4
$\begingroup$

Found this equality $$\tan 15^{\circ} = \tan 27^{\circ} \tan 33^{\circ} \tan 39^{\circ}$$ with a random search. It checks with WA, and these kind of equalities can be proved automatically.

I am looking for an elementary proof.

Some similarly looking equalities are using the identity;

$$\tan 3 x= \tan x\cdot \tan(\frac{\pi}{3}-x)\cdot \tan (\frac{\pi}{3}+x)$$

We notice that $27^{\circ} + 33^{\circ} = 60^{\circ}$, but does not seem enough.

Thank you for interest!

$\bf{Added:}$ Let me add the mechanical proof that works like here. One considers the basic angle $x = 3^{\circ}= \frac{2 \pi}{120}$, and we need to show that $\tan 5 x= \tan 9 x \tan 11 x \tan 13 x$, which is equivalent to: $z= e^{\frac{2 \pi i}{120}}$ is a root of the equation

$$\frac{z^{10}-1}{z^{10}+1} + \frac{z^{18} -1}{z^{18} + 1}\cdot \frac{z^{22} -1}{z^{22} + 1}\cdot \frac{z^{26} -1}{z^{26} + 1}=0$$

LHS factors and we see the cyclotomic polynomial $\Phi_{120}(z)$ in the numerator.

$\bf{Added:}$ Since in general we have to show some equality for functions of multiples of an angle $x$, where $x$ is a rational multiple of $2\pi$, we could prove an equivalent equality for the angle $m x$, where $m$ is relatively prime to the denominator of $\frac{x}{2 \pi}$, so we might reduce to other known identities. Perhaps this one works with say $m = 7$? Since $7\times 15^{\circ} = 90^{\circ} + 15^{\circ}$.

$\endgroup$
2
  • 2
    $\begingroup$ Maybe express all the tangents as roots of unity. $\tan x=-i(e^{2\pi ix}-e^{-2\pi ix})/(e^{2\pi ix}+e^{-2\pi ix})$. $\endgroup$ Commented Jan 11 at 1:19
  • $\begingroup$ @Gerry Myerson:Thank you, yes, that is the automatic way I guess. really interested in some juggling with angles and basic equalities like the one with triple the angle, something that a good high school student could understand. $\endgroup$
    – orangeskid
    Commented Jan 11 at 1:37

2 Answers 2

2
$\begingroup$

My original solution wasn't quite accurate. However, this one is, and the answer uses some nice elementary number theory.


Claim: $\cos(108^{\circ}) = \cos(48^{\circ})+\cos(168^{\circ})$.

In fact, for any $\theta$, we have $$ \cos(\theta)=\cos(\theta+60^{\circ})+\cos(\theta-60^{\circ})$$


Proof:

Clear denominators and multiply each side by $4$: $$4\sin(15{}^{\circ})\cos(27{}^{\circ})\cos(33{}^{\circ})\cos(39{}^{\circ})=4\sin(27{}^{\circ})\cos(15{}^{\circ})\sin(33{}^{\circ})\sin(39{}^{\circ})$$ Product-to-sum: $$(\sin(42{}^{\circ})+\sin(-12{}^{\circ}))(\cos(72{}^{\circ})+\cos(-6{}^{\circ}))=(\sin(42{}^{\circ})+\sin(12{}^{\circ}))(\cos(6{}^{\circ})-\cos(72{}^{\circ}))$$ Correct the signs using the parity of the functions and expand. After cancelling terms and rearranging, we are left with: $$2\sin(42^{\circ})\cos(72^{\circ})=2\sin(12^{\circ})\cos(6^{\circ})$$Turn sines into cosines, product-to-sum, and use complementary angles: $$2\cos(48^{\circ})\cos(72^{\circ})=2\cos(78^{\circ})\cos(6^{\circ})$$ $$\cos(24^{\circ})+\cos(120^{\circ})=\cos(72^{\circ})+\cos(84^{\circ})$$ $$\cos(24^{\circ})-\cos(72^{\circ})-\cos(84^{\circ})=\frac{1}{2}$$ $$\cos(24^{\circ})+\cos(96^{\circ})+\cos(108^{\circ})=\frac{1}{2}$$

By the claim, we can rewrite this as: $$\cos(24^{\circ})+ \cos(48^{\circ})+\cos(96^{\circ})+\cos(168^{\circ})=\frac{1}{2}$$ Rewrite the arguments in radians: $$\cos\left(2\pi \cdot \frac{1}{15}\right)+ \cos\left(2\pi \cdot \frac{2}{15}\right)+\cos\left(2\pi \cdot \frac{4}{15}\right)+\cos\left(2\pi \cdot \frac{7}{15}\right)=\frac{1}{2}$$ Here I paraphrase from the Wikipedia article on trig-identities:

The factors $1,2,4,7$ may start to make the pattern clear: they are those integers less than $15/2$ that are coprime to $15$. The last several examples are corollaries of a basic fact about the irreducible cyclotomic polynomials: the cosines are the real parts of the zeroes of those polynomials; the sum of the zeroes is the Möbius function evaluated at (in the very last case above) $15$; only half of the zeroes are present above.

If you don't want to rely on number theory, it is possible to keep using angle-addition to transform the LHS into $1/2$ using double-angle and a few other tricks; exercise.


I'm sure you could reverse-engineer additional identities by starting with similar evaluations of the Möbius function.

https://en.wikipedia.org/wiki/List_of_trigonometric_identities#Identities_without_variables

$\endgroup$
1
  • 2
    $\begingroup$ $$\cos48^\circ+\cos168^\circ=\cos48^\circ-\cos12^\circ=2\sin\left(\dfrac{12+48}2\right)^\circ\sin\left(\dfrac{12-48}2\right)^\circ=-\sin18^\circ=\cos(90+18)^\circ$$ $\endgroup$ Commented Jan 11 at 3:17
1
$\begingroup$

Using the identity $\tan x=\frac{\sin 2x}{1+\cos 2x},$ it is equivalent to showing that $$\frac{\sin 30^\circ}{1+\cos 30^\circ}=\frac{\sin 54^\circ}{1+\cos 54^\circ}\cdot\frac{\sin 66^\circ}{1+\cos 66^\circ}\cdot \frac{\sin 78^\circ}{1+\cos 78^\circ}.$$ But now each of the trig values can be evaluated directly. For example, $\sin 54^\circ=\sin(3\cdot 18^\circ), \sin(66^\circ)=\sin(30^\circ+36^\circ),\sin(78^\circ)=\sin(60^\circ+18^\circ),$ etc.

$\endgroup$
2
  • $\begingroup$ Interesting... is the etc part complicated or just routine? $\endgroup$
    – orangeskid
    Commented Jan 11 at 2:05
  • $\begingroup$ One trick to find $\sin 18^\circ$ is to observe that $x=18^\circ$ satisfies $5x=90^\circ,$ hence $\sin 3x=\sin(90^\circ-2x).$ Expressing both sides as polynomial in $\sin x$ allows to find the value of $\sin x.$ $\endgroup$
    – Pythagoras
    Commented Jan 11 at 2:12

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .