Logical Inference and Mathematical Proof

CSE 191, Class Note 03:

Logical Inference and Mathematical Proof
Computer Sci & Eng Dept
SUNY Buffalo

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 1 / 66

Need for inference

Why do we study propositional and predicate logic?

We want to use them to solve problems.
To solve a problem by using logic, we often need to start from some
“premises” and obtain a certain “conclusion using inference rules.
Example: Computer scientists often need to verify the correctness
of a program.
One possible approach is to prove the program is correct.
So one can start from the program and the semantics of the used
programming language (i.e., the premise), and use logic inference
to obtain a conclusion that the program does the right job.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 3 / 66
What is a inference rule?

Definition:
P1
P2
Q

is a inference rule if (P1 ∧ P2 ) → Q is a tautology.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 4 / 66

Implication rule
The first (and simplest) rule is:
Modus Ponens (MP) Rule:
P
P→Q
Q

This rule is called Modus Ponens (MP). Intuitively, if we have the

condition of an implication, then we can obtain its consequence.

Example:
P means: “there is a storm.”
P→Q means: “if there is a storm, then the office is closed.”
Q means: “the office is closed.”

Exercise: Show [P ∧ (P → Q)] → Q ≡ T.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 5 / 66
Another implication rule
Recall that P → Q ≡ ¬ P ∨ Q ≡ Q ∨ ¬ P ≡ ¬ Q → ¬ P.
The MP rule just studied above tells us that:
¬Q
¬Q→¬P
¬P
If we replace the ¬ Q → ¬ P in the above with the logically
equivalent proposition P → Q, then we get another implication
rule:

Modus Tonens (MT) Rule:

¬Q
P→Q
¬P

Exercise: Show [¬Q ∧ (P → Q)] → ¬P ≡ T.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 6 / 66

Logical equivalence vs. inference

By using inference rules, we can “prove” the conclusion follows from
the premises. In inference, we can always replace a logic formula with
another one that is logically equivalent, just as we have seen for the
implication rule.
Example:
Suppose we have: P → (Q → R) and Q ∧ ¬ R. Use inference to show
¬ P.

First, we note Q ∧ ¬ R ≡ ¬(¬ Q ∨ R) ≡ ¬(Q → R).

So we have the following inference:
(1) P → (Q → R) Premise
(2) Q∧¬R Premise
(3) ¬(Q → R) Logically equivalent to (2)
(4) ¬P Applying the second implication rule
(Modus Tonens) to (1) and (3)

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 7 / 66
Yet another implication rule

Hypothetical Syllogism (HS)

P→Q
Q→R
P→R

Intuitively, if P implies Q and Q implies R, then we can get that P

implies R.

Example:
P→Q means “if there is a storm, then the office is closed.”
Q→R means “if the office is closed, then I don’t go to work.”
P→R means “if there is a storm, then I don’t go to work.”

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 8 / 66

Conjunction and Simplification Rules

Conjunction rule
P
Q
P∧Q

Intuitively, this means when you have P and Q both being true, then
P ∧ Q is also true.

Simplification Rule
P∧Q
P

Intuitively, this means when you have P ∧ Q being true, clearly P is also
true.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 9 / 66
Disjunction rules

Disjunctive Syllogism (DS)

P∨Q
¬P
Q

Addition Rule:
P
P∨Q

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 10 / 66

Third disjunction rule

Resolution Rule
P∨Q
¬P∨R
Q∨R

This rule plays an important role in AI systems.

Intuitively, it means: if P implies R and ¬ P implies Q (why? Where
do we get these implications?), then we must have either Q or R.
Clearly, this is true since one of P and ¬ P must be true.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 11 / 66
Other rules

We have:
P∨Q
P→R
Q→S
R∨S

Intuitively it means we can do inference in each of two cases (P or

Q) independently.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 12 / 66

Inference Rules

Table: Rules of Inference

Rule of Inference Tautology Name

p
p→q (p ∧ (p → q)) → q Modus ponens (MP)
q
¬q
p→q (¬q ∧ (p → q)) → ¬p Modus tonens (MT)
¬p
p→q
q→r ((p → q) ∧ (q → r)) → (p → r) Hypothetical syllogism (HS)
p→r
p∨q
¬p ((p ∨ q) ∧ ¬p) → q Disjunctive syllogism (DS)
q
p
p → (p ∨ q) Addition
p∨q
p∧q
(p ∧ q) → p Simplification
p
p
q ((p) ∧ (q)) → (p ∧ q) Conjunction
p∧q
p∨q
¬p ∨ r ((p ∨ q) ∧ (¬p ∨ r)) → (q ∨ r) Resolution
q∨r

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 13 / 66
Valid Arguments

A mathematical proof is always like:

“If q1 and q2 . . . and qk are true, then q is true.”
The propositions q1 , . . . , qk are called the premises.
The proposition q is called the conclusion.
The mathematical proof is really to show that (q1 ∧ q2 . . . ∧ qk ) → q
is a tautology.
To do this, we can either:
Directly prove (q1 ∧ q2 . . . ∧ qk ) → q ≡ T by using logic equivalence
rules, (which will be very long); or
Present a valid argument, by using logic inference rules, defined
in the following slide.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 14 / 66

How to build a valid argument?

A valid argument is a sequence of propositions P1 , P2 , . . . , Pn such that:
Pn is the conclusion q.
Each Pi (1 ≤ i ≤ n) is:
either a premise (namely some qj );
or a proposition that can be obtained from previous propositions by
using a rule of inference;
or a proposition that is logically equivalent to a previous proposition.
Or a tautology.

Remark:
If P1 , P2 , . . . , Pn , Q is a valid argument, then we can always show:

[q1 ∧ q2 ∧ · · · ∧ qk ] → q ≡ T (1)

by using logic equivalence rules.

So a valid argument is just a shorter way to prove (1) is a tautology by
using logic equivalence rules.
c
Xin He (University at Buffalo) CSE 191 Discrete Structures 15 / 66
Inference example
Consider the following propositions:
p: It’s sunny this afternoon.
q: It’s colder than yesterday.
r: We will go swimming.
s: We will take a canoe trip.
t: We will be home by sunset.

Example:
Premises:

a. “It’s not sunny and it’s colder than yesterday” ¬p∧q

b. “We will go swimming only if it’s sunny.” r→p
c. “If we don’t go swimming then we will take canoe trip.” ¬r→s
d. “If we take a canoe trip, then we will be home by sunset.” s→t

Conclusion: “We will be home by sunset.” t.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 16 / 66

Inference example

(1) ¬p∧q Premise

(2) ¬p Simplification rule using (1)
(3) r→p Premise
(4) ¬r MT using (2) (3)
(5) ¬r→s Premise
(6) s MP using (4) (5)
(7) s→t Premise
(8) t MP using (6) (7)

This is a valid argument showing that from the premises (a), (b), (c)
and (d), we can prove the conclusion t.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 17 / 66
Inference example

Example:
Suppose P → Q; ¬ P → R; Q → S. Prove that ¬ R → S.

(1) P→Q Premise

(2) ¬P∨Q Logically equivalent to (1)
(3) ¬P→R Premise
(4) P∨R Logically equivalent to (3)
(5) Q∨R Apply resolution rule to (2)(4)
(6) ¬R→Q Logically equivalent to (5)
(7) Q→S Premise
(8) ¬R→S Apply HS rule to (6)(7)

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 18 / 66

A more concrete example

Example: Suppose:
(1) If it is Saturday today, then we play soccer or basketball.
(2) If the soccer field is occupied, we dont play soccer.
(3) It is Saturday today, and the soccer field is occupied.
Prove: “we play basketball or volleyball”.

First we formalize the problem:

P: It is Saturday today.
Q: We play soccer.
R: We play basketball.
S: The soccer field is occupied.
T: We play volleyball.
Premise: P → (Q ∨ R), S → ¬ Q, P, S
Need to prove: R ∨ T.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 19 / 66
A more concrete example

(1) P → (Q ∨ R) Premise
(2) P Premise
(3) Q∨R Apply MP rule to (1)(2)
(4) S→¬Q Premise
(5) S Premise
(6) ¬Q Apply MP rule to (4)(5)
(7) R Apply DS rule to (3)(6)
(8) R∨T Apply Addition rule to (7)

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 20 / 66

Solving a murder case

The following is a murder case solved by Sherlock Holmes, in “A Study

in Scarlet” (a detective mystery novel by Sir Arthur Conan Doyle).

Quote from “A Study in Scarlet”

“And now we come to the great question as to the reason
why. Robbery has not been the object of the murder, for
nothing was taken. Was it politics, then, or was it a woman?
That is the question which confronted me. I was inclined from
the first to the latter supposition. Political assassins are only
too glad to do their work and fly. This murder had, on the
contrary, been done most deliberately, and the perpetrator
has left his tracks all over the room, showing he had been
there all the time.”
From these, Sherlock Holmes concluded: “It was a woman”.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 21 / 66
Solving a murder case

Known premises:
1 If it’s a robbery, something would have been taken.
2 Nothing was taken.
3 If it’s not a robbery, it must be politics or a woman.
4 It it’s politics, the assassin would have left immediately.
5 If assassin left tracks all over the room, he cannot have left
immediately.
6 The assassin left tracks all over the room.
Show the conclusion: “It was a woman”.
We will discuss this example in class.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 22 / 66

Inference with quantifiers

Many inferences in Math and CS involve quantifiers.

Example 1:
All computer science majors must take CSE 191.
CSE 191 students study discrete structures.
So, all computer science majors must study discrete structures.

Example 2: The definition of limit in Calculus:

limx→a f (x) = b if and only if

∀  > 0 ∃ δ > 0 ∀ x (|x − a| ≤ δ) → (|f (x) − b| < )

Question:
How do you show limx→a f (x) 6= b?

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 24 / 66
Universal quantification rules

Consider the following two formulas:

∀ xP(x).
P(c) for an arbitrary c.
Starting from either of them we can obtain the other.

Example:
Everybody has a nose.
C has a nose (where C can be anybody).

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 25 / 66

Universal quantification rules

Universal instantiation rule:

∀ x P(x)
P(c) for an arbitrary c

Universal generalization rule:

P(c) for an arbitrary c
∀ x P(x)

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 26 / 66
Existential quantification rules

Consider the following two formulas:

∃ x P(x).
P(c) for some c.
Starting from either of them we can obtain the other.

Example:
There is a student living in Amherst.
John lives in Amherst.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 27 / 66

Existential quantification rules

Existential instantiation rule:

∃ x P(x)
P(c) for some c

Existential generalization rule:

P(c) for some c
∃ x P(x)

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 28 / 66
Rules of Inference for Quantified Statements

Table: Rules of Inference for Quantified Statements

Rule of Inference Name

∀xP(x)
Universal Instantiation
p(c) for an arbitrary element c
P(c) for an arbitrary element c
Universal generalization
∀xP(x)
∃xP(x)
Existential Instantiation
p(c) for some element c
P(c) for some element c
Existential generalization
∃xP(x)

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 29 / 66

Inference example
Example:
Premises:
1 “A student in this class has not read the book”.
2 “Everyone in this class passed the first exam”.
Conclusion: “Someone who passed the first exam has not read the
book”.

C(x): “x is in this class.

B(x): “x has read the book”.
P(x): “x has passed the first exam”.
Then:
1 ∃x(C(x) ∧ ¬B(x)).
2 ∀x(C(x) → P(x)).
Conclusion: ∃x(P(x) ∧ ¬B(x)).
We will show the inference in class.
c
Xin He (University at Buffalo) CSE 191 Discrete Structures 30 / 66
An arithmetic example

Example: Suppose:
all natural numbers are integers;
there exists a natural number;
Prove that there exists an integer.

We can formalize this problem as follows. (Let the universe of

discourse be all real numbers.)

N(x): x is a natural number.

I(x): x is an integer.
Premise: ∀ x (N(x) → I(x)), ∃ x N(x)
Need to prove: ∃ x I(x)

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 31 / 66

An arithmetic example

(1) ∃ x N(x) Premise

(2) N(c) Apply existential instantiation rule to (1)
(3) ∀ x (N(x) → I(x)) Premise
(4) N(c) → I(c) Apply universal instantiation rule to (3)
(5) I(c) Apply MP rule to (2)(4)
(6) ∃ x I(x) Apply existential generalization rule to (5)

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 32 / 66
From Inference to Proof

In fact, the process of logical inference is also the process of

giving a formal proof.
In mathematics, we need to do a lot of proofs.
However, formal proofs are too long, since in each step we can only
apply a simple inference rule.
Formal proofs are also too hard to follow since we can be easily
buried in details and thus miss bigger pictures.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 34 / 66

(Informal) mathematical proof

A mathematical proof is usually “informal”. (Compared to logical

inferences we just studied. But still much more formal than everyday
language.)
More than one rule may be used in a step.
Steps may be skipped.
Axioms may be assumed.
Rules for inference may not be explicitly stated.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 35 / 66
Some terminology

Theorem: statement that can be shown true.

Proposition: less important theorem.
Lemma: less important theorem used to prove other theorems.
Corollary: theorem that trivially follows another theorem.
Conjecture: statement that is proposed to be true, but has not
been proved.
Axiom: statement assumed to be true (i.e., true statement that
does not need a proof).

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 36 / 66

Convention on universal quantifier

Many mathematical theorems are about a property of elements in

a domain, and hence need universal quantifiers.
However, many such universal quantifiers are omitted.

Example:
If a > b, then a − b > 0.

This actually means, for all real numbers a and b, if a > b, then
a − b > 0.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 37 / 66
Proof method: direct proof

Definition
A direct proof for p → q starts by assuming p and finishes by
establishing q.

In the proof, we can use axioms, previously proven theorems, and

inference rules.

Example of direct proof:

Prove that if n is an odd integer, then n2 is also odd.

Proof: Since n is an odd integer, there exists integer k such that

n = 2k + 1. Hence:
n2 = (2k + 1)2 = 4k2 + 4k + 1 = 2 · (2k2 + 2k) + 1.
Since k is an integer, 2k2 + 2k is also an integer. So 2 · (2k2 + 2k) is
even. Therefore, n2 is an odd integer.
c
Xin He (University at Buffalo) CSE 191 Discrete Structures 38 / 66

Proof by contraposition

Another method for proof is proof by contraposition.

Note that p → q is logically equivalent to ¬ q → ¬ p.

Definition
A proof by contraposition for p → q is actually a direct proof for
¬ q → ¬ p.
It starts by assuming ¬ q, and finishes by establishing ¬ p.
In the proof, we can also use axioms, previously proven theorems,
and inference rules.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 39 / 66
Example for proof by contraposition

Example:
Prove that if n is an integer and 3n + 2 is odd, then n is odd.

Proof: We prove it by contraposition. Suppose n is not odd. Then it is

¯
even.
So there exists an integer k such that n = 2k.
Hence: 3n + 2 = 3(2k) + 2 = 6k + 2 = 2(3k + 1).
Since k is an integer, 3k + 1 is also an integer. Therefore, 3n + 2 is
even, i.e., 3n + 2 is not odd.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 40 / 66

Proof by contradiction

Yet another method for proof is proof by contradiction.

Note that p is logically equivalent to ¬ p → (r ∧ ¬ r).

Definition
A proof by contradiction for p is actually a direct proof for
¬ p → (r ∧ ¬ r).
It starts by assuming ¬ p and finishes by establishing both r and ¬ r.
In the proof, we can also use axioms, previously proven theorems, and
inference rules.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 41 / 66
Example for proof by contradiction (1)

Example:
Prove that there is no positive integer n such that n3 + 1 = 100.

Proof: We prove it by contradiction.

Suppose that there is a positive integer n such that n3 + 1 = 100.
If n ≤ 4, then n3 + 1 ≤ 43 + 1 = 65 < 100. Impossible.
If n ≥ 5, then n3 + 1 ≥ 53 + 1 = 126 > 100. Still impossible.
Contradiction.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 42 / 66

Example for proof by contradiction (2)

√
Example: Prove that 2 is not a rational number.

Proof (Hinted by Aristotle, 384-322 BC): We prove it by contradiction.

√
Suppose that r = 2 is a rational number. Then there exist
integers a and b such that r = a/b.
Further, we assume that a and b have no common divisors. (If
they have common divisors, divide both a and b by their greatest
common divisors.)
Hence, 2 = r2 = (a/b)2 = a2 /b2 .
So we get that a2 = 2 · b2 , which is an even number.
Therefore, a is even. Thus, there exists integer c such that a = 2c.
This implies that (2c)2 = a2 = 2 · b2 , i.e., 2c2 = b2 .
Hence, b2 is even, which means b is also even.
Since a and b are both even, they have a common divisor 2.
Contradiction.
c
Xin He (University at Buffalo) CSE 191 Discrete Structures 43 / 66
Example for proof by contradiction (3)

Definition
A prime number is a positive integer whose only divisors are 1 and
itself.

The first few prime numbers: 2, 3, 5, 7, 11, 13, 17 ...

At the beginning, the prime numbers are dense (i.e. there are
many of them). For example, there are 168 prime numbers
between 1 and 1000.
When the number gets bigger, the prime numbers are sparse (i.e.
there are few of them).
How many prime numbers are there? Finite? or infinite?

Theorem:
There are infinitely many prime numbers.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 44 / 66

Example for proof by contradiction (3)

Proof: (Euclid 325 -265 BC).
We prove this by contradiction. Assume there are only finitely many
primes: p1 , p2 , . . . , pn . Consider the number Q = p1 · p2 · . . . · pn + 1.
We ask: Is Q a prime number?
Because Q > pi for all 1 ≤ i ≤ n and p1 , . . . , pn are ALL prime
numbers, Q IS NOT a prime.
If Q is not a prime, it must have a prime factor.
So one of pi (1 ≤ i ≤ n) must be a factor of Q.
But Q divided by each pi has remainder 1.
So none of pi (1 ≤ i ≤ n) is a divisor of Q. Hence Q IS a prime.
A contradiction.

It is the first proof in the book “Proofs from THE BOOK”, where
THE BOOK refers to the imagined collection of the most elegant
proofs that the famous mathematician Paul Erdös claimed is
maintained by the God.
c
Xin He (University at Buffalo) CSE 191 Discrete Structures 45 / 66
Proof by cases

Note that (p1 ∨ p2 ∨ . . . ∨ pn ) → q is logically equivalent to

(p1 → q) ∧ (p2 → q) ∧ . . . ∧ (pn → q).

Definition:
A proof by cases for (p1 ∨ p2 ∨ . . . ∨ pn ) → q is actually a direct proof for
(p1 → q) ∧ (p2 → q) ∧ . . . ∧ (pn → q).
First, it lists n cases. In the ith case, it starts by assuming pi and
finishes by establishing q.
In the proof, we can also use axioms, previously proven theorems,
and inference rules.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 46 / 66

Example for proof by cases (1)

Example:
Prove that n + 100 > 3n if n is a positive integer with 1 ≤ n ≤ 3.

Proof: We prove it by cases.

Case 1: n = 1. Then, n + 100 = 101 > 3 = 3n .
Case 2: n = 2. Then, n + 100 = 102 > 9 = 3n .
Case 3: n = 3. Then, n + 100 = 103 > 27 = 3n .

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 47 / 66
Example for proof by cases (2)

Example:
Prove that if n is an integer, then n2 ≥ n.

Proof: We prove it by cases.

Case 1: n = 0. Then, n2 = 0 = n.
Case 2: n ≥ 1. Then, n > 0. So we can multiply both sides of inequality
by n, and get n2 ≥ n.
Case 3: n ≤ −1. Hence, n2 ≥ 0 > −1 ≥ n.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 48 / 66

Notion of Without Loss of Generality

In some “proof by cases”, there are too many cases.

Some of the cases are similar to each other.
To shorten the proof, we only give detailed proofs of some cases,
but omit the proof of other cases.
When we do this, we say Without Loss of Generality (often
abbreviated as WLOG).

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 49 / 66
Example of Without Loss of Generality

Example:
Prove that if x and y are integers and both x + y and x · y are even, then
both x and y are even.

Proof: We will use: proof by contraposition, the notion of WLOG, and

proof by cases.
First suppose that x and y are not both even. That is we assume either
x is odd, or y is odd (or both).
WLOG, we assume x is odd.
Case 1: y is even. Then x + y = odd + even = odd. This contradicts the
assumption that x + y is even.
Case 2: y is odd. Then x · y = odd · odd = odd. This contradicts the
assumption that x · y is even.
In both cases, we get a contradiction. Hence both x and y must be
even.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 50 / 66

Example of Without Loss of Generality

In this example, when we say WLOG, we are saying the other

case (that y is odd) can be proved by using similar method.
When using WLOG, you must make sure the omitted cases are
really similar to the proved cases.
Otherwise, your proof is incomplete and could be wrong.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 51 / 66
Constructive proof

Note that ∃ x P(x) is equivalent to “P(c) is true for some c”.

Definition
A constructive proof for ∃ x P(x) finds c such that P(c) is true. Hence,
we can conclude that ∃ x P(x) is also true.

Example:
Prove that there is a positive integer that can be written as the sum of
cubes of positive integers in two different ways.

Proof: We construct an example for such a positive integer:

1729 = 103 + 93 = 123 + 13

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 52 / 66

Non-constructive proof

To prove ∃ x P(x), we can just provide a specific c such that P(c) is

true.
But sometimes we cannot do this.
It is still possible to argue such element exists, even though we
cannot pin point the specific c.
This is called non-constructive proof.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 53 / 66
Non-Constructive proof example

Example:
Prove that there exist irrational numbers a and b such that ab is rational.
√ √ √2
Proof: We already know 2 is irrational. Consider the number 2 .
√ √2 √ √
Case 1: If 2 is rational. Then let a = 2 and b = 2. Then ab is
rational.
√ √2 √ √2 √
Case 2: If 2 √ is not rational. Then let a = 2 and b = 2.
√ 2 √ √ 2
Then ab = ( 2 ) 2 = 2 = 2 is rational.
√ √2
Note that, in this proof, we do not claim/know if 2 is a rational
number or not.
But in either case, we prove the proposition.
√ √2
Actually, 2 is irrational. But the proof requires high powered
mathematical theory, far beyond our reach.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 54 / 66

Uniqueness proof

In mathematics we often need to prove there exists a unique x

such that P(x) is true.
We often prove the following two things (which, when combined
together, are logically equivalent to the original statement):
There exists x such that P(x) is true. (Namely ∃ x P(x)).
If P(x) and P(y) are both true, then x = y. (Namely
∀ x ∀ y (P(x) ∧ P(y)) → (x = y).
In some books, we use the symbol ∃! for “there exists a unique ....”.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 55 / 66
Example for uniqueness proof

Example:
Prove that if a and b are real numbers and a is not 0, then there is a
unique real number r such that ar + b = 0.
(In other words: There is a unique solution for the equation ax + b = 0.

Proof:
First, we show that there exists such an r. Let r = −b/a. Then:

ar + b = a(−b/a) + b = −b + b = 0
So this r satisfies the condition.
Second, suppose that r and r0 both satisfy the condition.
Then, r − r0 = ((ar + b) − (ar0 + b))/a = (0 − 0)/a = 0, which
implies that r = r0 .

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 56 / 66

Logic Dilemmas
Definition
A statement p is a dilemma if we cannot prove p ≡ T nor p ≡ F.

Example: Barber’s dilemma

Premise: “The barber gives haircut to a person if and only if that
person doesn’t give haircut to himself”.
Statement p: “The barber gives haircut to himself.”
Question: Is p true or false?

If p is true: Then the premises states that he does not give haircut
to himself. This implies p is false.
If p is false: Then the premises states that he does give haircut to
himself. This implies p is true.
So we cannot say if p is true or false. And the statement is a
dilemma.
c
Xin He (University at Buffalo) CSE 191 Discrete Structures 57 / 66
Gödel’s Incompleteness Theorem.
Definition:
A “logic inference system” consists of a collection of “rules of
inferences” and a collection of “axioms”.

Definition:
A logic inference” system is “consistent” if for any proposition p, we can
prove either p ≡ T, or p ≡ F, (or neither), but not both.

An inconsistent logic inference system is useless, because we

can “prove” p ≡ T for any proposition p in such a system.

Definition:
A logic inference system is “complete” if for any proposition p, we can
prove either p ≡ T or p ≡ F, (or both).

The logic inference system we have studied is incomplete

because there are propositions that we cannot prove to be T or F.
c
Xin He (University at Buffalo) CSE 191 Discrete Structures 58 / 66

Gödel’s Incompleteness Theorem.

Question:
It is unsettling that our logic inference system (the foundation of all
mathematics, computer science ....) is INCOMPLETE.
Can we make it “stronger” by adding a few new “rules of
inferences”?
If these new rules can be derived from existing rules, we don’t
gain anything.
If the new rules are too “strong”, they will make our inference
system “inconsistent”.
It would be ideal if we have a logic inference system that is both
consistent and complete.
Can this be done?

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 59 / 66
Gödel’s Incompleteness Theorem.

Unfortunately:

Gödel’s Incompleteness Theorem (1931)

Any consistent logic inference system MUST BE incomplete.

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 60 / 66

Hilbert’s Problems

In 1900, David Hilbert (one of the greatest mathematician of his

time) presented 23 unsolved problems to the mathematicians of
the 20th century.
All these problems are extremely hard.
Some of Hilbert’s problems have been solved (either positively or
negatively). Some are not.

Hilbert’s Second Problem:

Prove that “The logic inference system for arithmetic is complete”.

Gödel’s Incompleteness Theorem is the negative answer to

Hilbert’s Second problem.
We will mention Hilbert’s first problem later (which remains
unsolved).

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 61 / 66
Hilbert’s Second Problem
What is Hilbert 2nd problem really asking?
Axioms of Arithmetic
Basic objects: The set of natural numbers: 1, 2, 3, 4 . . .
Operations: +, −, ×, ÷
Axioms:
if a = b then b = a. i.e ∀a∀b(a = b) → (b = a).
if a = b and b = c then a = c i.e ∀a∀b∀c(a = b) ∧ (b = c) → (a = c).
a + 0 = a i.e ∀a(a + 0 = a).
a + b = b + a i.e ∀a∀b(a + b = b + a).
(a + b) + c = a + (b + c) i.e ∀a∀b∀c(a + b) + c = a + (b + c).
..
.
plus all basic arithmetic laws you learned before the third grade.

These axioms plus our table of inference rules constitute ”the logic
inference system for arithmetic”.
c
Xin He (University at Buffalo) CSE 191 Discrete Structures 62 / 66

Logic Inference System for Arithmetic

This inference system looks really elementary (after all, you

learned everything in it before the 3rd grade!)
There cannot be hard problems in this system, right?
Completely wrong!

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 63 / 66
Some hard arithmetic problems
There are many integer solutions for the equation: x2 + y2 = z2 .
For examples: 32 + 42 = 52 , 52 + 122 = 132 . . .
Are there positive integer solutions for the equation: x3 + y3 = z3 ?

Fermat’s Last Theorem:

For any integer n ≥ 3, there is no positive integer solution x, y, z for the
equation
xn + yn = zn
Equivalently, we want to show the following proposition is T (the domain for
n, x, y, z is all positive integers.)

∀(n ≥ 3)¬∃x ∃y ∃z (xn + yn = zn )

This problem had been open for more than 350 years. It was proved by
Andrew Wildes (a Math professor at Princeton University) in 1995.
c
Xin He (University at Buffalo) CSE 191 Discrete Structures 64 / 66

Some hard arithmetic problems

Goldbach’s Conjecture:
Every even integer n ≥ 2 is the sum of two prime numbers.

Example: 4=2+2; 6=3+3; 8=3+5; . . ., 20=3+17; 22 = 3+19= 11+11;

. . ..
Let E(x) be the proposition x is even; P(y) be the proposition y is a
prime number. Then Goldbach’s Conjecture is to prove the
following proposition is T:

∀x(E(x) → (∃y ∃z (P(y) ∧ P(z) ∧ (x = y + z)))

It has been verified that this conjecture is true for n up to 1.6 · 1018 .
British publisher Tony Faber offered a $1,000,000 prize if a proof
was submitted before April 2002. The prize was not claimed.
It remains unsolved today.
c
Xin He (University at Buffalo) CSE 191 Discrete Structures 65 / 66
Hilbert’s 2nd Problem

Hilbert imagined that all propositions (involving only integers and

arithmetic operations) can be proved either T or F.
These propositions would include the propositions for Fermat’s
Last Theorem and Goldbach’s Conjecture (and many other
unsolved hard problems).
Gödel’s Incompleteness Theorem says: “this is a mission
impossible!”

c
Xin He (University at Buffalo) CSE 191 Discrete Structures 66 / 66