Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Background Suppose $(x_i)$ and $(y_j)$ are two sequences in a metric space $(X,d)$ that converge respectively to $x$ and $y$. I am trying to show that the sequence $(d( x_i, y_j))$ converges to $d(x,y)$

Thoughts I think probably the first thing to do is to show that the metric is itself a continuous function. I believe this is so because if $V$ is an open subset of $\mathbb{R}$ in the range of $d$, say $V := (a, b)$ then the preimage of $V$ is the set $$ d^{-1}(V) = \{(x,y) \in X \times X: a < d(x,y) < b \} $$ which is an open ball of radius $|b-a|$. Assuming this part is OK, since $d$ is continuous, it commutes with the limit operation but in this case, there are actually two limits: one as $i \rightarrow \infty$ and one as $j \rightarrow \infty$ I think maybe there is a couple of ways around this; one, by fixing $x := x_0$ and taking the limit as $j \rightarrow \infty$ and then by similarly fixing $y := y_0$ and take the limit as $i \rightarrow \infty$. Perhaps another way is to simply use the same sequence index and consider each point $(x_i, y_i)$ as a term in the sequence.

Question I think I understand intuitively what's going on, but I'm not sure of how to make the argument precise. Is the approach I outlined promising or is there a cleaner approach that I'm overlooking?

share|improve this question
Your inverse image of $V$ is incorrect. $d^{-1}(V)=\{(x,y)\in X\times X\mid a\lt d(x,y)\lt b\}$. –  Arturo Magidin Feb 15 '12 at 19:55
What ordering are you placing on $\mathbb{N}\times\mathbb{N}$ in order to consider $(d(x_i,y_j))$ a sequence? –  Arturo Magidin Feb 15 '12 at 19:57
@Arturo I'll fix V; with regard to ordering, well, perhaps I'm not stating the problem clearly. I'm trying to say that as $x_i$ approaches $x$ and $y_i$ approaches y then the distance between $x_i$ and $y_i$ approaches $d(x,y)$ –  ItsNotObvious Feb 15 '12 at 20:01
in other words: you want to prove that for all $\epsilon\gt 0$ there exist $N,M\gt 0$ such that if $i\geq N$ and $j\geq M$, then $|d(x,y) - d(x_i,y_j)|\lt\epsilon$. –  Arturo Magidin Feb 15 '12 at 20:06
add comment

3 Answers 3

up vote 2 down vote accepted

Assume that $\{x_n:n\in\mathbb{N}\}\subset X$ connverges to $x\in X$ and $\{y_n:n\in\mathbb{N}\}\subset X$ connverges to $y\in X$. Fix $\varepsilon >0$, then there exist $K\in\mathbb{N}$, $L\in\mathbb{N}$ such that for all $k>K$ and $l>L$ we have $$ d(x_k,x_0)<\varepsilon/2\qquad\qquad d(y_l,y_0)<\varepsilon/2 $$ Consider $N=\max(M,L)$, then for all $k>N$ and $l>L$ we have $$ |d(x_k,y_l)-d(x_0,y_0)|\leq d(x_k,x_0)+d(y,y_0)<\varepsilon $$ Since $\varepsilon >0$ is arbitrary we conclude $$ \lim\limits_{k,l\to\infty}d(x_k,y_l)=d(x_0,y_0). $$

P.S. In my proof I've used the following inequality: $$ |d(p,q)-d(r,s)|\leq d(p,r)+d(q,s) $$ which holds for all $p,q,r,s\in X$. It can be proved by the following way $$ d(p,q)\leq d(p,r)+d(r,q)\leq d(p,r)+d(r,s)+d(s,q)\Longrightarrow $$ $$ d(p,q)-d(r,s)\leq d(p,r)+d(q,s) $$ Similarly, $$ d(r,s)\leq d(r,p)+d(p,q)\leq d(r,p)+d(p,q)+d(q,s)\Longrightarrow $$ $$ d(r,s)-d(p,q)\leq d(r,p)+d(q,s) $$ Then $$ |d(p,q)-d(r,s)|\leq d(p,r)+d(q,s) $$

share|improve this answer
I was not familiar with the $|d(p,q)−d(r,s)| \leq d(p,r)+d(q,s)$ inequality that you used; this makes everything very clear. Thanks. –  ItsNotObvious Feb 15 '12 at 20:22
So, using this result, I believe we can say the following: $\lim_{i \to \infty} \lim_{j \to \infty} d(x_i, y_j) = \lim_{k \to \infty}d(x_k, y_k)$, correct? –  ItsNotObvious Feb 16 '12 at 16:30
Yes, if limit of the type $\lim\limits_{k,l\to\infty}$ exists, then exists repeared limits like $\lim\limits_{i\to\infty}\lim\limits_{j\to\infty}$, and limit over subsequences like $\lim\limits_{l=k,k\to\infty}$ –  Norbert Feb 16 '12 at 16:39
add comment

It is true that $d$ is a continuous function on $X\times X$, so if you show $(x_i,y_i)\to (x,y)$ in $X\times X$ the result follows. However, I would use a different approach. First note that $d(x_i,y_i)\leq d(x_i,x)+d(x,y)+d(y,y_i)$ by the triangle inequality. Furthermore, since $d(x_i,x)+d(x_i,y_i)\geq d(x,y_i)$ and $d(x,y_i)+d(y,y_i)\geq d(x,y)$ by the triangle inequality, we get $$d(x_i,y_i)\geq d(x,y_i)-d(x_i,x) \geq d(x,y)-d(x_i,x)-d(y,y_i)$$ so putting our two inequalities together $$d(x,y)-d(x_i,x)-d(y,y_i)\leq d(x_i,y_i)\leq d(x_i,x)+d(x,y)+d(y,y_i)$$ and the left and right expressions go to $d(x,y)$ because $x_i\to x$ and $y_i\to y$ (and by the definition of convergence this means $d(x_i,x),d(y,y_i)\to 0$). Thus $d(x_i,y_i)\to 0$.

share|improve this answer
add comment

$$ \color{red}{0\leqslant|d(x_i,y_j)-d(x,y)|\leqslant d(x_i,x)+d(y_j,y)\longrightarrow0} $$

share|improve this answer
add comment

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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