Is it possible to solve for x the following equation without root finding:
$$\frac{\sin(x)}{x}=\frac{5}{6}$$
$\begingroup$
$\endgroup$
-
3$\begingroup$ Wolfram alpha just does it numerically...getting $x\sim \pm 1.02674$. I expect there's no simpler form. $\endgroup$ – lulu May 14 '16 at 23:27
-
$\begingroup$ I'm surprised how challenging such simple equations! $\endgroup$ – ZeroCool May 14 '16 at 23:35
-
$\begingroup$ Exactly? Don't think so. But there are plenty of nice numerical methods, Taylor series approximations and so on. But you're aware of that already. $\endgroup$ – KR136 May 14 '16 at 23:39
-
$\begingroup$ It's not as surprising as you might think. There's a lot of numbers out there. Not all of them are as simple as, let's say, solutions to polynomials with integer coefficients. $\endgroup$ – Lee Mosher May 14 '16 at 23:41
-
$\begingroup$ If you assume small angles then you can obtain $x=\pm 1$ fairly fast else it's a bit grim as stated. You can't avoid finding roots though because even to lowest order the LHS expansion is $1 - x^2/3! + \mathcal{O}(x^4)$ $\endgroup$ – Alexander McFarlane May 15 '16 at 0:40
$\begingroup$
$\endgroup$
There is no closed form solution, but since $\sin(x)$ is a concave function on $(0,\pi)$ and $\sin'(0)=1$, there is a unique real and positive solution to $\sin(x)=\frac{5}{6}x$. We may use a step of Newton's method with starting point $x_0=1$ get a tight approximation of such solution: $$ x_1\approx 1-\frac{\sin 1-5/6}{\cos 1-5/6}\approx 1.02777. $$ An even better starting point is given by the solution of $$ \exp\left(\frac{5x^2}{x^2-30}\right)=\frac{5}{6} $$ i.e. by $$ x_0 = \sqrt{\frac{30 \log\left(\frac{6}{5}\right)}{5+\log\left(\frac{6}{5}\right)}}\approx 1.0273477$$ or by the solution of $$ \frac{60-7x^2}{60+3x^2}=\frac{5}{6}, $$ i.e. by $$ x_0 = 2\sqrt{\frac{5}{19}} \approx 1.026. $$ By using the Padé approximant in the origin $\frac{p(x)}{q(x)}$ with $\partial p=\partial q=4$, we get the almost exact solution $$ x\approx \sqrt{\frac{2}{977} \left(13405-\sqrt{166152805}\right)}.$$
answered May 14 '16 at 23:46
-
-
$\begingroup$ Hi Jack ! How did you arrive to the better starting point ? $\endgroup$ – Claude Leibovici May 15 '16 at 6:17
-
$\begingroup$ @ClaudeLeibovici: the first approximation was $\frac{\sin x}{x}\approx e^{-x^2/6}$, then I took an approximant for $\log\frac{\sin x}{x}$ capable to grasp the $x^4$ term, too. $\endgroup$ – Jack D'Aurizio May 15 '16 at 12:56
-
$\begingroup$
$\endgroup$
$$x=6/5\cdot\sin(6/5\cdot\sin(6/5\cdot \sin(6/5\cdot\sin(\cdots))))$$
But from convergence point of view Newton is faster
$\begingroup$
$\endgroup$
Equations which mix polynomial and trigonometric terms do not show analytical solutions (this is already the case for $x=\cos(x)$) and numerical methods should be used.
However, some rather good approximations can be made and, for your curiosity, I give you the links to two questions of mine (here and here).
Using the beautiful $$\sin(x) \simeq \frac{16 (\pi -x) x}{5 \pi ^2-4 (\pi -x) x}\qquad (0\leq x\leq\pi)$$ proposed by Mahabhaskariya of Bhaskara I, a seventh-century Indian mathematician, your equation becomes $$\frac{16 (\pi -x) x}{5 \pi ^2-4 (\pi -x) x}=\frac 56 x$$ which reduces to a simple quadratic $$-\frac{10 x^2}{3}+\left(\frac{10 \pi }{3}-16\right) x-\frac{25 \pi ^2}{6}+16 \pi=0$$ for which the solution to keep is $$x=\frac{1}{40} \left(-96+20 \pi +\sqrt{9216+3840 \pi -1600 \pi ^2}\right)\approx 1.02288 $$ while the exact solution, obtained using Newton method, would be $\approx 1.02674$.
Now, knowing that the solution is close to $1$, we could approximate $\sin(x)$ using its simplest Padé approximant built at $x=1$; this should give $$\sin(x)=\frac{\frac{1}{2} (x-1) (2 \cos (1)+\sin (1) \tan (1))+\sin (1)}{1+\frac{1}{2} (x-1) \tan (1)}$$ which, again, reduces to a quadratic the solution of which being $\approx 1.02676$.
Another way could be to use the Taylor series of $\sin(x)$ around $x=\frac \pi 3$ which is $$\sin(x)=\frac{\sqrt{3}}{2}+\frac{1}{2} \left(x-\frac{\pi }{3}\right)-\frac{1}{4} \sqrt{3} \left(x-\frac{\pi }{3}\right)^2-\frac{1}{12} \left(x-\frac{\pi }{3}\right)^3+O\left(\left(x-\frac{\pi }{3}\right)^4\right)$$ Limiting to first order, the solution would then be $$x_{(1)}=\frac{1}{2} \left(3 \sqrt{3}-\pi \right)\approx 1.02728$$ Limiting to second order, the solution would then be $$x_{(2)}=\frac{1}{3} \left(-\frac{2}{\sqrt{3}}+\pi +\sqrt{\frac{2}{3} \left(29-5 \sqrt{3} \pi \right)}\right)\approx 1.02674$$
answered May 15 '16 at 4:20
