4

Problem :

Find a geometric construction or a proof by hand to show :

$$\sin\left(1+\frac{1+\sqrt{5}}{2}\right)<1/2$$

As attempt I introduce the inequality :

$$\sin\left(1+\frac{1+\sqrt{5}}{2}\right)-1+\frac{1+\sqrt{5}}{2}-\frac{559}{500}<0$$

Then I introduce :

$$f(x)=\sin\left(1+x\right)-1+x-\frac{559}{500}$$

Or :

$$g(x)=\sin\left(x\right)+x-2-\frac{559}{500}$$

Then we can use power series around $x=610/233$ but it's tedious by hand .

How to solve the problem ?

6 Answers6

10

Since $\dfrac{1}{2}=\sin\dfrac{5\pi}{6}$ and $y=\sin x$ is strictly decreasing for $x\in\left(\dfrac{\pi}{2}, \pi\right)$, it suffices to prove that

$$ 1+\frac{1+\sqrt{5}}{2}>\frac{5\pi}{6} $$

Since $2.236^2=4.999696<5$, we know that $2.236<\sqrt{5}$, and hence $1+\dfrac{1+\sqrt{5}}{2}>2.618$. Hence, we have

$$\begin{align} 1+\frac{1+\sqrt{5}}{2} &>2.618 \\ &>\frac{1775}{678} & \text{ since }2.618=\frac{1775.004}{678}\\ &=\frac{5}{6}\cdot\frac{355}{113} \\ &>\frac{5\pi}{6} \end{align}$$

Yes there are some steps which have to be verified by multiplication, but definitely doable by hand.

IraeVid
  • 3,216
  • 1
    Out of curiosity, how would you get $\phi$ to three decimal places by hand? When I multiply them out I get that approximation times $678$ is $1775.004$ which is very close, so this seems like a crucial step I would want to see (although it looks like rounding to two decimal places also suffices). I did not downvote by the way, I am neutral at the moment. – Ninad Munshi Jul 02 '23 at 08:05
  • that still doesn't answer my question how – Ninad Munshi Jul 02 '23 at 08:15
  • 1
    Since $2.236^2=4.999696<5$ we know that $2.236<\sqrt{5}$. Plug this back into the equation and the conclusion follows. – IraeVid Jul 02 '23 at 08:21
  • 3
    One way is to notice that $\phi$ solves $\phi^2-\phi-1 = 0$ and apply to ${p \over q}$ with $p=1618, q=1000$. Expanding requires showing that $p^2-pq-q^2 = -76 < 0$. – copper.hat Jul 02 '23 at 08:23
  • 1
    +1 Excellent, would you edit either of these methods into the body of your post? Thank you for your contribution – Ninad Munshi Jul 02 '23 at 08:24
  • 1
    @NinadMunshi, sure, I have edited my post. – IraeVid Jul 02 '23 at 08:26
4

Note that $\phi^2=1+\phi$. It suffices to prove that $$ \phi^2>\frac{5\pi}{6}\implies \frac{6}{5}\phi^2>\pi. $$

By applying the cosine law to the golden triangle, we find $$ \frac{6}{5}\phi^2=\frac{3}{5(1-\cos(\frac{\pi}{5}))} =\frac{3}{5\left(1-\frac{\sqrt5+1}{4}\right)} =\frac{3}{5}(3+\sqrt5). $$ Different proofs that $\cos(\pi/5)=\frac12\phi$ can be found here.

By taking $\sqrt5$ to four decimal places ($\sqrt{5}\approx2.2360$, rounding down) and doing a little unpleasant computation by hand, it can be shown that $$ \frac{3}{5}(3+\sqrt5)>\frac{3}{5}(3+2.2360)=3.1416>\pi. $$ Hope this helps!

This answer is just an extension of my answer to this question, which asks why $\frac{6}{5}\phi^2\sim\pi$.

Scene
  • 1,566
  • The most part of the answer covers the fact that $(6/5)$ of the argument of $\sin$ can be written $\frac{3}{5}(3+\sqrt{5})$, but this can be obtained trivially without recurring to $\phi$, while the most important part, the inequality, provides no particular hint to prove them. – Vincenzo Tibullo Jul 02 '23 at 09:28
  • Sorry, I can't quite catch what you mean here by the $(6/5)$ argument of $\sin$ can be written $\frac{3}{5}(3+\sqrt{5})$. If there is some trivial way to obtain this result, please let me know. I have a terrible knack for overcomplication and circular reasoning! – Scene Jul 02 '23 at 11:35
  • $$ \frac{6}{5}\left(1+\frac{1+\sqrt{5}}{2}\right)=\frac{6}{5}\cdot\frac{3+\sqrt{5}}{2}=\frac{3}{5}\left(3+\sqrt{5}\right) $$ – Vincenzo Tibullo Jul 02 '23 at 12:05
1

First, notice that $${3(3+2.236)} = {5\cdot 3.1416}$$

Now, recall we need to prove $$\sin\left(1+\frac{1+\sqrt5}{2}\right)<\sin\frac{5\pi}{6}$$ which as Michael Rozenberg and others pointed out, is logically equivalent to showing $$1+\frac{1+\sqrt5}{2}>\frac{5\pi}{6}$$ Multiplying both sides by $6/5\pi$ and simplifying is equivalent to $$\frac{3(3+\sqrt5)}{5\pi}>1$$

Since $\sqrt 5 > 2.236$ and $\pi < 3.1416$, we have $$\frac{3(3+\sqrt5)}{5\pi} >\frac{3(3+2.236)}{5\cdot3.1416} = 1\quad\quad \blacksquare$$

Snared
  • 865
1

Here is another that avoids $\pi$ approximations other than $3 < \pi < 4$ and takes liberty with the notion of calculating by hand.

For $x \ge 0$ we have $\sin x \le u(x)=\sum_{k=1}^7 (-1)^{n+1} {x^{2k+1} \over (2k+1)!}$, so in principle we can upper bound $\sin$ when calculating with rationals.

Using $2 < \sqrt{5} < 3$ we have ${\pi \over 2} < 2 <{5 \over 2} < 1+ \phi < 3 < \pi$ and $\sin$ is decreasing on the interval $({\pi \over 2}, \pi)$.

If we can find a rational ${p \over q} \in (2,3)$ such that ${p_n \over q_n} \le 1+\phi$ and we have $u({p_n \over q_n}) < {1 \over 2}$ then we are finished.

Noting that $q(x)= x^2-x-1$ is increasing and non negative for $x \ge 0$, $q(\phi) = 0$ and $q({1618 \over 1000}) < 0$, we see that ${2618 \over 1000}$ satisfies the criteria and $u({2618 \over 1000}) = {4132642497822341133247468087 \over 9278984069824218750000000000} < {1 \over 2}$.

copper.hat
  • 172,524
0

I assume we know

  1. $\sqrt{5}>2.236$

  2. $\pi<3.141593$

  3. $\sin\frac{\pi}{6}=\frac{1}{2}$

  4. $\sin$ is strictly increasing on $\left[0,\pi/2\right]$

Now, $\sin\left(1+\frac{1+\sqrt{5}}{2}\right)=\sin\left(\pi-\left(1+\frac{1+\sqrt{5}}{2}\right)\right)$.

We can verify by hand that $\pi-\left(1+\frac{1+\sqrt{5}}{2}\right)\approx0.5\in\left[0,\pi/2\right]$.

So, by (4), it suffices to show that $$\pi-\left(1+\frac{1+\sqrt{5}}{2}\right)<\frac{\pi}{6}\text{ or }\frac{5}{6}\pi-\left(1+\frac{1+\sqrt{5}}{2}\right)<0.$$ But this last inequality holds because

$$\frac{5}{6}\pi-\left(1+\frac{1+\sqrt{5}}{2}\right)<\frac{5}{6}\left(3.141593\right)-\left(1+\frac{1+2.236}{2}\right)=-0.0000058\dot{3}.$$

  • 1
    Why even post this answer? It doesn't add anything on top of my answer, which follows the same approach except yours requires a better approximation of $\pi$. – Snared Jul 02 '23 at 12:25
-3

We need to prove that $$\sin\left(1+\frac{1+\sqrt5}{2}\right)<\sin\frac{5\pi}{6}$$ or $$1+\frac{1+\sqrt5}{2}>\frac{5\pi}{6}.$$ Can you solve the question from here?

IraeVid
  • 3,216
  • 3
    You do realise that the difference of the latter is around $4 \times 10^{-5}$, right? – copper.hat Jul 02 '23 at 07:35
  • 3
    -1 That is the problem isn't it? But how exactly do you go about showing that? – Ninad Munshi Jul 02 '23 at 07:36
  • @copper.hat Yes, of course. There are very many good approximations of $\pi.$ – Michael Rozenberg Jul 02 '23 at 07:37
  • @Ninad Munshi We can use $\frac{\pi}{4}=\arctan\frac{1}{2}+\arctan\frac{1}{3}$ and the Taylor series for $\arctan$. Or something similar. Try, it's not so hard. – Michael Rozenberg Jul 02 '23 at 07:50
  • 3
    If it's not so hard, it would be beneficial to OP to edit that process in :) It's the same reasoning I gave the other answer. But you should try it to verify it is tractable and not send OP on a wild goose chase that still leaves them in the lurch. I have made the assumption (incorrectly) many times that just because known numerical approximations existed my work was done. – Ninad Munshi Jul 02 '23 at 07:54
  • @Ninad Munshi I see now, you did not try. I think, it's not fair. If a topic-starter will want a full way, I'll show it, of course. – Michael Rozenberg Jul 02 '23 at 07:59
  • 2
    This post is being unreasonably downvoted. He is giving a hint to OP, and since he knows power series etc, he should be able to figure out on his own. – IraeVid Jul 02 '23 at 08:23
  • 3
    @IraeVid respectfully, I disagree. As it stands, this answer amounts to the first few words on your answer, and doesn't give any hints to the technical heart of the problem which is tangible numerical approximations. Without even a hint at that second piece, this answer is more appropriate as a comment. – Ninad Munshi Jul 02 '23 at 08:33
  • This answer is not wrong, but given its length, I would agree with @Ninad Munshi its more of hint, which respectfully resides within the comments (not as an answer). However, I would not downvote, rather leave it $0$ (neutral). – Dstarred Jul 02 '23 at 09:00