4

The precise question through which I came across this particular estimation is as follows:

If $x=\alpha$ is the maximum value of $x$ for which $ \left \lfloor{\sin^{-1}(\cos^{-1}(\tan^{-1}x))} \right \rfloor = 1$, then $\alpha$ lies in the interval:

A) $\left(0,\dfrac{1}{\sqrt3} \right)$

B) $\left(\dfrac{1}{\sqrt3},1 \right)$

C) $\left(1,\sqrt3 \right)$

D) $(\sqrt3, \infty)$


Finding the value of $\alpha$ isn't hard. The part where I'm stuck in is the estimation of its interval. $$ \begin{align} &\ \left \lfloor{\sin^{-1}(\cos^{-1}(\tan^{-1}x))} \right \rfloor = 1 \\ \implies &\ \sin^{-1}(\cos^{-1}(\tan^{-1}x)) \in \left[1,\frac{\pi}{2}\right] \\ \implies &\ \cos^{-1}(\tan^{-1}x) \in \left[\sin1,1\right] \\ \implies &\ \tan^{-1}x \in [\cos1,\cos(\sin1)] \\ \implies &\ x \in [\tan(\cos1),\tan(\cos(\sin1))] \end{align} $$ From here, it's very clear that $\alpha =\tan(\cos(\sin1))$ but now I am supposed to predict the interval in which it lies. The options seem to represent the numerical values of $\tan0, \tan\frac{\pi}{6},\tan\frac{\pi}{4},\tan\frac{\pi}{3},\tan\frac{\pi}{2} $.

So the problem essentially comes down to predicting whether $\cos(\sin1)$ lies in $\left(0, \frac{\pi}{6}\right), \left(\frac{\pi}{6}, \frac{\pi}{4}\right),\left(\frac{\pi}{4}, \frac{\pi}{3}\right), \left(\frac{\pi}{3}, \frac{\pi}{2}\right) $

Through rough estimation, I was able to arrive at (B), which is also the correct answer. $$ \frac{\pi}{4} < 1 < \frac{\pi}{3} $$ $$ \frac{1}{\sqrt2} < \sin1 < \frac{\sqrt3}{2} $$ $$ \begin{align} \cos(\sin1) &\ \in (\cos0.866, \cos0.707) \\ &\ \in (\cos49.6^{\circ}, \cos40.5^{\circ}) \\ &\ \approx \cos45^{\circ} \\ &\ = \frac{1}{\sqrt2} \in \left(\frac{\pi}{6}, \frac{\pi}{4}\right) \end{align}$$

I guess my approximations were a bit too rough. But anyway, what I am looking for is a concrete mathematical way to prove that $\cos(\sin1)$ lies in the interval $\left(\frac{\pi}{6}, \frac{\pi}{4}\right)$ without any inconclusive approximations.

Ankit Saha
  • 1,802
  • A concrete way would be to find the exact value, which I don't think you want. Don't think that approximations are not mathematical, approximations are also a part of maths. – ultralegend5385 Aug 25 '21 at 16:24
  • 1
    @ultralegend5385 I am fully aware that approximation is also a part of math, in fact, it is an entire branch of math. I don't mind reasonable approximations or approximations that can conclusively decide the answer. In the last line, I was referring to approximations similar to what I made. I got that cos(sin1) ∈ (cos49.6°, cos40.5°) which is approximately cos45° ∈ (π/6, π/4) but this does not imply that cos(sin1) ∈ (π/6, π/4) because cos40.5° can still be greater than π/4 and cos49.6° can still be less than π/6. So, this approximation isn't conclusively implying anything. – Ankit Saha Aug 25 '21 at 16:59

1 Answers1

3

I think this might be what you're looking for. $$\sin(1)\approx\sin(57^\circ)\approx\sin(60^\circ)$$ Infact, this approximation has an error of $3\%$. (Keep in mind that this approximation overestimates.)

Then, you see our final expression becomes $\cos(\sqrt{3}/2)$, i.e. $\cos(0.866)$. Notice that $$\dfrac{\pi}{4}<0.866<\dfrac{\pi}{3}$$ implying ($E$ is our expression $\cos(\sin 1)$). $$\dfrac{\sqrt{2}}{2}>E>\dfrac12$$ Now, you notice that $\dfrac12\approx\dfrac{\pi}{6}$ (error $4\%$). Also, $\pi/4\approx 0.78$ so you're sure that $E<\pi/4$. This was essentially your problem. We don't know if $E$ be less than $\pi/6$, being greater than $1/2$.

Well, in such cases, either you can just look at the options and mark your answer. Or you look at your inequalities and see that $0.866$ lies somewhere in between and not close to one bound. So, your final inequality, $E$ must also be in between and so it isn't harm to replace the bound by something just greater than it. Or you can also see that you overestimated in your first line and so in the final one, the actual quantity should be slightly closer to $\sqrt{2}/2$.

Hope this helps. Ask anything if not clear :)

  • For what it's worth, I worked on this for several minutes a few days ago (when the question was asked) and I didn't discover anything that seemed especially cute or clever that was also elegant and rigorous, so I abandoned it, thinking someone else would probably come up with something along those lines. However, I forgot to check back here, and only now happened by chance to see your answer. I don't remember exactly the things I tried, but probably one was similar to this, and if not, certainly none were any better. I suspect this problem doesn't have a mathematically cute solution. – Dave L. Renfro Aug 31 '21 at 17:15
  • @Dave: I also suspected their might be one, and I still do. Hopefully someone (or even you or I) will be able to find an elegant way, which we don't see as of now. – ultralegend5385 Sep 01 '21 at 00:14