4
$\begingroup$

Find the values of $x$ such that $\frac{11}{2} x - \cos x = 0$.

I really don't know how to find the solutions of this equation. I would appreciate if you could help me.

$\endgroup$
  • 2
    $\begingroup$ The question basically asks to find the intersection point of two graphs.........y=(11/2)x and y=cosx $\endgroup$ – Sujan Dutta Aug 19 '16 at 5:13
  • $\begingroup$ @SujanDutta ... and that's obviously solvable how? $\endgroup$ – fleablood Aug 19 '16 at 5:15
14
$\begingroup$

An equation which mix polynomial and trigonometric terms do not show explict solutions and numerical methods should be required.

If you consider the function $$f(x)=\frac{11}2x-\cos(x)$$ its derivative $$f'(x)=\frac{11}2+\sin(x)$$ is always positive so only one solution.

Apply Newton method which, starting from a guess $x_0$, will update it according to $$x_{n+1}=x_n-\frac{f(x_n)}{f'(x_n)}$$ Let us be lazy and start using $x_0=0$. So, this simple method will generate the following iterates $$\left( \begin{array}{cc} n & x_n \\ 1 & 0.18181818181818181818 \\ 2 & 0.17891658537664712943 \\ 3 & 0.17891585607026614457 \\ 4 & 0.17891585607022005433 \end{array} \right)$$

Quite fast, isn't it ? You will learn that we could even do it faster.

Edit for your curiosity

Beside Newton, there are higher order iterative methods. Using them, still using $x_0=0$, the first iterate for a method of order $k$ is given in the following table $$\left( \begin{array}{ccc} k & x_1^{(k)} & \approx & \text{method} \\ 2 & \frac{2}{11} & 0.1818181818 &\text{Newton}\\ 3 & \frac{238}{1331} & 0.1788129226 &\text{Halley} \\ 4 & \frac{28814}{161051} & 0.1789122700 &\text{Householder}\\ 5 & \frac{10459726}{58461513} & 0.1789164437 &\text{no name}\\ 6 & \frac{1265622382}{7073843073} & 0.1789158127 &\text{no name}\\ 7 & \frac{2297105198362}{12839025177495} & 0.1789158574 &\text{no name}\\ 8 & \frac{5053631400166}{28245855390489} & 0.1789158562 &\text{no name}\\ 9 & \frac{21402128966834974}{119621197578720915} & 0.1789158561&\text{no name} \end{array} \right)$$

Moreover, "knowing" that $x$ has to be be quite small, we could have used the Taylor expansion of $\cos(x)$ around $x=0$ and limiting to $O\left(x^3\right)$ write $$f(x)\approx \frac{x^2}{2}+\frac{11 x}{2}-1$$ the solution of which being $x=\frac{\sqrt{129}-11}{2} \approx 0.178908$ which is already quite good.

Even better, using Padé approximants around $x=0$ and writing $$f(x)\approx \frac{P_{2}(x)}{Q_{n}(x)}$$ and solving $P_2(x)=0$ leads to quite accurate results as shown i the following table $$\left( \begin{array}{ccc} n & x & \approx \\ 2 & \frac{77-\sqrt{3265}}{111} & 0.1789173107 \\ 3 & \frac{3 \left(-1199+\sqrt{1645097}\right)}{1402} & 0.1789158655 \\ 4 & \frac{118635-\sqrt{13560037665}}{12226} & 0.1789158556 \\ 5 & \frac{-1624095+\sqrt{6130243687065}}{4761109} & 0.1789158560 \end{array} \right)$$

$\endgroup$
  • $\begingroup$ Thanks alot, but i was actually helping to get an analitical solution, not a numerical one. But it's helpful, thanks. $\endgroup$ – Oscar Gutierrez Aug 20 '16 at 20:03
  • 3
    $\begingroup$ @OscarGutierrez. Even $x=\cos(x)$ does not show analytical solution. $\endgroup$ – Claude Leibovici Aug 21 '16 at 4:35
10
$\begingroup$

We can approximate the value of $x$ without Newton's Method or any calculus at all. We can generalize the idea behind solving for $x$ in $x = \cos(x)$ by repeatedly hitting the cosine button on a calculator and noticing that it converges.


Rearranging the equation a bit, we are looking for the $x$ such that $$ x = \frac{2}{11}\cos(x)\;. $$ Intuitively we know that such a value of $x$ exists since the graph of $y=\cos(x)$ and the line $y=\frac{11}{2}x$ intersect.

Now let's consider the function $f$ defined as $f(\_) = \frac{2}{11}\cos(\_)$. Suppose that we repeatedly apply $f$ to some input $a$, and let this infinite composition equal $x$. So we are going to say $$\begin{align} &f\left(f\left(\dotsb f(a)\right)\dotsb\right) \\=\;\;& \frac{2}{11}\cos\left(\frac{2}{11}\cos\left(\dotsb \frac{2}{11}\cos(a)\right)\dotsb\right) \\=\;\;& x \end{align}$$ But if we suppose this converges as the number of times we apply $f$ approaches infinity (so $x$ actually exists and is a number), applying $f$ one more time shouldn't change the value of anything. So we now have $$\begin{align} &x \\=\;\;& f\left(f\left(\dotsb f(a)\right)\dotsb\right) \\=\;\;& f\left(f\left(f\left(\dotsb f(a)\right)\dotsb\right)\right) \\=\;\;& \frac{2}{11}\cos\left(\frac{2}{11}\cos\left(\frac{2}{11}\cos\left(\dotsb \frac{2}{11}\cos(a)\right)\dotsb\right)\right) \\=\;\;& \frac{2}{11}\cos(x) \end{align}$$

So the $x$ that this may converge to, the $x$ that we are looking for, is precisely the $x$ the solves the equation $x = \frac{2}{11}\cos(x)$. Furthermore we know such an $x$ exists by the reasoning of intersecting graphs above! And since there is no $a$ in the equation $x = \frac{2}{11}\cos(x)$, that must mean the value of $x$ doesn't depend on $a$. This means we can approximate the value of $x$ in $x = \frac{2}{11}\cos(x)$ by taking any number and repeatedly applying $f$ to it.

We can now easily calculate this value of $x$ by programming the function $f(\_) = \frac{2}{11}\cos(\_)$ into a calculator repeatedly apply $f$ to some (any) input. Calculating $f^n$ for different values of $n$ (mostly to compare to Newton's Method in Claude Leibovici's answer) I get:

\begin{array}{c|c} n & f^n(1) \\\hline 1 & 0.\color{#A00}{0982367828} \\ 2 & 0.1\color{#A00}{809415720} \\ 3 & 0.178\color{#A00}{8499431} \\ 4 & 0.17891\color{#A00}{79884} \\ 5 & 0.178915\color{#A00}{7870} \\ 6 & 0.17891585\color{#A00}{83} \\ \vdots & \\ 100 & 0.1789158560 \end{array}

So this doesn't converge as quickly as Newtons' Method, but I think that this can be punched into a calculator much quicker.

$\endgroup$
  • 1
    $\begingroup$ Nice, reminds me when I discovered this when repeatedly pressing $\sqrt{Ans+1}$ on my calculator. math.stackexchange.com/questions/1599163/…. This is called the fixed point theorem, which you most likely are already are aware of. This link might be of interest to the OP. $\endgroup$ – Ahmed S. Attaalla Aug 21 '16 at 6:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.