
subsequence of length n+1 or strictly decreasing subsequence of length n+1.
Example: Sequence
8, 11, 9, 1, 4, 6, 12, 10, 5, 7
consists of 10 = 32+1 elements, has to contain either strictly increasing subsequence of
length 4 or strictly decreasing subsequence of length 4.
1, 4, 6, 12
1, 4, 6, 7
11, 9, 6, 5
Proof: Assume a1, a2, …, an2+1 is a sequence consisting of n2+1 distinct numbers. Assign to
each element ak an ordered pair (ik,dk), where ik is the length of the longest strictly increasing
subsequence starting from ak and dk is the length of the longest strictly decreasing
subsequence starting from ak.
Example: 8, 11, 9, 1, 4, 6, 12, 10, 5, 7
a2 = 11  (i2, d2) = (2, 4)
a4 = 1  (i4, d4) = ?
• Let assume there does not exist strictly increasing and decreasing subsequence of length
n+1. Thus ik and dk are positive integers  n, where k = 1, 2, ..., n2+1.
Example 14
• As 1  ik, dk  n, following the product rule, there are n2 distinct ordered pairs (ik,dk).
 As there are n2 + 1 ordered pair (ik,dk), so according to the Dirichlet principle, two of them are
identical.
It means there exists 2 elements as and at in the sequence where s < t such that is = it and ds = dt.
We show that it is impossible:
• As elements in the sequence are distinct, so
either as < at or as > at.
– If as < at, then because is = it , we could build increasing subsequence of length it+1
starting from as, by concatenating it to increasing subsequence of length it, starting from
at .
... , as , ..., at , ....
The longest subsequence starting from as has the length at least of it + 1, it measn is > it.
It is contradiction to the assumption is= it.
– Similarly, if as > at, we could show that ds must be greater than dt, and also come to a
contradiction.
Theorem is thus proven.