Professional Documents
Culture Documents
Methods of Proof
Methods of Proof
Methods of Proof
Michael P. Allocca
Formal proof is a concept that takes on new meaning when one achieves a
higher level of mathematical sophistication. It is the cornerstone of all branches
of mathematics and the standard by which we accept new ideas. It is not un-
common for students in precollegiate math classes to understand mathematical
proof to be a string of simple algebraic justifications. For example, you may be
quite familiar with arguments such as the following:
x3 − 8 = 0
so x3 = 8
√3
so x = 8
so x = 2
02 = 0, which is even.
22 = 4, which is even.
42 = 16, which is even.
62 = 36, which is even.
(−2)2 = 4, which is even.
1
You may recall that a positive integer (also called a natural number) is
prime if its only divisors are 1 and itself (otherwise, it is called composite). You
might also read this proposition with justifiable skepticism. Adopting a similar
strategy as before, consider a few quick computations:
This may be surprising at first, and it is reasonable to now think that the
proposition may actually be true. In fact, if you compute n2 − 3n + 43 replacing
n with the next thirty positive integers, you will end up with thirty more
prime numbers! This appears to be overwhelming evidence that the proposition
is true. However, if you continued to test examples, you would find a surprising
outcome when n = 42:
1.1 Groundwork
One of the major challenges in understanding how a formal system of proof-
writing works is establishing a starting point based on reasonable assumptions.
We will define a proof that depends on this starting point as follows.
In other words, axioms are claims that we are unable to formally prove, but
we may communally accept as an evident “fact” and use them to justify further
claims. Loosely speaking, we take an “axiomatic approach” to proof-writing in
the sense that a core set of facts are accepted to be true, and they form building
blocks to proving theorems. A rather famous execution of this paradigm exists
in the study of Euclidean geometry, which has logical roots that trace back to
five axioms (or “postulates”):
2
1. A straight line can be drawn to join any two points.
2. Any straight line segment can be extended continuously in a straight line.
3. Given any straight line segment, a circle can be constructed using one
endpoint as its center and the segment itself as the radius.
4. All right angles are congruent.
5. Given any straight line and a point not on it, there is only one
straight line that passes through the point and never intersects
the line.
Every one of these axioms may be considered to be very reasonable, if not
a bit “obvious.” However, you will not be able to prove any of them using only
the other four. The boldfaced axiom is more commonly known as the “parallel
postulate,” which has an equivalent assertion that two parallel lines will never
intersect. For thousands of years, attempts were made to prove the parallel
postulate based on the other four, to no avail. The entire system of geometry
that you likely experienced in high school is based on the assumption that all
five of these are true. Yet, in reality, and in the context of contemporary work
such as Einstein’s special relativity, it was discovered that the parallel postulate
is, in fact, not always true. This creates an important logical distinction. If
you remove the parallel postulate from your system of axioms, an entirely new
system of “Non-Euclidean geometry” is formed with surprising consequences.
For example, in Non-Euclidean geometry, the sum of all angles in a triangle
need not be 180 degrees!
As we move forward in our study of proof methods, for our purposes we will
assume that a set of relevant axioms has been established and all subsequent
results can eventually be derived from these assumptions.
We now establish four main rules of proof writing. First, you may recall a
powerful form of logical deduction from an earlier chapter:
Modus Ponens
p ∧ (p ⇒ q) ⇒ q
3
You will leverage Modus Ponens quite often as the primary form of deduc-
tive reasoning in a proof, and it will take on multiple forms, such as formal
conversation with your reader or algebraic logic.
The next rule is very reasonable, but worth singling out since it is commonly
invoked in any formal proof.
Given Rule
You can state assumptions, axioms, proved theorems, or established facts
at any time.
The next rule will be important when using different techniques later in this
chapter.
Replacement Rule
You can replace a statement at any time with any of its logical equiva-
lencies.
Example 4 It is not true that John or Jane will inherit the money. So neither
John, nor Jane will inherit the money.
¬(p ∨ q)
so ¬p ∧ ¬q
Tautology Rule
As a new proof writer, you might want to be careful in how you execute
the Tautology Rule. While precision is valuable in a proof, so is concision.
Tautologies should be relevant and advance your argument with a specific goal
in mind.
Example 5
n∈N
so n is odd or n is even
4
In this example, your reader knows that n is a natural number. In the
universe of natural numbers, n’s “parity” can only be even or odd, never both.
So the second statement follows the logical form p ∨ ¬p, which is a tautology.
This is a common strategy in proof writing, which usually leads to arguing
different cases. In this example, the writer would likely list “Case 1: n is odd”
with appropriate justifications afterwards, followed by “Case 2: n is even” in
their proof.
With these rules of proof-writing established, we move on to examining the
major methods of proof.
Direct Proof of p ⇒ q
Assume p is true.
···
Conclude q is true.
Therefore, p ⇒ q is true.
5
Proof:
A few comments are in order. Line (1) illustrates the standard way to begin
a direct proof, as we assumed that the antecedent is true. Line (2) illustrates the
replacement rule in effect. However, this was not a replacement of line (1) with a
sophisticated logical equivalency, but rather a simple one - the formal definition
of an even integer. This is an important part of proof-writing and illustrative of
a successful strategy in “definition-chasing.” The same can be said for line (3),
as steps in an algebraic argument are logical replacements of the expressions in
preceding steps. Finally, line (4) can be seen as another replacement (using the
definition of “even”), or also a type of execution of Modus Ponens to establish
the truth of the claim’s consequent. Specifically, the underlying logic of line (4)
is as follows:
Proof:
a = even = 2k
a2 = 4k2 = even
While this may resemble your scrap work when strategizing how to prove
the claim, it is an example of very poor proof prose. In deciphering the first
6
line, your reader will be saying “a equals even,” which has no mathematical
meaning. Further, the variable k is introduced without telling the reader what
it represents. The second line has similar issues, and even further glosses over
an important detail of explaining why 4k2 has to be an even number. An
even number must be written as a multiple of 2, which has not fully been
demonstrated by writing a2 as 4k2 .
Let’s try another direct proof.
Example 7 Show that if a + b is even and b + c is even, then a + c is even.
A few comments are in order. Take note that we again use the replacement
rule by simply reciting definitions. This is usually the best place to start if you
are unsure of how to begin a proof. Ask yourself questions like “What does it
mean if this number is even?” Also, after you suppose the antecedent is true
and establish any information relevant to it, you should feel free to “stick your
head above the water” and remind yourself (and your reader) what your goal
is. For example, after the second sentence in this proof, we could have written
“We aim to show that a + c = 2m for some integer, m” with no ill effects on
the proof’s validity or style.
Finally, an effective writing style often entails rotating words that connect
each line of logical reasoning. A proof that reads “So.. then.. therefore.. hence..
thus.. ” is far more legible than one that overuses the same words, like “then..
then.. then.. then.. then.. ”.
It is also worth noting that the underlying logic ((p ∧ q) ⇒ r) in the claim in
Example 7 is also equivalent to p ⇒ (q ⇒ r), we could have followed the direct
proof method for this pair of conditionals and arrived at a similar argument.
Specifically, in the former, we would start off by supposing p and q. In the latter,
we would first suppose p, then we would suppose q. This would essentially lead
to the same proof.
Let’s work through some more examples of direct proofs.
Example 8 Suppose a and b are integers. Prove that a − b is an odd integer
if one of a and b is even and the other one is odd.
You should note that the antecedent here is not stated first, but it is still
attached to the word ‘If.’ Thus, the claim in Example 8 is worded “r, if p and
q, and the underlying logic is again of the form (p ∧ q) ⇒ r.
7
a − b = 2k1 − (2k2 + 1)
= 2k1 − 2k2 − 1
= 2k1 − 2k2 − 1 + 2 − 2
= 2(k1 − k2 − 1) + 1
You might note that the preceding proof began in a surprising way. To
satisfy the assumption that one of the two integers is even, we simply supposed
a was the even one and b was the odd one. It’s important to circle back and
address the situation in which a is the odd one and b is the even one. Since
the justifications are the same in this case, only with different variables, your
reader does not need to see it duplicated. In situations like this, we sometimes
simply write “Without loss of generality, assume a is even and b is odd.” This
communicates to your reader that you are maintaining concision with your proof
while acknowledging that an identical argument can cover the second case.
ab = (2k1 )(2k2 + 1)
= 4k1 k2 + 2k1
= 2(2k1 k2 + k1 )
a2 = (2k + 1)2
= (2k + 1)(2k + 1)
= 4k2 + 2k + 2k + 1
= 4k2 + 4k + 1
= 2(2k2 + 2k) + 1
Hence a2 is odd, since it has the form 2s + 1, where s = 2k2 + 2k, which is
an integer since k ∈ Z.
8
Just as parity arguments are common in proofs, divisibility is an invaluable
definition to a proof writer.
It is worth noting that this definition does not change if a and b, and k are
natural numbers.
The following example is considered to be a “Lemma” to many other proofs.
That is, it is a theorem that is usually proven separately and used axiomatically
in a larger proof.
Example 11 Prove that if a|b and b|c, then a|c.
Note that the claim did not establish a certain universe, but the context
(using the definition of “divides”) makes it clear that a, b and c should not
be assumed to be real numbers, or complex numbers. Since there is still an
ambiguity regarding whether they should be natural numbers or integers, we
will proceed by proving the result for integers, which is a more general (and
stronger) result.
Proof: Suppose a|b and b|c. Then b = ak and c = bl for some integers
k and l. So c = (ak)l = a(kl). Since kl ∈ Z, a|c.
B Exercises
Exercise 2 Prove directly that if gcd(a, 65) = 1 then a12 ≡ x(mod65).
C Exercises
Exercise 4 Prove directly that the polynomial x2 + 1 does not have a root in R.
9
1.3 A Note on Quantifiers
One of the most important skills to develop in reading and writing proofs is
the ability to ascertain unspecified context in some statements. Quantifiers, in
particular, are often hiding in plain sight, even when phrases such as “for all”
and “there exists” are absent. You may recall from a previous chapter that we
typically work with three quantifiers for predicates:
∀x − for all (Universal quantifier) x
∃x − there exists (Existential quantifier) x
∃!x − there exists a unique x
10
Hence, a+c is even, since k+l−y is an integer. Since a, b, and c were arbitrary,
it follows that if a + b is even and b + c is even, then a + c is even.
Proofs of this nature are often called “constructive proofs” in the sense that
you are constructing a particular x that makes p true. In practice this is usually
accomplished by choosing a particular value for x from the universe in which
you are working, but sometimes it suffices to just argue that one exists, even if
you haven’t chosen its value. This is illustrated by two separate proofs of the
following claim.
Example 12 There is a real root of f(x) = x2 − 1 between x = 0 and x = 2.
Note that the phrase “there is” tells us to read this claim as “∃ x ∈ (0, 2) 3
f(x) = 0,” where the symbol 3 is shorthand for “such that.” The context of the
claim also implies that we are working over the universe of real numbers, R.
Proof: Let f(x) = x2 − 1. Now let x = 1. f(1) = 12 − 1 = 0, so
∃ x ∈ (0, 2) 3 f(x) = 0.
Proof: Let f(x) = x2 − 1. Note that f(0) = −1 < 0, and f(2) = 3 > 0.
Hence, by the intermediate value theorem, since f is a continuous function on
(0, 2), ∃ x ∈ (0, 2) 3 f(x) = 0.
The first proof shows existence by example, illustrating the only acceptable
type of “proof by example” that is logically sound (note that our earlier attempts
11
to “prove by example” were erroneously attempted on universally quantified
claims). The second proof leverages a well-known theorem in Calculus; It shows
existence without illustrating a specific example. This is sometimes necessary
in more advanced proofs, but generally if you can construct a specific example,
then you might make a more clear argument to your reader.
Finally, the third quantifier, ∃!, establishes a truth set with only one element.
To prove a statement with this type of quantifier, we may extend our strategy
for proving existentially quantified statements by adding an extra condition:
uniqueness.
If we parse this claim, we will find that there are actually two quantifiers.
Logically, it reads
∀x ∈ R, ∃! y ∈ R 3 x + y = 0
We may address proofs with multiple quantifiers by adhering to our rules,
addressing them in order of left to right. The first quantifier is universal, so
we must choose an arbitrary value. Then, based on this arbitrarily chosen
value, x, we must show there exists a unique y that satisfies the claim.
Proof: Let x ∈ R.
(Existence) Let y = −x. Then x + y = x + (−x) = 0, which is the additive
identity. Therefore, x has an additive inverse.
(Uniqueness) Suppose y and z are additive inverses of x. Then x + y = 0 and
x + z = 0. So x + y = x + z, which implies that (x + y) − (x + z) = 0. Thus,
y − z = 0, which means that y = z.
12
Exercises for Section 1.3
A Exercises
Exercise 5 Rewrite Examples 8, 9, and 10 as claims with quantifiers and re-
prove them.
B Exercises
Exercise 6 Prove that for every positive real number, there exists another pos-
itive real number less than it.
C Exercises
Exercise 8 Prove that, in the universe of differentiable functions, there exists
a unique function, f(x), whose first derivative is x3 + 2 and has the property
that f(0) = 3.
Proof of p ⇒ q by Contrapositive
Assume ¬q is true.
···
Conclude ¬p is true.
Hence, ¬q ⇒ ¬p is true.
Therefore p ⇒ q is true.
13
Proof:
√ Let a ∈ Z. Suppose a2 is even. Then a2 = 2k for some k ∈ Z. So
a= 2k, which should be even.
a2 = (2k + 1)2
= 4k4 + 4k + 1
= 2(2k2 + 2k) + 1
This claim has the logical form ¬p ⇒ (¬q ∧ ¬r). With so many negations
in this conditional, it seems natural to instead prove its contrapositive, ¬(¬q ∧
¬r) ⇒ ¬¬p, or equivalently, (q ∨ r) ⇒ p.
ab = (7s)b
= 7sb
= 7(sb)
14
Case 2: Assume 7 | b. Then b = 7t, for some integer t. then
ab = a(7t)
= 7at
= 7(at)
You may have noticed that in this proof, we assume (q ∨ r) is true, yet we
only explored two cases. You may recall that in order for (q ∨ r) to be true, q
must be true (case 1), r must be true (case 2), or both q AND r must be true.
We did not, however, need to explore this third case, because the arguments
presented in case 1 and case 2 do not change based on whether 7 divides the
other integer.
By now you may realize that the contrapositive method really shines when
the negation of statements are simpler to work with. Other times, however, its
necessity can be quite subtle.
Example 16 Let a ∈ Z. Prove that if a2 + 2a − 3 is even then a is odd.
a2 + 2a − 3 = (2k)2 + 2(2k) − 3
= 4k2 + 4k − 3
= 4k2 + 4k − 4 + 1
= 2(2k2 + 2k − 2) + 1
15
2. Let a ∈ Z. Prove that if 5a + 1 is even, then a is odd.
3. Let a ∈ Z. Prove that if a2 + 3a + 7 is even then a is odd.
4. Let a ∈ Z. Prove that if a3 + 3a2 − 9 is odd then a is even.
B Exercises
Exercise 11 Let a, b ∈ Z. Prove that if a + b ≥ 21 then a ≥ 10 or b ≥ 10.
Exercise 12 Let x, y ∈ R. Prove that if y3 + yx2 ≥ x3 + xy2 then y ≥ x.
Proof of Proof of p ⇔ q
1. Prove p ⇒ q
2. Prove q ⇒ p
Proof: Let a ∈ Z.
[⇒] Suppose a is even. Then a = 2k for some integer, k. So a2 = (2k)2 =
4k2 = 2(2k2 ), which is an even integer since 2k2 is an integer. Since a was
arbitrary, it follows that if a is an even integer, then a2 is even.
[⇐] Suppose a is odd. Then a = 2k + 1 for some k ∈ Z. So a2 = (2k + 1)2 =
4k4 + 4k + 1 = 2(2k2 + 2k) + 1. Since 2(2k2 + 2k) ∈ Z, it follows that a2 is odd.
Thus, by contraposition, if a2 is even, then a is even.
16
Example 18 Let a be an integer. Show that a is even ⇔ a + 1 is odd.
Proof: Let a ∈ Z.
[⇒] Suppose a is even. Then a = 2k, k ∈ Z. So a + 1 = 2k + 1. Since k ∈ Z,
a + 1 is odd.
[⇐] Suppose a + 1 is odd. Then a + 1 = 2k + 1, k ∈ Z. So a = 2k, by right
cancellation. Since k ∈ Z, a is even.
Although this is a short proof, you may feel as though there’s some redun-
dancy in effect. If we examine the underlying logic, the proof really looks like
this:
Proof: Let a ∈ Z.
The second condition is, in fact, a reversal of the first. That really means
that each conditional could more simply be justified as a biconditional. Thus,
we may condense the proof as follows.
Proof: Let a ∈ Z.
a is even ⇔ a = 2k, k ∈ Z
⇔ a + 1 = 2k + 1, k ∈ Z
⇔ a is odd
Not all biconditional proofs can be argued with a single string of bicondi-
tionals, but when it is possible it is usually preferable as it adds much-valued
concision to a mathematical argument.
Exercise 14 Show that a positive integer, n, is even if, and only if, 4 | n2 .
17
1.6 Proofs by Contradiction
Regardless of mathematical background or familiarity with formal logic, many
people still rely on the efficacy of contradictions in every day reasoning and
informal arguments. For example, consider returning from a walk where you
were unexpectedly caught in the rain. Let’s suppose a friend of yours does
not believe that it is raining, noting that it was just sunny a few minutes ago.
You might ask this friend to suppose that it’s not raining outside. Then, your
clothes would be dry. This is an obvious “contradiction,” as the friend can
plainly see that your clothes are soaking wet. Therefore, your original claim
that it is raining outside must be true.
The underlying logic in this scenario reflects a powerful proof technique,
highlighted by inviting your audience to “suppose not.” Surprisingly, the rigor
of this proof method lies behind the logical equivalency of any statement, p, to
the statement (¬p) ⇒ (q ∧ ¬q), regardless of what q is. This can be verified by
truth table, as explored in an earlier chapter. More significantly, it means that
the proof of p can be accomplished by a direct proof of yet another conditional
statement!
Proof of p by Contradiction
Note that in this proof, q turned out to be the statement “a is the greatest
even integer”, and we showed that, as a consequence of supposing the claim, p,
is false, a is the greatest even integer and a is NOT the greatest even integer
(q ∧ ¬q), which is a contradiction.
Proofs by contradiction can be quite intuitive when you know in your gut
that something is true and make it a habit of asking yourself “what are the
consequences of this statement NOT being true?” If those consequences are an
obvious contradiction, then you have formed a proof.
18
Stylistically, many writers prefer to use the symbol ⇒⇐ to inform their
reader that they have arrived at a contradiction.
Now consider a rather famous proof by contradiction, dating back to ancient
Greek times when it was believed that all numbers were rational. That is, a
rational number must be expressed as the ratio of one integer to another, a b.
√
Theorem 4 2 is an irrational number.
√ √
Proof: Suppose not. Suppose 2 is rational. Then 2 = a b for some
a2
a, b ∈ Z, b 6= 0 and a, b have no common factors. So 2 = b2 , therefore 2b2 = a2 .
Hence a2 is even. So, by Example 14, a is even. So there exists k ∈ Z such that
a = 2k. Hence
2b2 = a2
= (2k)2
= 4k2
Hence, b2 = 2k2 , which shows that b2 is even. By Example 14, b is even. But
√ even, 2 is a common factor of a and b ⇒⇐.
since a is also
Therefore, 2 is irrational.
Example 20 Prove that if a is an integer, then a cannot be both even and odd.
There are multiple ways to parse out the logic of this claim, but at its core
it is a simple conditional statement. To use contradiction, we would suppose
that the implication is NOT true, so we suppose the antecedent and NOT the
consequent.
Proof: Suppose not. Suppose a is an integer and a is both even and odd.
Then a = 2k and a = 2l + 1 for some integers, k and l. So 2k = 2l + 1. Thus,
2k − 2l = 1, which means 2(k − l) = 1. Since k − l ∈ Z, it follows that 1 is even,
which is a contradiction. Hence, If a is an integer, then a cannot be both even
and odd.
√
Example 21 Let x ∈ R. Prove that if x2 < 5 then |x| < 5.
√ √
√ Suppose not. Suppose x < 5 and |x| ≥ 5. Then either x ≥ 5
2
Proof:
or x ≤ − √ 5. √
If x ≥ 5, then x2 ≥ 5. If x ≤ − 5 then x2 ≥ 5. In either √ case we have
x2 ≥ 5 which contradicts our assumption that x2 < 5. Thus |x| < 5.
19
Exercises for Section 1.6
A Exercises
Exercise 15 Let a ∈ Z. Prove, by contradiction, that if 5a + 7 is odd, then a
is even.
B Exercises
C Exercises
Exercise 19 Use Exercise 18 to prove that the set of prime numbers is infinite.
In order to convince yourself this claim is true, you may jot down some quick
notes such as the following.
While this should increase your confidence that the claim is true, as we
have established earlier there no way to do an exhaustive proof of a universally
quantified statement like this, similar to what we saw in Example 2. We see a
similar quandaries with innocent-looking algebraic claims.
20
n(n+1)
Example 23 Claim: ∀n ∈ N, 1 + 2 + 3 + · · · + n = 2 .
You may be again tempted to see if the equation holds true for some randomly-
chosen values for n, but that hardly counts as a proof. While there are many
ways to prove this claim, it is a great example of the effectiveness of a new
type of reasoning called inductive reasoning. The idea is as follows. To show a
statement p(n) is true for all n ∈ N, you want to show that p(1) is true, and
p(2) is true, and p(3) is true, etc. Since this is an exhastive task with no end
in sight, we might wonder whether we can “automate” the verification of each
with the effectiveness of logical implication. That is, what if we showed p(1)
is true, but p(1) is true implies p(2) is true, and p(2) is true implies p(3)
is true, and in general p(k) is true implies p(k + 1) is true? If k were to be
arbitrary, this would effectively set up an infinite string of “dominoes” that all
fall in perpetual sequence, assuring us that every single statement p(n) would
be true. This powerful argument is the heart of proof by induction.
Show that:
1. p(1) is true.
2. ∀k ∈ N, p(k) ⇒ p(k + 1)
Then p(n) is true ∀n ∈ N
Proof: By induction on n:
1(2)
1= 2 , so the claim is true for n = 1.
k(k+1)
Now let k ∈ N and suppose that 1 + 2 + 3 + · · · + k = 2 . That is, suppose
that the claim is true for n = k. Then
k(k + 1)
1 + 2 + 3 + · · · + k + (k + 1) = + (k + 1)
2
k(k + 1) 2k + 2
= +
2 2
k2 + 3k + 2 (k + 1)(k + 2)
= =
2 2
Hence, the claim is true for n = k + 1. Therefore the claim is true for all n by
induction.
21
As this proof illustrates, arithmetic can be very helpful in transitioning from
the inductive hypothesis to the main result in your inductive step. Let’s try
another one.
Example 24 Claim: ∀n ∈ N, 1 + 2 + 22 + 23 + · · · + 2n = 2n+1 − 1.
Again, you may be tempted to check whether this pattern holds for n =
1, 2, 3, 4, · · · , but an exhaustive proof is not possible. This is precisely the type
of claim that induction is most suited to proving.
Proof: By induction on n:
1+1
1+2=2 − 1, so the claim is true for n = 1.
Now let k ∈ N and suppose that 1 + 2 + 22 + 23 + · · · + 2k = 2k+1 − 1. That is,
suppose that the claim is true for n = k. Then
Hence, the claim is true for n = k + 1. Therefore the claim is true for all n by
induction.
12 < 2(1) + 1
22 < 2(2) + 1
32 > 2(3) + 1
42 > 2(4) + 1
52 > 2(5) + 1
.
.
.
While at first the number on the left is smaller, If we were to continue this
pattern, we would find that the number on the left seems to be larger most of
the time. We now have reason to make the following claim.
Example 25 Claim: ∀n ≥ 3, n2 > 2n + 1.
This seems like a situation that calls for inductive reasoning, but the claim
is clearly not true for all natural numbers. However, it appears that once n
passes a “tipping point,” the inequality always holds. So why not apply the
22
same inductive logic, but begin with a different “base case?” This is precisely
the idea behind the generalized principle of induction.
Show that:
1. p(n0 ) is true.
2. ∀k ∈ N, (k ≥ n0 ∧ p(k)) ⇒ p(k + 1)
Then p(n) is true ∀n ≥ n0
Functionally, this is simply an inductive proof where the base case can refer
to any natural number that acts as the tipping point. Let’s use this to prove
Example 25.
(k + 1)2 = k2 + 2k + 1
> (2k + 1) + 2k + 1 (by the inductive hypothesis)
= 4k + 2
> 2k + 3
Hence, the claim is true for n = k + 1. Therefore the claim is true for all n ≥ 3
by generalized induction.
23