4

Gödel’s Undecidability Theorem

Author: Stephen F. Andrilli, Mathematical Sciences Department, La Salle

University

Prerequisites: The prerequisites for this chapter are logic, properties of

integers, and techniques of proof. See Chapter 1, and Sections 1.6 and 1.7 in
particular, of Discrete Mathematics and Its Applications.

Introduction
In high-school geometry and in your discrete mathematics course, you encoun-
tered many types of mathematical proof techniques. These included such meth-
ods as proof by contrapositive, proof by contradiction, and proof by induction.
After becoming familiar with these methods, many students think mathemati-
cians have so much “power” at their disposal that, surely, any given mathemat-
ical statement must either already have been proven or disproven.
But this is not the case! For instance, consider the following simple state-
ment.
Goldbach Conjecture: Every even positive integer greater than 2
is the sum of two (not necessarily distinct) primes.
Note that 4 = 2 + 2, 6 = 3 + 3, 8 = 5 + 3, 10 = 5 + 5, 12 = 7 + 5, and so on.
Can we continue, expressing all larger even integers as the sum of two primes?
The answer is not yet known, despite extensive work by many mathematicians.

57
58 Applications of Discrete Mathematics

Goldbach’s Conjecture* is one of hundreds of mathematical “conjectures” whose

truth value we have not been able to determine.
Now, of course, Goldbach’s Conjecture is either true or false. Therefore,
considering the huge battery of proof techniques available, it seems logical for
us to predict with some confidence that eventually mathematicians will develop
a proof that it is true or else find a counterexample to show that it is false.
Until recently, the general attitude of most mathematicians was that every
mathematical conjecture would, after enough effort, be resolved one way or
the other. That is, it would either “become” a theorem or be identified as
a false statement. However, as we will see later, the work of a brilliant 25-
year-old German mathematician named Kurt Gödel** forever shattered this
“belief”, and led to a profound difference in the way we perceive mathematics
as a whole.

Mathematical Systems: Examples from Geometry

To understand Gödel’s work, we must first discuss the nature of general math-
ematical systems. Every statement in mathematics occurs in a context. This
context might not always be stated directly, but is often implicit from previous
statements. For example, a theorem in a high-school geometry can only be
understood fully when we know what the terms in the theorem mean. These
terms may have been defined previously in the text, or they may have been
accepted as undefined terms. Undefined terms are terms that are so fun-
damental that we cannot fully describe them using other more basic terms.
Terms such as “point”,“line”, “equal”, and “between” are typical examples of
undefined terms used in many texts.
Sometimes we cannot prove a theorem without the help of previous the-
orems. We may even need to use some axioms (sometimes called postulates).
Axioms are statements that we accept as true, because they are so fundamen-
tal that we cannot prove them from other more basic results. In high-school
geometry, most axioms are based on common-sense notions. For example, a

* The Goldbach Conjecture appeared in a letter from Christian Goldbach (1690–

1764) to Leonhard Euler in 1742. Goldbach was a historian and professor of math-
ematics at the Imperial Academy of St. Petersburg (now Leningrad). He tutored
Czar Peter II, and later worked in the Russian Ministry of Foreign Affairs. Among
his research areas were number theory, infinite series, and differential equations.
** Kurt Gödel was born in Brno, Czechoslovakia in 1906. During the 1930s he
taught at the University of Vienna and did research at the Institute for Advanced
Study in Princeton, New Jersey. Gödel spent time in a sanatorium for treatment of
depression in 1934 and again in 1936. In 1940, to escape Nazi fascism, he emigrated
to the United States, and continued his work at the Institute for Advanced Study. He
died in Princeton in 1978.
 Chapter 4 Gödel’s Undecidability Theorem 59

typical axiom in many geometry texts is: Every two distinct points lie on a
unique straight line.
In any high-school geometry text, the undefined terms, definitions, ax-
ioms, and theorems, taken together, form a mathematical system. Most
high-school geometry courses introduce students to the system of Euclidean
geometry, so named because most of the results are essentially those handed
down from classical times in The Elements of Euclid.* Within such a mathe-
matical system, we can make new definitions and prove additional theorems.

Example 1 After accepting the axiom “Every two distinct points lie on a
unique straight line”, and the undefined terms listed above, we can introduce
a new term: “midpoint”. The midpoint between two given points is defined as
the (unique) point on the line connecting the given points that divides the line
segment between the points into two equal parts.
Additionally, once we have defined terms like “right triangle” and “hy-
potenuse”, we can prove theorems such as the familiar Pythagorean Theorem:
In any right triangle, the square of the hypotenuse is equal to the sum of the
squares of the other two sides.

In this way, we can achieve the development of an entire mathematical

system beginning with a few simple building blocks!

Non-Euclidean Geometries
One of the axioms that Euclid adopted in The Elements is equivalent to the
following statement, commonly known as the Parallel Postulate:
Parallel Postulate: If l is a line in a given plane, and P is a point
of the plane not on l, then in this plane there is one and only one line
going through P that is parallel to l.
This seems like a perfectly reasonable assumption — so reasonable, in fact, that
even into the nineteenth century many mathematicians tried to prove that it
was really a consequence of earlier axioms and definitions. Their efforts were
ultimately unsuccessful. They next tried to deny this postulate to determine
whether the resulting mathematical system contained a logical contradiction.
But instead they obtained a set of new “strange” results!
We know now that it is possible to create several valid non-Euclidean ge-
ometries in which the Parallel Postulate is contradicted in different ways. For

* Euclid, who lived about 300 B.C., taught mathematics at Alexandria during the
reign of the Egyptian Pharaoh Ptolemy Soter, and was most probably the founder
of the Alexandrian School of mathematics. His classic work The Elements contains
much of the basic geometrical knowledge accumulated up to that time.
60 Applications of Discrete Mathematics

example, we could decide to include the following axiom in our mathematical

system instead of the Parallel Postulate:
Multi-Parallel Postulate: If l is a line in a given plane and P is a
point in the plane not on l, then in the plane there is more than one
line going through P parallel to l.
Adopting this axiom leads to a type of geometry known as hyperbolic ge-
ometry, which was discovered independently before 1830 by Nikolai Lobachev-
sky* and János Bolyai**. We can now prove theorems that are consequences
of this new postulate. For instance, in hyperbolic geometry we can prove that,
with l and P as in the statement of the Multi-Parallel Postulate, there are an
infinite number of lines through P parallel to l. We can also show that the sum
of the angles of any triangle is less than 180◦ .
Another type of non-Euclidean geometry is created when we replace the
Parallel Postulate with:
No-Parallel Postulate: If l is a line in a given plane and P is a
point in the plane not on l, then in the plane there is no line going
through P parallel to l.
Such a geometry is called elliptic geometry, and was first created by
Bernhard Riemann†. In this geometry, every pair of distinct lines meets in a
point. Also, the angles of any triangle sum up to more than 180◦.

* Nikolai Lobachevsky (1792–1856) taught at the University of Kazan, Russia,

beginning in 1816. He later became Rector of the University. Lobachevsky published
his first results on hyperbolic geometry in 1829. Another of his research interests was
developing methods for approximating roots of algebraic equations.
** János Bolyai (1802–1860) studied at the Royal College of Engineering in Vienna
and served in the army engineering corps. He was an expert swordsman and a superb
violinist! Bolyai’s father Farkas spent his lifetime to no avail trying to prove that the
Parallel Postulate was a consequence of Euclid’s earlier postulates. János continued
this work, but realized that it might be impossible to prove. He began constructing a
non-Euclidean geometry, a first draft of which was written in 1823. Bolyai’s work went
unnoticed at the time, and was largely overshadowed by Lobachevsky’s publication.
His contribution to non-Euclidean geometry was only recognized posthumously. Aside
from his work in geometry, Bolyai also broke new ground in the theory of complex
variables.
† Georg Friedrich Bernhard Riemann (1826–1866) was one of the most impor-
tant mathematicians of the last century. He taught at Göttingen (University), and
aside from his ground-breaking work in geometry, he developed many new concepts
in complex and real analysis, and what is now topology and differential geometry. His
discovery of elliptic geometry dates from about 1854 and was first published in 1867.
Albert Einstein later based much of his theory of relativity on the geometrical ideas
of Riemann.
 Chapter 4 Gödel’s Undecidability Theorem 61

Consistency of Mathematical Systems

As a result of these new observations in non-Euclidean geometry, mathemati-
cians began to realize that abstract mathematical systems (with undefined and
defined terms, axioms and theorems) could be created without any natural or
obvious association with “reality” as we ordinarily perceive it. But this leads to
a fundamental difficulty: once we choose an arbitrary set of axioms, how do we
know that the system “makes sense”? That is, how can we be sure that none
of the axioms are contradicted by the others?
It is important that any mathematical system we work in be consistent,
that is, free from self-contradiction. It can be shown that Euclidean geometry
is inwardly consistent, as are both of the non-Euclidean geometries discussed
earlier. However, not all sets of axioms lead to consistent systems.

Example 2 Consider this set of axioms with undefined terms “point” and
“line”:
(i) There are exactly three points in the system.
(ii) There are fewer than three lines in the system.
(iii) Every pair of points lies on exactly one line.
(iv) Not all points lie on the same line.
Notice that “line” as used here is a general term. A “line” does not necessarily
have to agree with our “common-sense” notion of an object that consists of an
infinite number of points, but may only contain a finite set of points. These
axioms are not consistent, since the only way that axioms (i), (ii), and (iii) can
be satisfied simultaneously is to have all three points lying on the same line.
But this contradicts axiom (iv).

One way to be sure that none of the axioms in a given mathematical system
contradict the others is to exhibit a model for the system that satisfies all of
the axioms simultaneously.

Example 3 Consider the following collection of axioms, with undefined

terms “point” and “line”:
(i) At least two lines exist.
(ii) Every line contains at least two points.
(iii) Given any two distinct points, there is a line containing the first
point but not the second point.
Notice also that the last axiom is “symmetric”, because if we reverse the order
of the points, axiom (iii) insists there is a line containing the original second
point that does not contain the original first point.
62 Applications of Discrete Mathematics

Figure 1. A model of four points and four lines.

The model of four “points” and four “lines” in Figure 1 shows that these
axioms are consistent.
For example, given points A and B, line 1 goes through A but not B, while
line 2 goes through B but not A. Exercise 3 asks you to find another model for
this system using fewer than four points.

Mathematicians are not interested in a system unless it is logically consis-

tent. However, for a system that involves a large (possibly infinite) number of
objects, a “model” of the system is often too difficult to construct. Such mod-
els would be extremely complicated, especially if we were trying to represent a
system powerful enough to contain an entire branch of mathematics.
This can be seen even in a branch of mathematics as familiar as arithmetic.
For instance, suppose we want to include an axiom that states that for every
integer, there is a next largest integer. If we attempt to construct this system,
then for every integer x that we list, we must list the integer y immediately
following it as well. But then for this new integer y, we must list the integer
z immediately following y, and so on. Thus, we would have to display an
entire infinite sequence of integers, which is, of course, not physically possible.
Mathematicians realized that a new approach was needed.
In 1910, two mathematicians, Alfred N. Whitehead* and Bertrand Rus-
sell** published a monumental treatise, Principia Mathematica. In this work,

* Alfred North Whitehead (1861–1947) was a philosopher and mathematician who

held various academic positions in London from 1911 to 1924. From 1910–1913,
he worked with Russell on Principia Mathematica. In 1924, he accepted a faculty
post at Harvard University, and in the late 1920s he wrote a number of treatises on
metaphysics.
** Bertrand Russell (1872–1970) was a philosopher and logician, who published nu-
merous books on logic and the theory of knowledge. He was a student of Whitehead’s
before they collaborated on Principia Mathematica. Russell was active in many social
causes, and was an advocate of pacifism (as early as World War I) and of nuclear
disarmament.
 Chapter 4 Gödel’s Undecidability Theorem 63

they took a different approach to the consistency question. They attempted

to derive all of arithmetic from principles of symbolic logic, which had been
rigorously developed in the nineteenth century. (Symbolic logic includes such
rules of inference as “modus ponens”, “disjunctive syllogism”, etc.) In this
way, Whitehead and Russell could establish the consistency of arithmetic as a
consequence of the basic laws of symbolic logic.

Resolving Logical Paradoxes

Whitehead and Russell’s effort took on added importance because, at the turn of
the century, mathematicians became more and more concerned about resolving
paradoxes appearing in mathematical systems. Paradoxes are statements that
are apparently self-contradictory. These paradoxes worried mathematicians be-
cause they seemed to imply that mathematics itself is ultimately inconsistent!
A typical paradox is Russell’s Paradox. To understand this, let a set be
described as a collection of objects (leaving “object” as an undefined mathemat-
ical term). Next, consider the set X of all sets having more than one element.
Now X certainly contains more than one set. Thinking of these sets as elements
of X, we see that X itself has more than one element. Hence, X ∈ X. Thus,
it is possible to find sets that are elements of themselves. This leads us to the
following.

Russell’s Paradox: Let S be the set of all sets that are not elements
of themselves: i.e., S = {X | X ∈ / X}. Then either S ∈ S, or S ∈ / S.
But either possibility leads to a contradiction. For, if S ∈ S, then by
definition of S, S ∈/ S. On the other hand, if S ∈/ S, then by definition
of S, S ∈ S.

In their Principia Mathematica, Whitehead and Russell devised methods

to avoid Russell’s Paradox and other similar paradoxes. They introduced the
idea of “set types”. They began with a fundamental collection of objects as
above. These objects were considered to be of “type 0”. Any set of objects of
type 0 is considered to be a set of “type 1”. Any set of objects of types 0 or 1
is considered to be a set of “type 2”. In general, all objects in a set of “type
n” must be of type less than n. Finally, only those collections of objects that
have a “type” will be admitted into the system as sets. Such a scheme enabled
Whitehead and Russell to avoid situations like Russell’s Paradox where a set
could be an element of itself.
In effect, what Whitehead and Russell did was to avoid mathematical para-
doxes and “shift” the problem of the consistency of arithmetic to the consistency
of a few basic logical principles. If these logical principles standing alone were
consistent, then any mathematical system (in particular, arithmetic) derived
solely from them would also be consistent. With this new method for checking
64 Applications of Discrete Mathematics

consistency, many mathematicians felt it would only be a matter of time before

other branches of mathematics would be proven consistent as well, and then,
within those systems, eventually all statements could be proven or disproven.

Gödel’s Undecidability Theorem

In 1931, Kurt Gödel published a paper “On Formally Undecidable Proposi-
tions of Principia Mathematica and Related Systems” in which he proved the
following theorem.

Theorem 1 Gödel’s Undecidability Theorem Any mathematical sys-

tem containing all the theorems of arithmetic is an incomplete system. (That
is, there is a statement within this system that is true, but can never be proven
true.)

In other words, any system powerful enough to contain all of the rules of
arithmetic, (e.g., the commutative and associative laws of addition and multi-
plication), must contain a mathematical statement that is true (has no coun-
terexample), but for which no proof can ever be found using all of the results
available in that system!
Another way of expressing Gödel’s Undecidability Theorem is to say that,
given a certain mathematical statement in this system, we may never learn
whether it is true or false — and even if it is true, we may not ever have the
resources to prove that it is true from within the system. Because of Gödel’s
Undecidability Theorem, we can no longer confidently predict that all open
conjectures, such as Goldbach’s Conjecture, will eventually be resolved one
way or the other.
In his paper, Gödel produced a statement (let us call it G, for Gödel) that
is true in any system similar to the system developed in Principia Mathematica
but can never be proven true within that system. A natural question is: How
do we know that G is true if we can’t prove G is true? To explain this, we need
to give some idea of Gödel’s actual proof.
Gödel’s actual paper is extremely complicated. The Undecidability Theo-
rem is preceded by the establishment of several primitive objects called “signs”,
several variables of various “types” (as in Principia Mathematica), 45 definitions
of functions (or relations), many other definitions, five axioms, one lemma, and
seven theorems! Nevertheless, we can get the central thrust of Gödel’s argument
by studying the method he used for labeling items in Principia Mathematica.
His technique is often referred to as Gödel numbering.
 Chapter 4 Gödel’s Undecidability Theorem 65

Gödel Numbering
Gödel assigned a distinct positive integer, a Gödel number, to each symbol,
formula, and proof in the system. He assigned the first few odd integers to
these seven basic symbols:
symbol meaning Gödel number
0 zero 1
f successor 3
¬ not 5
∨ or 7
∀ for all 9
( left parenthesis 11
) right parenthesis 13
Other useful symbols such as ∧ (and), ∃ (there exists), and = (equals)
can be “built” by using an appropriate combination of the seven basic symbols
above. However, for the sake of simplicity, we will depart from Gödel’s actual
numbering, and assign Gödel numbers to these symbols as follows:
symbol meaning Gödel number
∧ and 15
∃ there exists 17
= equals 19
Numerical variables (such as x, y, and z) that represent natural numbers
are assigned larger odd values as their Gödel numbers — we assign 21 to x, and
23 to y.
Natural numbers after 0 can be represented in this system as follows:

f0 , f f 0, f f f 0, etc.
    
one two three
Formulas of this type that contain a sequence of symbols are assigned Gödel
numbers by creating a product of powers of successive primes 2, 3, 5, 7, 11, . . ..

Example 4 Find the Gödel number for the sequence f f f 0.

Solution: The symbols in the sequence

fff0

have Gödel numbers 3, 3, 3, 1, respectively, and therefore we assign the Gödel

number
23 33 53 71
to the entire sequence f f f 0.
66 Applications of Discrete Mathematics

Example 5 Find the Gödel number of the formula ∃x(x = f 0).

Solution: The symbols in the formula

∃x(x = f 0)

have Gödel numbers 17, 21, 11, 21, 19, 3, 1, 13, respectively. Hence we would
assign the Gödel number

217 321 511 721 1119 133 171 1913

to the formula ∃x(x = f 0).

By the way, the formula ∃x(x = f 0) in the last example asserts that there
is some natural number that is the successor of zero. Since ‘1’ is the successor of
zero, we would consider this to be a true proposition. With the method of this
example, we can assign to every logical proposition (and hence every theorem)
an associated Gödel number.
We can generalize this even further by associating a Gödel number with a
sequence of propositions. If there are k propositions in a sequence with Gödel
numbers n1 , n2 , . . . , nk , then we assign the Gödel number 2n1 3n2 5n3 . . . pnk k
(where pk is the kth prime in the list 2, 3, 5, 7, 11, . . .) to this sequence.

Example 6 Find the Gödel number of the sequence

⎧ ⎫
⎨ x = f0 ⎬
S = ¬(f 0 = 0) .
⎩ ⎭
¬(x = 0)

Solution: We first find the Gödel number of each individual proposition in

the sequence. Now,
the Gödel number of x = f 0 is k1 = 221 319 53 71 ,
the Gödel number of ¬(f 0 = 0) is k2 = 25 311 53 71 1119 131 1713 , and
the Gödel number of ¬(x = 0) is k3 = 25 311 521 719 111 1313 .
Now, the Gödel numbers of these individual propositions are placed as expo-
nents on the primes 2, 3, 5, respectively, to obtain the Gödel number

2k1 3k2 5k3

(an incredibly large number) for the entire sequence S.

Note that, in Example 6, if the first two statements in the sequence S have
already been proven (or accepted as axioms), then the third statement follows
 Chapter 4 Gödel’s Undecidability Theorem 67

logically from the first two. Hence, if x = f 0, we can consider the sequence
S to be a proof of the statement ¬(x = 0). In this way, every proof in this
mathematical system can be assigned a Gödel number.
In this manner, every basic symbol, variable, sequence of symbols and vari-
ables (e.g., proposition), and sequence of propositions (e.g., proof) that can be
expressed in a system such as Principia Mathematica has its own unique Gödel
number. Conversely, given any Gödel number, we can reconstruct the unique
sequence of symbols corresponding to it. (This follows from the Fundamental
Theorem of Arithmetic, which asserts that each positive integer has a unique
expression as a product of primes, where the primes are listed in increasing
order.) In other words, there is a one-to-one correspondence between valid se-
quences of symbols and valid Gödel numbers in our Principia Mathematica -like
system.

A Few Additional Definitions

Suppose we consider all the possible formulas in our system having exactly one
variable (say, x), and put them in increasing order according to Gödel number.
Let R(n) represent the nth such Gödel formula.
Gödel became interested in what would happen if the particular sequence
of the form f f f · · · f 0 representing the natural number n were substituted in
place of the variable x in the nth Gödel formula R(n). This produces a new
formula, which we label as

SUBST[n, R(n)],

where “SUBST” stands for “substitution”. (Gödel actually used [R(n); n] in

his paper to represent this new formula.) This new formula may or may not be
provable within our mathematical system. That is, there may or may not be
some sequence of previously established propositions that logically proves the
new formula SUBST[n, R(n)].
Suppose, for a particular n, the Gödel formula SUBST[n, R(n)] is provable.
We express this by writing

PR(SUBST[n, R(n)]),

where “PR” stands for “provable”. (Gödel expressed this as Bew[R(n); n].)

The Heart of Gödel’s Argument

Consider the set Z of all integers n for which PR(SUBST[n, R(n)]) is true —
that is, for which there exists a finite sequence of propositions to prove the
statement obtained when the sequence f f f · · · f 0 for n is substituted for the
variable x in the nth Gödel formula. (In his paper, Gödel actually used the
68 Applications of Discrete Mathematics

complement of this set as the set Z.) Now, a given integer n is either in Z or
it is not. Consider the formula ¬(x ∈ Z) (i.e., it is not true that x ∈ Z). It
is not immediately obvious that this formula involving x has a Gödel number,
but Gödel showed that there is a way of expressing the formula x ∈ Z in terms
of more basic symbols in the system. Therefore, let us suppose that ¬(x ∈ Z)
has a Gödel number, and that it is the qth Gödel formula (in the ordering of
single-variable Gödel formulas discussed earlier). Then, ¬(x ∈ Z) is R(q).
Finally, let us consider SUBST[q, R(q)] — the new formula obtained by
putting the appropriate sequence of symbols of the form f f f · · · f 0 for q into
the formula R(q). We will call this new formula G.
Now, either G is true or ¬G is true. However,

Theorem 2 Neither G nor ¬G is a provable statement.

Proof: We give a proof by contradiction. Suppose first that G is a provable
statement. Then, PR(SUBST[q, R(q)]) is true — and so by definition of Z,
we have q ∈ Z. But upon actually substituting q into R(q), we obtain the
statement ¬(q ∈ Z), which is supposed to be provable. Since we cannot have
both q ∈ Z and ¬(q ∈ Z) provable in our system, we have a contradiction.
On the other hand, suppose ¬G is a provable statement. Then G is cer-
tainly not true, and therefore not provable. Hence, PR(SUBST[q, R(q)]) is not
true, and so by definition of Z, it follows that q ∈
/ Z. That is, q ∈
/ Z is a prov-
able statement. But, since G is not provable, substituting q into R(q) gives us
the unprovable statement ¬(q ∈ Z). Since we cannot have q ∈ / Z both provable
and unprovable, we have a contradiction.

We have seen that either G or ¬G is true but that neither of these two
statements is provable. Thus, in our Principia Mathematica -like system, there
is at least one statement that is true, but is not provable!

Applications of Gödel’s Undecidability Theorem

Gödel’s Undecidability Theorem has profound implications. It tells us that even
such a fundamental branch of mathematics as arithmetic has its limitations. It
is not always possible to find a proof for every true statement within that system.
Thus, we may never be able to establish whether certain conjectures within the
system are true. You might wonder whether we could enlarge our system to a
more powerful one in the hopes of finding proofs for certain statements. But,
we could then apply Gödel’s Undecidability Theorem to this larger system to
show that it also has limitations.
 Chapter 4 Gödel’s Undecidability Theorem 69

Rudy Rucker writes in [4] that “Gödel once told the philosopher of science
Hao Wang that the original inspiration for his . . . theorem was his realization
that ‘truth’ has no finite description.” That is, no mathematical system with
a finite number of axioms such as we have described here can contain a full
depiction of reality. No mathematical system like Principia Mathematica can
be both consistent and complete.
In addition to the Undecidability Theorem, Gödel also demonstrated in his
paper that a system like Principia Mathematica is not even powerful enough to
establish its own consistency! That is, the consistency of the system cannot be
proven using only the principles available within the system itself.
In recent years, the implications of Gödel’s Undecidability Theorem in com-
puter science have been studied widely, especially in the area of artificial intelli-
gence. If we imagine giving to a computer an initial set of formulas (axioms) and
a set of logical principles (operations) for calculating new formulas (theorems),
then Gödel’s Undecidability Theorem implies that, no matter how powerful the
computer, there are always some true statements that the computer will never
be able to derive. Although computers have been useful to mathematicians in
proving computationally difficult results, Gödel’s Undecidability Theorem de-
stroys the myth many people have that any difficult mathematical problem will
eventually be resolved by a powerful enough computer.
Computer scientists have drawn an even closer parallel between mathe-
matical systems and computers. The initial state (register contents, data) of a
computer is analogous to the initial set of formulas (axioms) in a mathematical
system. Similarly, the logical operations (via hardware and software) that a
computer is allowed to perform are analogous to the principles of logic permit-
ted in the mathematical system. Extending this analogy, the mathematician
(and pioneer computer scientist) Alan Turing* extended Gödel’s results to prove
that there is no general algorithm that can always correctly predict whether a
randomly selected computer program will run or not.
In 1961, the British philosopher J. Anthony Lucas tried to use Gödel’s
Undecidability Theorem to show that a machine will never be able to “think”

* Alan Turing (1912–1954) was a British mathematician and logician, who received
his Ph.D. from Princeton in 1938. In his landmark 1937 paper “On Computable
Numbers”, he proposed the notion of a general computer (later known as a Turing
machine), that could execute finite mathematical algorithms, and proved there exist
mathematical problems that can not be solved by such a machine. Turing also worked
on the enormous problems of breaking secret codes during World War II with the
Foreign Office of the British Department of Communications. He was the subject of a
recent Broadway play (starring Derek Jacobi) by Hugh Whitemore entitled “Breaking
the Code”. After the war, Turing worked at the National Physical Lab at Teddington
on the design of a large (automatic) computer, and became Deputy Director of the
Computing Lab at the University of Manchester in 1949. Turing died from a dose of
potassium cyanide poisoning, self-administered, but possibly accidental.
70 Applications of Discrete Mathematics

in quite the same manner as a human, and hence that the achievement of
an artificial intelligence is impossible. However, Lucas’ argument is refuted
in Douglas Hofstadter’s book [2]. In this book, Hofstadter also considers a
number of other related artificial-intelligence questions, such as whether the
human mind is essentially a computer system (i.e., a mathematical system). If
so, are there some truths that the unaided human mind will never fathom?

Suggested Readings

1. M. Gardner, “Douglas R. Hofstadter’s ‘Gödel, Escher, Bach’ ”, Scientific

American, 1979, pp. 16–24. (A quick tour through Hofstadter’s book [2].)
2. D. Hofstadter, Gödel, Escher, Bach: An Eternal Golden Braid, 20th An-
niversary Edition, Basic Books, New York, 1999. (Pulitzer Prize winning
tour de force that relates the work of Gödel to the work of the artist M.C.
Escher and the composer Johann Sebastian Bach.)
3. E. Nagel and J. Newman, Gödel’s Proof, New York University Press, New
York, 1958. (The classic reference on Gödel’s Undecidability Theorem.)
4. R. Rucker, Mind Tools: The Five Levels of Mathematical Reality, Hough-
ton-Mifflin, Boston, 1987. (See especially the section on Gödel’s Undecid-
ability Theorem, pp. 218–226.)
5. R. Smullyan, Forever Undecided: A Puzzle Guide to Gödel, Knopf, New
York, 1987. (Understanding the concepts behind Gödel’s Undecidability
Theorem by solving successively more difficult puzzles.)
6. J. van Heijenoort, From Frege to Gödel, A Source Book in Mathemat-
ical Logic, 1879–1931, Harvard University Press, Cambridge, MA, 2002.
(Contains a complete English translation of Gödel’s 1931 paper, as well as
Gödel’s abstract of the paper, and a related note that he wrote about the
paper.)

Exercises

1. Verify that Goldbach’s Conjecture is true for all the even integers from 14
to 50.
2. Another unproven conjecture in number theory is the following: Let f :
N −→ N be defined by
 Chapter 4 Gödel’s Undecidability Theorem 71


n/2 n even
f (n) =
3n + 1 n odd;
then, for every n, there is an integer i such that f i (n) = 1. Verify that this
conjecture is true for n = 22 and n = 23.
3. Exhibit a model for the axioms in Example 3 containing fewer than four
points.

In Exercises 4–6, find a finite model for the system having the given set of
axioms. Let “point” and “line” be undefined terms. (Remember, a “line” does
not necessarily have to be straight or contain an infinite number of points.)
4. (i) There are exactly four points in the system.
(ii) Not all points are on the same line.
(iii) Through any three points there is exactly one line.
5. (i) There are at least two lines in the system.
(ii) Each line contains at least two points.
(iii) Every pair of points is on a line.
(iv) No point is on more than two lines.
6. (i) There is at least one line and a point not on that line.
(ii) Each line contains exactly three points.
(iii) Each pair of points lies on exactly one line.
Hint: Use seven points and seven lines.
7. In Exercise 5, can you find a model containing exactly 4 points?

In Exercises 8 and 9, show that the given set of axioms is inconsistent.

8. (i) There are exactly four points in the system.
(ii) Each pair of points lies on exactly one line.
(iii) Every line contains exactly three points.
9. (i) There is at least one line in the system.
(ii) Every two points lie on exactly one line.
(iii) Every point lies on exactly two distinct lines.
(iv) Every line contains exactly three points.
10. Explain why the following statement is a paradox: “This sentence is false.”
11. Suppose there are two types of people living on a certain island: truth-
tellers (who never lie) and liars (who never tell the truth).
a) An inhabitant of the island says, “I am lying.” Explain the paradox.
 b) An inhabitant of the island says, “You will never know that I am
telling the truth.” Is this a paradox?
 72 Applications of Discrete Mathematics

12. Using the Gödel numbering scheme established earlier in this chapter, give
the Gödel number for each of the following statements:
a) x = f f 0.
b) ¬((x = 0) ∨ (y = 0)).
c) ∀x∃y(¬(f x = y)).
d) ∀x((x = f 0) ∨ ¬(x = f 0)).
13. Give the Gödel number for each of the following proofs (i.e., sequence of
statements):⎧ ⎫
⎨ x = f0 ⎬
a) S = y = fx
⎩ ⎭
y = ff0
⎧ ⎫
⎨ ∃y(x = f y) ⎬
b) T = ∀y(¬(0 = f y))
⎩ ⎭
¬(x = 0)
14. In each case, state the proposition that has the given Gödel number.
a) 25 311 521 719 1123 1313 .
b) 29 321 517 723 1111 1321 1719 193 2323 2913 .
c) 211 321 519 71 1113 137 175 1911 2323 2919 311 3713 .
15. In each case, state the sequence of propositions that has the given Gödel
number.
a) 2k1 3k2 5k3 , where k1 = 221 319 53 73 111 , k2 = 25 311 523 719 1121 1313 , and
k3 = 25 311 523 719 113 133 171 1913 .
b) 2k1 3k2 5k3 , where k1 = 29 323 517 721 115 1311 1721 1919 2323 2913 , k2 =
2 3 5 7 , and k3 = 217 321 55 711 1121 1319 173 191 2313 .
23 19 3 1

16. Suppose we enlarged our Principia Mathematica -like system to include

Gödel’s statement G as an axiom. If the new mathematical system is
consistent (that is, if G is true), what does Gödel’s Undecidability Theorem
tell us about the new system?

Computer Projects

1. Write a program that takes as input an even positive integer greater than 2
and writes it as a sum of two primes (Goldbach’s Conjecture).
2. Write a program that implements the function defined in Exercise 2.
3. Write a program that takes a list of primes and gives as output the string of
symbols that constitutes the corresponding Gödel statement or recognizes
that the list of primes is unacceptable.