0

Is there an analytical solution of this function?

$x^2 - 1 = \sin(x)$

I could only find numerical solutions for this equation.

amWhy
  • 209,954
Natjo
  • 131
  • 1
    I'm reasonably sure it doesn't have analytic solutions, but don't know how to prove it off the top of my head and don't have time to work out a proof right now. – Stella Biderman Mar 22 '17 at 21:41
  • 1
    It's not a function, but an equation. But indeed, it is quite clear you are asking about the existence, or lack thereof, of analytical solutions. – amWhy Mar 22 '17 at 21:45
  • See here: http://math.stackexchange.com/questions/2082103/trigonometric-polynomial-equations-and-the-algebraic-nature-of-trig-functions – Simply Beautiful Art Mar 22 '17 at 21:49

3 Answers3

4

There are analytic results for this equation involving summations over all the solutions of this equation in the complex plane. We can write the equation as $f(z) = 0$ with $f(z) = \sin(z) + 1 - z^2$. We can then consider the contour integral:

$$I(R) = \oint_{C(R)} \frac{1}{z^n}\frac{f'(z)}{f(z)}dz$$

where $C$ is a contour with the origin as center and radius R. Now, in this case for $n>1$, $I(R)$ tends to zero of we send $R$ to infinity such that the contour avoids the zeros of $f(z)$ . The contour integral can also be computed by summing all the residues at the poles of the integrand, which yields the summation of the reciprocals of the $n$th power of the roots plus the residue at $z = 0$. The summation over all the roots is thus minus the residue at $z = 0$.

We can compute the residue at $z=0$ by computing the series expansion of $\log\left[f(z)\right]$:

$$\log\left[f(z)\right] = z - \frac{3}{2}z^2 + \frac{7}{6}z^3 - \frac{19}{12}z^4 + \frac{15}{8}z^5 + \mathcal{O}(z^6)$$

The residue of the integrand at $z = 0$ is the coefficient of $z^{-1}$ in the expansion of the integrand, which is $n$ times the coefficient of $z^n$ of $\log\left[f(z)\right]$. Denoting the roots of $f(z)$ by $\alpha_j$, we thus have:

$$ \begin{split} \sum_{j}\frac{1}{\alpha_j^2} &= &3\\ \sum_{j}\frac{1}{\alpha_j^3} &= -&\frac{7}{2}\\ \sum_{j}\frac{1}{\alpha_j^4} &= &\frac{19}{3}\\ \sum_{j}\frac{1}{\alpha_j^5} &= -&\frac{75}{8} \end{split} $$

Count Iblis
  • 10,366
  • I'm impressed by your answer, unfortunately it seems as I don't understand everything. I can follow your reasoning, but when you say there are analytic results, how do you mean this? Since I can't read out a certain number as a solution from your answer. – Natjo Mar 27 '17 at 14:10
  • @Jonas So, the analytic results are on the summations over all the roots, which is similar to the case of polynomial equations. E.g. there doesn't exist an analytic solution for the general 5th degree equation (in terms of root extractions), but you can write down exact results for such things as the sum of all the roots, the sum of all squares of roots, the sum of all reciprocals etc. etc. In this case there are an infinite number of roots and the listed summations converge and we can write down exact expressions for them. – Count Iblis Mar 27 '17 at 20:02
0

If $f(x) = x^2 - 1 - \sin(x) $, since $0 \le 1+\sin(x) \le 2$, then $f(x) > 0$ for $x > \sqrt{2}$ or $x < -\sqrt{2}$.

Numerically, according to Wolfy, there are only two real root s at $x≈-0.636733$ and $x≈1.40962$.

marty cohen
  • 107,799
0

As alreasy said in comments and answer, there is no analytical solutions.

However, the function can be approximated using Pade approximants. To keep the problem simple, let us use degree $2$ for numerator and degree $n$ for denominator.

For $n=1$, the approximant would be $$\frac{\frac{7 x^2}{6}-\frac{5 x}{6}-1}{1-\frac{x}{6}}\implies x_{1,2}=\frac{1}{14} \left(5\pm\sqrt{193}\right)$$ For $n=2$ $$\frac{\frac{47 x^2}{42}-\frac{6 x}{7}-1}{\frac{x^2}{42}-\frac{x}{7}+1}\implies x_{1,2}=\frac{1}{47} \left(18\pm\sqrt{2298}\right)$$ For $n=3$ $$\frac{\frac{1043 x^2}{940}-\frac{203 x}{235}-1}{\frac{11 x^3}{2820}+\frac{5 x^2}{188}-\frac{32 x}{235}+1}\implies x_{1,2}=\frac{2 }{1043}\left(203\pm\sqrt{286314}\right)$$ I give below a table od the decimal representation of the solutions as functions of $n$ $$\left( \begin{array}{ccc} n & x_1 & x_2 \\ 0 & -0.61803 & 1.61803 \\ 1 & -0.63518 & 1.34946 \\ 2 & -0.63697 & 1.40292 \\ 3 & -0.63679 & 1.41531 \\ 4 & -0.63674 & 1.40812 \\ 5 & -0.63673 & 1.40966 \end{array} \right)$$ to be compared to the solutions given by marty cohen in his answer.