6
$\begingroup$

I am asked to provide an iterative algorithm which would lead to finding a real root of this polynomial:

$$6x^5-x^3+6x-6=0$$ It is required to rely on the contraction mapping principle and Banach fixed-point theorem.

At the moment I think I can rewrite $$f(x) = \sqrt[5]{x^3/6-x+1}$$ and try to prove that in a complete metric space $\langle \, \mathbb{R}, d \, \rangle$ I have a contraction mapping $f:\mathbb{R} \rightarrow \mathbb{R}$ and consequently a unique fixed-point is my solution. I already see it's going to be messy (contraction mapping proof part) and the idea of it makes me sick...

Besides those 6'es in the initial polynomial temps to rewrite $$ x = \sqrt[3]{6} \sqrt[3]{x^5+x-1}$$ but I see no good coming out of it.

Since I'm very new to functional analysis and metric spaces I decided to ask you for suggestions about the most suave way to do it.

The final answer (fixed point) will be $x \approx 0.78$ so I'd love to get the contraction mapping over $[0;1]$, $[-1;1]$, $[0;2]$, $[-2;2]$ or something like that which would be easy to prove (contractility that is) without calculator... But I can't find it!

$\endgroup$
  • $\begingroup$ Keep in mind that you can also write $x=1+\frac{x^3}{6}-x^5$ - you don't necessarily have to show that you have a contraction mapping over all of $\mathbb{R}$, but just over an interval where you know that there must be a root... $\endgroup$ – Steven Stadnicki Mar 8 '13 at 22:10
  • $\begingroup$ @StevenStadnicki Thanks for the suggestion. It seems that a $[0;1]$ interval could do here which is pretty good. But how did you realize that particularly this expression of $x$ is advantageous? $\endgroup$ – Pranasas Mar 8 '13 at 22:17
  • $\begingroup$ @Pranasas, if you let $F(x)=1+\frac{x^3}{6}-x^5$, then the problem is to find a fixed point of $F$. Thats why steven used this expression. $\endgroup$ – Tomás Mar 8 '13 at 22:42
  • $\begingroup$ @Pranasas For a few reasons: (a) it was the one reorganization of the original equation you hadn't looked at; (b) it's the simplest - it remains polynomial, rather than being an awkward root expression; and (c) over the region near $x=0$ then the powers of $x$ are going to be smaller than $x$ is, so it seemed plausible that it might be a contraction mapping. $\endgroup$ – Steven Stadnicki Mar 8 '13 at 23:32
  • $\begingroup$ @StevenStadnicki So what interval would you suggest? The fixed point is at $x \approx 0.78$. I realized $[0;1]$ is no good because for example $f(x)=1+x^3/6-x^5$ at $x= \sqrt{0.1}$ is not within $[0;1]$. Another option $[0;2]$ is not good either. $[0;1.1]$ would pretty much require a calculator to prove and I hate it. $\endgroup$ – Pranasas Mar 9 '13 at 11:40
3
$\begingroup$

Taking the fifth root was a good idea: It helps to make derivatives small.

Consider the polynomial $$p(x):={x^3\over 6}-x+1$$ on the interval $X:=[0,1]$. From $p'(x)={x^2\over2}-1<0$ $\ (x\in X)$ it follows that $$ p(x)\geq p(1)={1\over6}\qquad(x\in X)\ .\tag{1}$$ Now define $$f(x):=\bigl(p(x)\bigr)^{1/5}\qquad(x\in X)\ .$$ Then $$f'(x)={1\over5}\bigl(p(x)\bigr)^{-4/5}\biggl({x^2\over2}-1\biggr)\ .\tag{2}$$ It follows that $f'(x)<0$ for all $x\in X$. Therefore $$0<{1\over 6^{1/5}}=f(1)\leq f(x)\leq f(0)=1\qquad(x\in X)\ ,$$ which proves that $f$ maps the complete metric space $X$ into itself.

To finish the proof we need an estimate of $|f'(x)|$. From $(1)$ and $(2)$ it follows that $$|f'(x)|\leq{1\over5}6^{4/5}<0.84\qquad(x\in X)\ .$$ Therefore we may conclude that in the interval $X$ there is exactly one point $\xi$ such that $f(\xi)=\xi$. It follows that $p(\xi)=\xi^5$, or that $\xi$ is a root of the polynomial $q(x)=6x^5-x^3+6x-6$.

$\endgroup$

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.