I am learning Metric Fixed Point Theory by Mohammed A Khamsi and William A Kirk.

I need help in understanding a step in the proof of the following theorem(Chapter 3, Theorem 3.2, Page No. 43):

Let $(M,d)$ be a complete metric space, let $T:M\rightarrow M$ be a contraction mapping with Lipschitz constant $k\in(0,1)$, and suppose $x_0\in M$ is the fixed point of $T$. Let $\{\epsilon_n\}$ be a sequence of positive numbers for which $\lim_{n\rightarrow\infty}\epsilon_n=0$, let $y_0\in M$, and suppose $\{y_n\}\subset M$ satisfies $d(y_{n+1},T(y_n))\leq \epsilon_{n}$. Then $\lim_{n\rightarrow\infty}=x_{0}$

In proof we assume $y_0=x$. By applying the triangle inequality, contraction and the assumption $d(y_{n+1},T(y_n))\leq \epsilon_{n}$ we get, $$d(T^{m+1}(x),y_{m+1})\leq\sum_{i=0}^mk^{m-i}\epsilon_{i}$$ Again by applying triangle inequality, we get, $$d(y_{m+1},x_{0})\leq\sum_{i=0}^mk^{m-i}\epsilon_{i}+d(T^{m+1}(x),x_{0})$$ Since $\{\epsilon_n\}\rightarrow 0$, for given $\epsilon> 0$, there exists $N\in\mathbb{N}$ such that for $m\geq N$, $\epsilon_m\leq\epsilon$ $$\sum_{i=0}^mk^{m-i}\epsilon_{i}\leq k^{m-N}\sum_{i=0}^Nk^{N-i}\epsilon_{i}+\epsilon\sum_{i=N+1}^mk^{m-i}\epsilon_{i}$$

I understand up to this step. From this how do I get, $$\lim_{m\rightarrow\infty}\sum_{i=0}^mk^{m-i}\epsilon_{i}\leq\epsilon\left(\frac{k^N-1}{1-k}\right)$$

I am not sure how to make $m$ larger and obtain the limit of the sum. I have trouble in figuring out $k^{m-N}$ and $k^{m-i}$ as $m\rightarrow\infty$. Kindly help. Sorry, if I have not posed my question properly.

  • $\begingroup$ Got it. Thank you. It is actually k^N not k^(N) - 1. It is a mistake. $\endgroup$ – Vicky Apr 19 '15 at 13:53
  • $\begingroup$ you might wanna edit your post, especially the statement of the theorem. $\endgroup$ – Saaqib Mahmood May 17 '16 at 5:52

The $\epsilon_i$ at the very end of the displayed inequality

$$\sum_{i=0}^mk^{m-i}\epsilon_{i}\leq k^{m-N}\sum_{i=0}^Nk^{N-i}\epsilon_{i}+\epsilon\sum_{i=N+1}^mk^{m-i}\epsilon_{i}$$

should not be there: it’s

$$\sum_{i=0}^mk^{m-i}\epsilon_{i}\leq k^{m-N}\sum_{i=0}^Nk^{N-i}\epsilon_{i}+\epsilon\sum_{i=N+1}^mk^{m-i}\;.$$

Also, $\epsilon\left(\frac{k^N-1}{1-k}\right)$ can’t be right: $k\in(0,1)$, so the fraction is negative. Here’s a correct calculation:

$$\begin{align*} \lim_{m\to\infty}\sum_{i=0}^mk^{m-i}\epsilon_{i}&\le \lim_{m\to\infty}\left(k^{m-N}\sum_{i=0}^Nk^{N-i}\epsilon_{i}+\epsilon\sum_{i=N+1}^mk^{m-i}\right)\\ &=\lim_{m\to\infty}k^{m-N}\sum_{i=0}^Nk^{N-i}\epsilon_{i}+\epsilon\lim_{m\to\infty}\sum_{i=N+1}^mk^{m-i}\\ &\overset{(*)}=\left(k^{-N}\sum_{i=0}^Nk^{N-i}\epsilon_{i}\right)\lim_{m\to\infty}k^m+\epsilon\lim_{m\to\infty}\sum_{\ell=0}^{m-N-1}k^\ell\\ &=0+\epsilon\sum_{\ell=0}^\infty k^\ell\\ &=\epsilon\left(\frac1{1-k}\right)\;. \end{align*}$$

In the step marked $(*)$ I simply let $\ell=m-i$: $N+1\le i\le m$ is then precisely equivalent to $0\le\ell\le m-N-1$. And of course $m-N-1\to\infty$ as $m\to\infty$, so the last limit on that line is just the sum of the infinite geometric series.


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.