Suppose $X = \{\{x_n\}_{n \in \mathbb{N}} : x_n \in \mathbb{R}$ $\forall n\geq 1\}$. Given that $d: X \times X \rightarrow \mathbb{R}$ such that $d(\{x_n\}_{n \in \mathbb{N}}, \{y_n\}_{n \in \mathbb{N}}) = \sum_{n = 1}^{\infty} 2^{-n}\frac{|x_n - y_n|}{1 + |x_n - y_n|}$ is a metric, show that $(X, d)$ is a complete metric space.

My attempt:

Consider $\{x^j\}_{j\geq 1}$ Cauchy in $d$. Fix $n \geq 1$. Then for all $0 < \epsilon < \frac{1}{2}$, there is an $j_{\epsilon} \in \mathbb{N}$ so that $d(x^{(j)}, x^{(k)}) < \frac{\epsilon}{2^n}$ for all $j, k \geq j_{\epsilon}$. So for $j, k \geq j_{\epsilon}$ we have

$\sum_{m = 1}^{\infty} 2^{-m}\frac{|x_m^{(j)} - x_m^{(k)}|}{1 + |x_m^{(j)} - x_m^{(k)}|} < \frac{\epsilon}{2^n}$

$\implies 2^{-m}\frac{|x_m^{(j)} - x_m^{(k)}|}{1 + |x_m^{(j)} - x_m^{(k)}|} < \frac{\epsilon}{2^n}$

$\implies |x_m^{(j)} - x_m^{(k)}| < \frac{\epsilon}{1-\epsilon} \leq 2\epsilon$

So $\{x^{(j)}\}_{j\geq 1}$ is cauchy $\implies x_n^{(j)} \rightarrow x_n$ as $x \rightarrow \infty$.

Is this correct so far? I have come this far but I'm not sure how to complete the proof. Any assistance is much appreciated.

  • $\begingroup$ Why did you choose $j_{\epsilon, n}$ to depend on $n$? $\endgroup$ – P. J. Mar 12 at 7:30
  • $\begingroup$ @P.J. My bad. I've edited it accordingly. $\endgroup$ – SupremePickle Mar 12 at 7:51

Given any $e>0,$ take $m$ large enough that $2^{-m}<e/2.$ You have shown that each $(x^j_n)_j\to x_n$ as $j\to\infty.$ So for $n\le m,$ take $j_n $ such that $(j\ge j_n\implies |x^j_n-x_n|<e/2)$. And let $k=\max_{n\le m}j_n.$ Now we have $$ j\ge k \implies d(x^j,x)=\sum_{n=1}^m 2^{-n}\frac {|x^j_n-x_n|}{1+|x^j_n-x_n|}+\sum_{n=m+1}^{\infty}2^{-n}\frac {|x^j_n-x_n|}{1+|x^j_n-x_n|}\le$$ $$\le\sum_{n=1}^m2^{-n}(e/2)+\sum_{n=m+1}^{\infty}2^{-n}<$$ $$<e/2+2^{-m}<e.$$

  • $\begingroup$ The idea is that if $m$ is large enough then the sum of the terms for $n\ge m+1$ in the series for $d(x^j,x)$ will be arbitrarily small, regardless of the values of $j$ and $x^j$ and $x,$ so we then attend to a finite set of series $\{(x^j_n)_j:n\le m\}.$ $\endgroup$ – DanielWainfleet Mar 12 at 8:39
  • $\begingroup$ The topology on $X=\Bbb R^{\Bbb N}$ generated by $d$ is the (Tychonoff) product topology. We can use $d$ to show $X$ is separable by showing that the set of all $(y_n)_n\in X$ such that (i) every $y_n\in \Bbb Q$ and (ii) $\{n\in\Bbb N : y_n\ne 0\}$ is finite, is a countable dense subset of $X$. $\endgroup$ – DanielWainfleet Mar 12 at 11:24
  • $\begingroup$ Ah this makes sense. Thank you. $\endgroup$ – SupremePickle Mar 12 at 18:17

Your $j_{\epsilon,n}$ should be just $j_{\epsilon}$. Take $m=n$ in your argument. Rest is fine.

You now have $x_n=\lim_{j \to \infty} x_n^{j}$ for each $n$. Now let $x=(x_n)$. Then $d(x,x^{j}) \leq \sum_{m=1}^{N} 2^{-m} \frac {|x_m^{j}-x_m|} {1+|x_m^{j}-x_m|} + \sum_{m=N+1}^{\infty} 2^{-m} \frac {|x_m^{j}-x_m|} {1+|x_m^{j}-x_m|}$. The second term here is at most $\sum_{m=N+1}^{\infty} 2^{-m} <\epsilon$ if $N$ is large enough. The first term is a finite sum and each term tends to $0$. I will let you write out a proof in terms of $\epsilon$ and $n_0$.

  • $\begingroup$ I'm struggling to put this down formally.. Any chance you could elaborate? $\endgroup$ – SupremePickle Mar 12 at 7:51
  • 1
    $\begingroup$ @Kraftsman Choose $N$ such that the second term is less than $\epsilon /2$. Then choose $j_0$ such that $j >j_0$ implies that each of the terms in the first sum is less than $\epsilon /{2N}$. $\endgroup$ – Kavi Rama Murthy Mar 12 at 7:54

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.