Proving theorems and special cases (Part 12): The sum and difference formulas for sine

In a recent class with my future secondary math teachers, we had a fascinating discussion concerning how a teacher should respond to the following question from a student:

Is it ever possible to prove a statement or theorem by proving a special case of the statement or theorem?

Usually, the answer is no. In this series of posts, we’ve seen that a conjecture could be true for the first 40 cases or even the first $10^{316}$ cases yet not always be true. We’ve also explored the computational evidence for various unsolved problems in mathematics, noting that even this very strong computational evidence, by itself, does not provide a proof for all possible cases.

However, there are plenty of examples in mathematics where it is possible to prove a theorem by first proving a special case of the theorem. For the remainder of this series, I’d like to list, in no particular order, some common theorems used in secondary mathematics which are typically proved by first proving a special case.

3. Theorem 1. $\sin(\alpha + \beta) = \sin \alpha \cos \beta + \cos \alpha + \sin \beta$

Theorem 2. $\sin(\alpha - \beta) = \sin \alpha \cos \beta - \cos \alpha \sin \beta$

For angles that are not acute, these theorems can be proven using a unit circle and the following four lemmas:

Lemma 1. $\cos(x - y) = \cos x \cos y + \sin x \sin y$

Lemma 2. $\cos(x + y) = \cos x \cos y - \sin x \sin y$

Lemma 3. $\sin(\pi/2 - x) = \cos x$

Lemma 4. $\cos(\pi/2 - x) = \sin x$

Specifically, assuming Lemmas 1-4, then:

$\sin(\alpha + \beta) = \cos(\pi/2 - [\alpha + \beta])$ by Lemma 4

$= \cos([\pi/2 - \alpha] - \beta)$

$= \cos(\pi/2 - \alpha) \cos \beta + \sin(\pi/2 - \alpha) \sin \beta$ by Lemma 1

$= \sin \alpha \cos \beta + \cos \alpha \sin \beta$ by Lemmas 3 and 4.

Also,

$\sin(\alpha - \beta) = \cos(\pi/2 - [\alpha - \beta])$ by Lemma 4

$= \cos([\pi/2 - \alpha] + \beta)$

$= \cos(\pi/2 - \alpha) \cos \beta - \sin(\pi/2 - \alpha) \sin \beta$ by Lemma 2

$= \sin \alpha \cos \beta - \cos \alpha \sin \beta$ by Lemmas 3 and 4.

However, we see that what I’ve called Lemma 3, often called a cofunction identity, can be considered a special case of Theorem 2. However, this is not circular logic since the cofunction identities can be proven without appealing to Theorems 1 and 2.

Proving theorems and special cases (Part 11): The Law of Cosines

In a recent class with my future secondary math teachers, we had a fascinating discussion concerning how a teacher should respond to the following question from a student:

Is it ever possible to prove a statement or theorem by proving a special case of the statement or theorem?

2. Theorem. In $\triangle ABC$ where $a = BC$ , $b = AC$ , and $c = AB$ , we have $c^2 = a^2 + b^2 - 2 a b \cos (m \angle C)$ .

This is typically proven using the Pythagorean theorem:

Lemma. In right triangle $\triangle ABC$ , where $\angle C$ is a right angle, we have $c^2 = a^2 + b^2$ .

Though it usually isn’t thought of this way, the Pythagorean theorem is a special case of the Law of Cosines since $\cos 90^\circ = 0$ .

There are well over 100 different proofs of the Pythagorean theorem that do not presuppose the Law of Cosines. The standard proof of the Law of Cosines then uses the Pythagorean theorem. In other words, a special case of the Law of Cosines is used to prove the Law of Cosines.

Proving theorems and special cases (Part 10): Angles in a convex n-gon

In a recent class with my future secondary math teachers, we had a fascinating discussion concerning how a teacher should respond to the following question from a student:

Is it ever possible to prove a statement or theorem by proving a special case of the statement or theorem?

1. Theorem. The sum of the angles in a convex n-gon is $180(n-2)$ degrees.

This theorem is typically proven after first proving the following lemma:

Lemma. The sum of the angles in a triangle is $180$ degrees.

Clearly the lemma is a special case of the main theorem: for a triangle, $n=3$ and so $180(n-2) = 180 \times 1 = 180$ . The proof of the lemma uses alternate interior angles and the convention that the angle of a straight line is 180 degrees.

Using this, the main theorem follows by using diagonals to divide a convex n-gon into $n-2$ triangles. (For example, drawing a diagonal divides a quadrilateral into two triangles.) The sum of the angles of the n-gon must equal the sum of the angles of the $n-2$ triangles.

So it is possible to prove a theorem by proving a special case of the theorem. Using the sum of the angles of a triangle to prove the formula for the sum of the angles of a convex n-gon is qualitatively different than the previous computational examples seen earlier in this series.

Proving theorems and special cases (Part 9): The Riemann hypothesis

In a recent class with my future secondary math teachers, we had a fascinating discussion concerning how a teacher should respond to the following question from a student:

Is it ever possible to prove a statement or theorem by proving a special case of the statement or theorem?

Usually, the answer is no. In this series of posts, we’ve already seen that a conjecture could be true for the first 40 cases or even the first $10^{316}$ yet ultimately prove false for all cases.

For the next few posts, I thought I’d share a few of the most famous unsolved problems in mathematics… and just how much computational work has been done to check for a counterexample.

4. The Riemann hypothesis (see here, here, and here for more information) is perhaps the outstanding unsolved problem in pure mathematics, and a prize of $1 million has been offered for its proof.

The Riemann zeta function is defined by

$\zeta(s) = \displaystyle \sum_{n=1}^\infty \frac{1}{n^s}$

for complex numbers $s$ with real part greater than 1. For example,

$\zeta(2) = \displaystyle \frac{1}{1^2} + \frac{1}{2^2} + \frac{1}{3^2} + \dots = \displaystyle \frac{\pi^2}{6}$

and

$\zeta(4) = \displaystyle \frac{1}{1^4} + \frac{1}{2^4} + \frac{1}{3^4} + \dots = \displaystyle \frac{\pi^4}{90}$

The definition of the Riemann zeta function can be extended to all complex numbers (except a pole at $s = 1$ ) by the integral

$\zeta(s) = \displaystyle \frac{1}{\Gamma(s)} \int_0^\infty \frac{x^{s-1}}{e^s - 1} dx$

and also analytic continuation.

The Riemann hypothesis conjectures to all solutions of the equation $\zeta(s) = 0$ other than negative even integers occur on the line $s = \frac{1}{2} + i t$ . At present, it is known that the first 10 trillion solutions are on this line, so that every solution with $t < 2.4 \times 10^{12}$ is on this line. Of course, that’s not a proof that all solutions are on this line.

A full description of known results concerning the Riemann hypothesis requires much more than a single post. I’ll refer the interested reader to the links above from MathWorld, Wikipedia, and Claymath and the references embedded in those links. An excellent book for the layman concerning the Riemann hypothesis is Prime Obsession: Bernhard Riemann and the Greatest Unsolved Problem in Mathematics, by John Derbyshire.

Proving theorems and special cases (Part 8): The Collatz conjecture

In a recent class with my future secondary math teachers, we had a fascinating discussion concerning how a teacher should respond to the following question from a student:

Is it ever possible to prove a statement or theorem by proving a special case of the statement or theorem?

Usually, the answer is no. In this series of posts, we’ve already seen that a conjecture could be true for the first 40 cases or even the first $10^{316}$ cases yet ultimately prove false for all cases.

For the next few posts, I thought I’d share a few of the most famous unsolved problems in mathematics… and just how much computational work has been done to check for a counterexample.

3. The Collatz conjecture (see here and here for more information) is an easily stated unsolved problem that can be understood by most fourth and fifth graders. Restated from a previous post:

Here’s the statement of the problem.

Start with any positive integer.
If the integer is even, divide it by $2$ . If it’s odd, multiply it by $3$ and then add $1$ .
Repeat until (and if) you reach $1$ .

Here’s the question: Does this sequence eventually reach $1$ no matter the starting value? Or is there a number out there that you could use as a starting value that has a sequence that never reaches $1$ ?

For every integer less than $5 \times 2^{60} = 5,764,607,523,034,234,880$ , this sequence returns to 1. Of course, this is not a proof that the conjecture will hold for every integer.

Proving theorems and special cases (Part 7): The twin prime conjecture

In a recent class with my future secondary math teachers, we had a fascinating discussion concerning how a teacher should respond to the following question from a student:

Is it ever possible to prove a statement or theorem by proving a special case of the statement or theorem?

For the next few posts, I thought I’d share a few of the most famous unsolved problems in mathematics… and just how much computational work has been done to check for a counterexample.

2. The twin prime conjecture (see here and here for more information) asserts that there are infinitely many primes that have a difference of 2. For example:

3 and 5 are twin primes;

5 and 7 are twin primes;

11 and 13 are twin primes;

17 and 19 are twin primes;

29 and 31 are twin primes; etc.

While most mathematicians believe the twin prime conjecture is correct, an explicit proof has not been found. The largest known twin primes are

$3,756,801,695,685 \times 2^{666,669} \pm 1$ ,

numbers which have 200,700 decimal digits. Also, there are 808,675,888,577,436 twin prime pairs less than $10^{18}$ .

Most mathematicians also believe that there are infinitely many cousin primes, which differ by 4:

3 and 7 are cousin primes;

7 and 11 are cousin primes;

13 and 17 are cousin primes;

19 and 23 are cousin primes;

37 and 41 are cousin primes; etc.

Most mathematicians also believe that there are infinitely many sexy primes (no, I did not make that name up), which differ by 6:

5 and 11 are sexy primes;

7 and 13 are sexy primes;

11 and 17 are sexy primes;

13 and 19 are sexy primes;

17 and 23 are sexy primes; etc.

(Parenthetically, a “sexy” prime is probably the most unfortunate name in mathematics ever since Paul Dirac divided a bracket into a “bra” and a “ket,” thereby forever linking women’s underwear to quantum mechanics.)

While it is unknown if there are infinitely many twin primes, it was recently shown — in 2013 — that, for some integer $N$ that is less than 70 million, there are infinitely many pairs of primes that differ by $N$ . In 2014, this upper bound was reduced to 246. Furthermore, if a certain other conjecture is true, the bound has been reduced to 6. In other words, there are infinitely many twin primes or cousin primes or sexy primes… but, at this moment, we don’t know which one (or ones) is infinite.

In November 2014, Dr. Terence Tao of the UCLA Department of Mathematics was interviewed on the Colbert Report to discuss the twin prime conjecture… and he does a good job explaining to Stephen Colbert how we can know one of the three categories is infinite without knowing which category it is.

From the Colbert Report: http://thecolbertreport.cc.com/videos/6wtwlg/terence-tao

Proving theorems and special cases (Part 6): The Goldbach conjecture

In a recent class with my future secondary math teachers, we had a fascinating discussion concerning how a teacher should respond to the following question from a student:

Is it ever possible to prove a statement or theorem by proving a special case of the statement or theorem?

For the next few posts, I thought I’d share a few of the most famous unsolved problems in mathematics… and just how much computational work has been done to check for a counterexample.

1. The Goldbach conjecture (see here and here for more information) claims that every even integer greater than 4 can be written as the sum of two prime numbers. For example,

4 = 2 + 2,

6 = 3 + 3,

8 = 3 + 5,

10 = 3 + 7,

12 = 5 + 7,

14 = 3 + 11, etc.

This has been verified for all even numbers less than $4 \times 10^{18} = 4,000,000,000,000,000,000$ . A proof for all even numbers, however, has not been found yet.

Here are some results related to the Goldbach conjecture that are known:

1. Any integer greater than 4 is the sum of at most six primes.

2. Every sufficiently large even number can be written as the sum or two primes or the sum of a prime and the product of two primes.

3. Every sufficiently large even number can be written as the sum of two primes and at most 8 powers of 2.

4. Every sufficiently large odd number can be written as the sum of three primes.

Proving theorems and special cases (Part 5): Mathematical induction

Today’s post is a little bit off the main topic of this series of posts… but I wanted to give some pedagogical thoughts on yesterday’s post concerning the following proof by induction.

Theorem: If $n \ge 1$ is a positive integer, then $5^n - 1$ is a multiple of 4.

Proof. By induction on $n$ .

$n = 1$ : $5^1 - 1 = 4$ , which is clearly a multiple of 4.

$n$ : Assume that $5^n - 1$ is a multiple of 4, so that $5^n - 1 = 4q$ , where $q$ is an integer. We can also write this as $5^n = 4q + 1$ .

$n+1$ . We wish to show that $5^{n+1} - 1$ is equal to $4Q$ for some (different) integer $Q$ . To do this, notice that

$5^{n+1} - 1 = 5^1 5^n - 1$

$= 5 \times 5^n - 1$

$= 5 \times (4q + 1) - 1$ by the induction hypothesis

$= 20q + 5 - 1$

$= 20q + 4$

$= 4(5q + 1)$ .

So if we let $Q = 5q +1$ , then $5^{n+1} - 1 = 4Q$ , where $Q$ is an integer because $q$ is also an integer.

My primary observation is that even very strong math students tend to have a weak spot when it comes to simplifying exponential expressions (as opposed to polynomial expressions). For example, I find that even very good math students can struggle through the logic of this sequence of equalities:

$2^n + 2^n = 2 \times 2^n = 2^1 \times 2^n = 2^{n+1}$ .

The first step is using the main stumbling block. Students who are completely comfortable with simplifying $x + x$ as $2x$ can be perplexed by simplifying $2^n + 2^n$ as $2 \times 2^n$ . I attribute this to lack of practice with this kind of simplification in lower grade levels.

Here’s another algoebraic stumbling block that I’ve often seen: at the beginning of the $n+1$ case, some students will make the following mistake:

$5^{n+1} - 1 = 5^1 5^n - 1 = 5 (4q) = 4 (5q) = 4Q$ .

Because these students end with a multiple of 4, they fail to notice that the second equality is incorrect since

$5^1 5^n - 1 \ne 5^1 (5^n - 1)$ .

Again, I attribute this to lack of practice with simplifying exponential expressions in lower grade levels… as well as being a little bit over-excited upon seeing $5^n - 1$ and wishing to use the induction hypothesis as soon as possible.

Proving theorems and special cases (Part 4): Mathematical induction

In a recent class with my future secondary math teachers, we had a fascinating discussion concerning how a teacher should respond to the following question from a student:

Is it ever possible to prove a statement or theorem by proving a special case of the statement or theorem?

Usually, the answer is no… even checking many special cases of a conjecture does not mean that the conjecture is correct. In the previous two posts, we saw that a statement that’s true for the first 40 cases or even the first $10^{316}$ cases may not be true for all cases.

This is the reason that mathematical induction is important, as it provides a way to build from previous cases to prove that the next case is still correct, thus proving that all cases are correct.

Theorem: If $n \ge 1$ is a positive integer, then $5^n - 1$ is a multiple of 4.

Proof. By induction on $n$ .

$n = 1$ : $5^1 - 1 = 4$ , which is clearly a multiple of 4.

$n$ : Assume that $5^n - 1$ is a multiple of 4, so that $5^n - 1 = 4q$ , where $q$ is an integer. We can also write this as $5^n = 4q + 1$ .

$n+1$ . We wish to show that $5^{n+1} - 1$ is equal to $4Q$ for some (different) integer $Q$ . To do this, notice that

$5^{n+1} - 1 = 5^n 5^1 - 1$

$= 5 \times 5^n - 1$

$= 5 \times (4q + 1) - 1$ by the induction hypothesis

$= 20q + 5 - 1$

$= 20q + 4$

$= 4(5q + 1)$ .

So if we let $Q = 5q +1$ , then $5^{n+1} - 1 = 4Q$ , where $Q$ is an integer because $q$ is also an integer.

QED

In the above proof, we were able to build from the $n$ case to reach the $n +1$ case. In this sense, to answer the student’s question, it is possible to prove a theorem by first proving a special case of the theorem.

By contrast, when trying to “prove” that $n^2 - n + 41$ is prime for all integers $n$ , the proposition is true for $1 \le n \le 40$ , but it’s just a coincidence… there was no string of logic that connected these first 40 cases other than the coincidence that they all were correct.

Proving theorems and special cases (Part 3): Skewes’ number

In a recent class with my future secondary math teachers, we had a fascinating discussion concerning how a teacher should respond to the following question from a student:

Is it ever possible to prove a statement or theorem by proving a special case of the statement or theorem?

Usually, the answer is no… even checking many special cases of a conjecture does not mean that the conjecture is correct.

In the first two posts of this series, we showed conjectures could be true for the first 40 cases or even the first 900 million odd cases but fail on the next case. In today’s post, I’ll describe a conjecture that, for hundreds of years, was thought to be true for all integers $n$ and has since been shown to be true for all $n \le 10^{316}$ . However, despite being true for so many special cases, the conjecture is false.

Let’s absorb the above paragraph again. The conjecture “ $n^2 - n + 41$ is always prime” was true for the first 40 cases before failing. By contrast, the conjecture I’m about the describe is true for the first (approximately) 10,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,
000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,
000,000,000,000,000,000,000,000,000 cases before failing.

The conjecture in question concerns the number of prime numbers $\pi(n)$ that are less than or equal to $n$ . For example:

There are four prime numbers less than 10 (namely 2, 3, 5, 7), and so $\pi(10) = 4$ .

There are eight prime numbers less than 10 (namely 2, 3, 5, 7, 11, 13, 17, 19), and so $\pi(20) = 8$ .

One of the classic problems in mathematics, often called the Prime Number Theorem, is estimating the value of $\pi(n)$ when $n$ is large. It turns out that one way of approximating $\pi(n)$ is through the logarithmic integral

$\hbox{Li}(n) = \displaystyle \int_2^n \frac{dx}{\ln x}$

Indeed, the Prime Number Theorem states that

$\pi(n) \sim \hbox{Li}(n)$ ,

$\displaystyle \lim_{n \to \infty} \frac{\pi(n)}{\hbox{Li}(n)} = 1$ .

This asymptotic relationship was proven by the eminent mathematician Karl Friedrich Gauss.

With that as prelude, here’s the false conjecture. For decades, luminaries of mathematics, including Gauss and Bernhard Riemann, conjectured that

$\pi(n) < \hbox{Li}(n)$ for all integers $n$

based on the available numerical evidence at the time. However, this conjecture was proven false in the 20th century by the British mathematician John Littlewood. No only did Littlewood show that there is a value of $n$ so that $\pi(n) > \hbox{Li}(n)$ , but he also showed that the graphs of $\pi(n)$ and $\hbox{Li}(n)$ cross over each other infinitely many times!

Littlewood proved his theorem over 100 years ago in 1914. Today, even with modern computing power, we still do not precisely know the first value of $n$ for which $\pi(n) > \hbox{Li}(n)$ . This number is called Skewes’ number, in honor of the mathematician who first found (in 1955) an upper bound on this first crossing point. The bound that he found was absolutely enormous: he showed that $\pi(n) > \hbox{Li}(n)$ for some $n$ less than

$\displaystyle 10^{\displaystyle 10^{10^{34}}}$

More recent work has established that the first crossover likely occurs in the vicinity of $n \approx 1.397 \times 10^{316}$ .

Whoever first evaluates Skewes’ number exactly will surely have a nice feather in his/her cap for completing a task that mystified both Gauss and Riemann.

References:

http://mathworld.wolfram.com/PrimeNumberTheorem.html

http://mathworld.wolfram.com/SkewesNumber.html

http://mathworld.wolfram.com/PrimeCountingFunction.html

http://mathworld.wolfram.com/LogarithmicIntegral.html