Proof of Cauchy sequence.

142 Views Asked by At

I'm facing this problem,

Let $\{ x_n \}_{ n\in \mathbb{N} }$ be a sequence of real numbers and $0 < r< 1$ such that $|x_{n+1} - x_n | < r^n$ for all $n \in \mathbb{N}$. Prove that $\{ x_n \}_{ n\in \mathbb{N}}$ is a Cauchy sequence.

My work based on other similar problems is:

Let's consider $|x_m-x_n|$ where $m$ y $n$ do not need to be consecutive numbers, assume $m\leq n$. Then $k=n-m\geq 0$,

$\\ \begin{aligned} |x_m-x_n| &=|x_m-x_{m+k}| \\ & =|x_m-x_{m+1}+x_{m+1}-x_{m+2}+...+x_{m+k-1}-x_{m+k}| \\ & \leq|x_{m}-x_{m+1}|+|x_{m+1}-x_{m+2}|+...+|x_{m+k-1}-x_{m+k}| \\ & < r^{m}+r^{m+1}+...+r^{m+k-1} \\ & = r^{m}(1+r^1+...+r^{k-1}) \\ & = (r^{m})\sum_{i=0}^{k-1}r^{i} \\ & \leq r^m(1/(1-r)). \\ \end{aligned}$

$\\ $ If $m>N_{\epsilon}$, then we have $r^{m}<r^{N_{\epsilon}}$ because $0<r<1$ (I think I need a strong argument to guarantee this). Then $r^m(1/(1-r))<r^{N_{\epsilon}}(1/(1-r))$

$\\ $ Can I guarantee that always exists $N_{\epsilon}\in \mathbb{N}$, such that $r^{N_{\epsilon}}(1/1-r)<\epsilon$ ? If so, then I get what I want

$$\\ |x_m-x_n|<r^m(1/(1-r))<r^{N_{\epsilon}}(1/(1-r))<\epsilon$$ $\ $ Whenever $m,n>N_{\epsilon}$

My greatest doubt. If $m>N_{\epsilon}$, then we have $r^{m}<r^{N_{\epsilon}}$ (Whenever $0<r<1$). Will this always be true? I'm thinking of mathematical induction to prove this.

1

There are 1 best solutions below

0
On

Generally if you didn't express doubts about some steps this proof would have been quite widely accepted.

Your first concern is if you can assume that if $m>N_\epsilon$ then $r^m<r^{N_\epsilon}$ or a proof of that. Now we have that $N_\epsilon = m + d$ for some $d>0$ so $r^m-r^{N_epsilon} = r^m - r^{m+d} = r^m(1-r^d)$. So what you need to know is that $0<r^d<1$ which is quite straight forward to prove with induction.

Your second concern is if you can guarantee that $r^{N_\epsilon}(1/(1-r))$ can be made arbitrarily small by selecting $N_\epsilon$. First of all $r$ is just a constant so we should focus on $r^{N_\epsilon}$. It's a well known fact and you could use the estimate that if $h>0$ then $(1+h)^n \ge 1+nh$ (which can be shown unsing induction) to see that

$$r^{N_\epsilon} \le {1\over (1+ (1/r-1))^{N_\epsilon}} \le {1\over1 + N_\epsilon(1/r-1)} \le {1\over N_\epsilon}{1\over1/r-1}$$

Which can be made arbitrarily small (while we know that it's also positive).