13

Without a calculator, determine whether chords $AB, AC \text{ and }BD$ can divide a circle into five equal-area regions.

enter image description here

Context: I just made up this question.

I have only been able to answer the question with a calculator. Surely there must be some clever way to answer the question without a calculator.

Answer that uses a calculator:

Proof by contradiction: assume the answer is yes.
Let radius $=1$.
Call the centre $O$.
$\alpha=\angle{AOB}$
$\beta=\angle{AOC}$
$\text{Area}_{\text{minor segment }AB}=\dfrac12(\alpha-\sin{\alpha})=\dfrac{\pi}{5}\implies \alpha\approx2.1131$
$\text{Area}_{\text{segment }ADC}=\dfrac12(\beta-\sin{\beta})=\dfrac{2\pi}{5}\implies \beta\approx2.8248$
$\text{Area}_{\text{triangle}}=\dfrac{\pi}{5}\text{ and }AB=2\sin{\dfrac{\alpha}{2}}\implies ...\implies \angle{BAC}=\arctan{\left(\dfrac{\pi}{5\sin^2{\dfrac{\alpha}{2}}}\right)}\approx{0.692}$
But $\angle{BAC}=\dfrac12{\angle{BOC}}=\dfrac12{\left(2\pi-\alpha-\beta\right)}\approx{0.673}$, contradiction.
Therefore the answer is no.

UPDATE: I found essentially the same question, except it does not request a calculator-free solution. So it doesn't answer my question.

Dan
  • 22,158

3 Answers3

4

Second try (cleaner) Please refer to the diagram in my other answer.

By naming $2\theta=\widehat{BOA},\, 2\varphi=\widehat{AOC}$ and by imposing that the circle segments on $AB,AC,BD$ have the correct areas we end up with $$ \theta-\frac{1}{2}\sin(2\theta)=\frac{\pi}{5},\qquad \varphi-\frac{1}{2}\sin(2\varphi)=\frac{2\pi}{5}.\tag{A}$$ In the triangle $ABE$ we have $AB=2\sin\theta$ and $\widehat{ABE}=\pi-(\theta+\varphi)$, so $$ [ABE] = -\sin^2(\theta)\tan(\theta+\varphi) \tag{B}$$ and given $(A)$ it is enough to show that $(B)$ cannot be equal to $\frac{\pi}{5}$.
$f(x)=x-\frac{\sin(2x)}{2}$ is non-negative, convex and increasing from $0$ to $\pi/2$ on $[0,\pi/2]$, with $f'(x)=2\sin^2(x)$. We have $\theta\approx \pi/3$ and by a single step of Newton's method with starting point $\pi/2$ we have $\varphi\approx 9\pi/20$. Precisely $\theta > \pi/3$ and $\varphi < 9\pi/20$, and the error of these approximations is controlled by the fact that $1\leq f'\leq 2$ in $[\pi/4,\pi/2]$. In particular

$$ \theta < \frac{\pi}{3}+\left(\frac{\pi}{5}-f(\pi/3)\right) $$ $$ \varphi > \frac{9\pi}{20}-\left(f(9\pi/20)-\frac{2\pi}{5}\right)\tag{C}$$ such that $[ABE]$ is very close to $\frac{3}{4}\tan\left(\frac{13}{60}\pi\right)$.
The impossibility of the pentasection then follows from the proof of the following inequality: $$ \tan\left(\frac{13\pi}{60}\right)<\frac{4\pi}{15}=\frac{16\pi}{60}\tag{D} $$which is equivalent to $$ \frac{1-\tan(\pi/30)}{1+\tan(\pi/30)} < \frac{4\pi}{15}\tag{E} $$ or to $$ \tan\left(\frac{\pi}{30}\right) > \frac{15-4\pi}{15+4\pi} \tag{F}$$ which just follows from $$ \tan\left(\frac{\pi}{30}\right) > {\color{red}{\frac{\pi}{30}}} > \frac{15-4\pi}{15+4\pi} \tag{G}$$ since $\pi > \frac{15}{8}(\sqrt{113}-9).$

Jack D'Aurizio
  • 353,855
  • Shouldn't equation (C) be $\varphi < ...$, because $f(x)$ is concave up? Also, I'm not sure how you go from (C) to (D), and from (D) to (E); perhaps you could elaborate? Thanks. – Dan Aug 06 '22 at 03:25
  • Nice solution, fou sûre ! Thanks – Claude Leibovici Aug 06 '22 at 17:45
  • @ClaudeLeibovici This answer seems more elegant than mine, so I would love to accept it. But I'm still struggling to understand it: I hope someone can address the questions that I asked in my first comment on this answer. – Dan Aug 06 '22 at 22:06
  • @Dan: the Lipschitz-continuity of $f$ grants that $[ABE]$ is at most $0.02$ apart from $\frac{3}{4}\tan\left(\frac{13}{60}\pi\right)$. This gives that at the current state this is not a complete answer, one needs to show not only that $\frac{4\pi}{15}-\tan\left(\frac{13\pi}{60}\right)$ is positive, but that it is greater than $0.02$. Which is true, but does not directly follow from (G). – Jack D'Aurizio Aug 06 '22 at 22:11
  • By estimating the difference between $\tan(\pi/30)$ and $\pi/30$, as well as the difference between $\frac{15-4\pi}{15+4\pi}$ and $\frac{\pi}{30}$ one gets an improved version of (G) and also an improved version of (D) which actually settles the matter. – Jack D'Aurizio Aug 06 '22 at 22:14
  • @JackD'Aurizio Thanks. Shouldn't equation (C) be $\varphi < ...$ instead of $\varphi>...$? – Dan Aug 06 '22 at 22:33
  • No, that should be correct. $\varphi$ is less than $9\pi/20$, but due to the Lipschitz-continuity of $f$ it is greater than $9\pi/20$ minus something positive (but very small) – Jack D'Aurizio Aug 06 '22 at 22:36
  • @JackD'Aurizio In (C) you have $\theta<...$ but $\varphi>...$. I don't understand why the directions of the inequalities are different. There should be no difference, because $f(x)$ is concave up at both $x=\theta$ and at $x=\varphi$. – Dan Aug 07 '22 at 00:27
  • The direction is different because the approximation for $\varphi$ is obtained by applying Newton's method with starting point $\pi/2$, which is obviously greater than $\varphi$, while $\pi/3$ is actually less than $\theta$. – Jack D'Aurizio Aug 07 '22 at 09:58
  • @JackD'Aurizio According to Wolfram, the solution to $\varphi-\frac{1}{2}\sin(2\varphi)=\frac{2\pi}{5}$ is $\varphi=1.41239...$ but you wrote $\varphi > \frac{9\pi}{20}-\frac{1}{2}\left(f(9\pi/20)-\frac{2\pi}{5}\right)$, the right side is $1.41243...$. There is definitely something wrong here. – Dan Aug 07 '22 at 14:01
  • @Dan: oh, I get it. Indeed $f'=2$ at $\pi/2$, but $f'$ is increasing, so the constant $\frac{1}{2}$ has to be replaced by something slightly larger. Fixing. – Jack D'Aurizio Aug 07 '22 at 14:15
0

enter image description here

Let us assume that the radius of the circle is $1$ and let $\widehat{BOA}=2\theta$.
Then $\theta$ is the unique solution of $\theta-\frac{1}{2}\sin(2\theta)=\frac{\pi}{5}$ and $AB=\sin(2\theta)$.
$[AEB]=\frac{\pi}{5}$ then implies $ME=\frac{\pi}{5\sin\theta}$ and $AE=EB=\sqrt{\sin^2\theta+\frac{\pi^2}{25\sin^2\theta}}$.
The height of the circle segment with base $AB$ is $1-\cos\theta$, so by naming $P$ the antipode of $N$ we have $EP=1-\cos\theta+\frac{\pi}{5\sin\theta}$ and $EN=1+\cos\theta-\frac{\pi}{5\sin\theta}$. By the chords theorem $EN\cdot EP = EC\cdot EA$, hence

$$ EC = \frac{1-\left(\cos\theta-\frac{\pi}{5\sin\theta}\right)^2}{\sqrt{\sin^2\theta+\frac{\pi^2}{25\sin^2\theta}}}$$ and by the similarity between $EAB$ and $ECD$ we have

$$ CD = \frac{1-\left(\cos\theta-\frac{\pi}{5\sin\theta}\right)^2}{\sin^2\theta+\frac{\pi^2}{25\sin^2\theta}}\cdot 2\sin\theta$$ and

$$ \widehat{DOC} = 2\arcsin\left(\frac{1-\left(\cos\theta-\frac{\pi}{5\sin\theta}\right)^2}{\sin^2\theta+\frac{\pi^2}{25\sin^2\theta}}\cdot \sin\theta\right).$$ The length of $OE$ is $\frac{\pi}{5\sin\theta}-\cos\theta$, hence the area of the quadrilateral $DOCE$ equals

$$ [DOCE] = \left(\frac{\pi}{5\sin\theta}-\cos\theta\right)\left(\frac{1-\left(\cos\theta-\frac{\pi}{5\sin\theta}\right)^2}{\sin^2\theta+\frac{\pi^2}{25\sin^2\theta}}\cdot \sin\theta\right) $$ and the area of the portion of the circle bounded by $E,D,C$ equals

$$\arcsin\left(\frac{1-\left(\cos\theta-\frac{\pi}{5\sin\theta}\right)^2}{\sin^2\theta+\frac{\pi^2}{25\sin^2\theta}}\cdot \sin\theta\right)-\left(\frac{\pi}{5\sin\theta}-\cos\theta\right)\left(\frac{1-\left(\cos\theta-\frac{\pi}{5\sin\theta}\right)^2}{\sin^2\theta+\frac{\pi^2}{25\sin^2\theta}}\cdot \sin\theta\right).$$

Few steps of Newton's method give $\theta\approx 1.05657$, hence the last expression is $\approx 0.552681$ while $\pi/5\approx 0.628319$, proving that the wanted pentasection is impossible.

Jack D'Aurizio
  • 353,855
  • Thank you, but doesn't this method require the use of a calculator? – Dan Aug 04 '22 at 22:53
  • Yes, at the current moment a calculator is needed. On the other hand one might try to show (without a calculator) that there is no $\theta$ such that the last expression equals at the same time $\frac{\pi}{5}$ and $\theta-\sin\theta\cos\theta$. – Jack D'Aurizio Aug 04 '22 at 23:24
  • On the other hand I suspect that at some point some numerical computation is still needed, since the distance between $0.5527$ and $0.6283$ is fairly small. – Jack D'Aurizio Aug 04 '22 at 23:26
  • I am dumb since I do not see the equation you are solving for $\theta$. Could you please write it ? Thanks & cheers :-) – Claude Leibovici Aug 05 '22 at 04:03
  • @ClaudeLeibovici: $\theta$ is defined by $\theta-\frac{1}{2}\sin(2\theta)=\frac{\pi}{5}$, so $\theta$ is approximately $\pi/3$, and we have to prove that the last expression cannot be equal to $\pi/5$. – Jack D'Aurizio Aug 06 '22 at 16:07
  • @ClaudeLeibovici: anyway my second answer reduces the problem to a more appealing one, i.e. showing that if $\theta-\frac{1}{2}\sin(2\theta)=\frac{\pi}{5}$ and $\varphi-\frac{1}{2}\sin(2\varphi)=\frac{2\pi}{5}$ then $\sin^2(\theta)\tan(\theta+\varphi)\neq -\frac{\pi}{5}$. – Jack D'Aurizio Aug 06 '22 at 16:09
0

$\huge{\text{Preliminaries}}$

This answer does not require a calculator, but assumes that we can add/subtract/multiply/divide numbers of up to five digits, and uses the following facts:
$3.1415<\pi<3.1416$
$1.414<\sqrt2<1.415$
$1.732<\sqrt3<1.733$
$\sqrt{5-2\sqrt5}<0.727$ which can be proved:

$\sqrt5>2.236$
$5-2\sqrt5<0.528$
$\sqrt{5-2\sqrt5}<\sqrt{0.528}<\sqrt{0.727^2}=0.727$

$\tan{\dfrac{\pi}{5}}=\sqrt{5-2\sqrt5}$ which is proved here.

$\huge{\text{Proof by contradiction: assume the answer is yes}.}$

$\huge{\text{Define $\alpha$ and $\beta$}.}$

Let radius $=1$.
Call the centre $O$.
$\alpha=\angle{AOB}$
$\beta=\angle{AOC}$

enter image description here

$\huge{\text{Show that }\alpha>2.1.}$

$2\times\text{Area}_{\text{minor segment }AB}=\alpha-\sin{\alpha}=\dfrac{2\pi}{5}$

enter image description here

Between $P\left(\dfrac{2\pi}{3},\dfrac{2\pi}{3}-\dfrac{\sqrt3}{2}\right)$ and $Q\left(\dfrac{3\pi}{4},\dfrac{3\pi}{4}-\dfrac{\sqrt2}{2}\right)$, the curve $y=x-\sin{x}$ is concave up, so the curve lies to the right of line segment $PQ$.
We can show that $P$ is below $y=\dfrac{2\pi}{5}$, and $Q$ is above $y=\dfrac{2\pi}{5}$:

$\dfrac{2\pi}{3}-\dfrac{\sqrt3}{2}<\dfrac{2(3.15)}{3}-\dfrac{1.7}{2}=1.25<1.256=\dfrac{2(3.14)}{5}<\dfrac{2\pi}{5}$
$\dfrac{2\pi}{5}<\dfrac{2(3.2)}{5}=1.28<1.45=\dfrac{3(3)}{4}-0.8<\dfrac{3\pi}{4}-\dfrac{\sqrt2}{2}$

Let $a=$ x-coordinate of intersection of $y=\dfrac{2\pi}{5}$ and line segment $PQ$.

$\alpha >a=\dfrac{\frac{2\pi}{5}+\frac{9\sqrt3}{2}-4\sqrt2}{1+\frac{6}{\pi}(\sqrt3-\sqrt2)}>\dfrac{0.4(3.14)+4.5(1.732)-4(1.415)}{1+\frac{6}{3.14}(1.733-1.414)}=\dfrac{3.39}{\frac{5.054}{3.14}}>\dfrac{3.39}{1.61}>2.1$

$\huge{\text{Show that }\beta>2.82.}$

$2\times\text{Area}_{\text{segment }ADC}=\beta-\sin{\beta}=\dfrac{4\pi}{5}$

Between $R\left(\dfrac{5\pi}{6},\dfrac{5\pi}{6}-\dfrac{1}{2}\right)$ and $S\left(\pi,\pi\right)$, the curve the curve $y=x-\sin{x}$ is concave up, so the curve lies to the right of line segment $RS$.
We can show that $R$ is below $y=\dfrac{4\pi}{5}$, and $S$ is above $y=\dfrac{4\pi}{5}$:

$\dfrac{5\pi}{6}-\dfrac{1}{2}<\dfrac{5(3.6)}{6}-\dfrac{1}{2}=\dfrac{4(3)}{5}<\dfrac{4\pi}{5}$
$\dfrac{4\pi}{5}<\pi$

Let $b=$ x-coordinate of intersection of $y=\dfrac{4\pi}{5}$ and line segment $RS$.

$\beta >b=\dfrac{\frac{4\pi}{5}+3}{1+\frac{3}{\pi}}>\dfrac{0.8(3.1415)+3}{1+\frac{3}{3.1415}}>\dfrac{5.5132}{1.955}>2.82$

$\huge{\text{Show that }\angle{BAC}<0.6816}.$

$\angle{BAC}=\dfrac{1}{2}{\angle{BOC}}=\dfrac{1}{2}{\left(2\pi-\alpha-\beta\right)}<\dfrac{1}{2}(2(3.1416)-2.1-2.82)=0.6816$

$\huge{\text{Show that }\alpha<2.12.}$

$2\times\text{Area}_{\text{minor segment }AB}=\alpha-\sin{\alpha}=\dfrac{2\pi}{5}$

enter image description here

Between $(0,0)$ and $(\pi,\pi)$, the curve $y=x-\sin{x}$ is concave up, so the curve lies to the left of the tangent to the curve at $x=\dfrac{2\pi}{3}$.
Let $c=$ x-coordinate of intersection of $y=\dfrac{2\pi}{5}$ and the tangent to the curve at $x=\dfrac{2\pi}{3}$.

$\alpha<c=\dfrac{\sqrt3}{3}+\dfrac{22\pi}{45}<\dfrac{1.74}{3}+\dfrac{22(3.15)}{45}=2.12$

$\huge{\text{Show that }5-5\cos{\alpha}<7.62.}$

enter image description here

For $\dfrac{\pi}{2}<x<\pi$, the curve $y=5-5\cos{x}$ is concave down, so the curve lies below the tangent to the curve at $x=\dfrac{2\pi}{3}$, which has equation $y=\dfrac{5\sqrt3}{2}\left(x-\dfrac{2\pi}{3}\right)+7.5$.

$5-5\cos{\alpha}<5-5\cos{2.12}<\dfrac{5\sqrt3}{2}\left(2.12-\dfrac{2\pi}{3}\right)+7.5<5.3(1.733)-\dfrac{5(3.141)(1.732)}{3}+7.5<7.62$

$\huge{\text{Show that }\angle{BAC}>0.6817}.$

In the diagram of the circle and chords, $\text{Area}_{\text{triangle}}=\dfrac{\pi}{5}\text{ and }AB=2\sin{\dfrac{\alpha}{2}}\implies\angle{BAC}=\arctan{\left(\dfrac{2\pi}{5-5\cos{\alpha}}\right)}$
$\angle{BAC}=\arctan{\left(\dfrac{2\pi}{5-5\cos{\alpha}}\right)} > \arctan{\left(\dfrac{2\pi}{7.62}\right)}>\arctan{\left(\dfrac{2(3.14)}{7.62}\right)}>\arctan{0.82}$

enter image description here

Between $T\left(\sqrt{5-2\sqrt5}, \dfrac{\pi}{5}\right)$ and $U\left(1, \dfrac{\pi}{4}\right)$, the curve $y=\arctan{x}$ is concave down, so the curve lies above line segment $TU$, which has equation $y=\dfrac{\frac{\pi}{20}}{\sqrt{5-2\sqrt5}-1}(1-x)+\dfrac{\pi}{4}$.

$\angle{BAC}>\arctan{0.82}>\dfrac{\frac{\pi}{20}}{\sqrt{5-2\sqrt5}-1}(1-0.82)+\dfrac{\pi}{4}>\dfrac{\frac{3.141}{20}}{0.727-1}(0.18)+\dfrac{3.141}{4}>0.6817$

$\huge{\text{Contradiction}}.$

$\huge{\text{Therefore the answer is no}}.$

Dan
  • 22,158