I had some issues proving this, so I relaxed the conditions to prove the statement in $\mathbb{R}$ first. Here is my attempt:

Note: I used the following theorem:

Theorem. Every Cauchy sequence of real numbers converges.

Proof. By the theorem, we know $\{p_n\}$ and $\{q_n\}$ converges since they are Cauchy. So by definition,

$(1)$ $p_n \rightarrow p$ for given $\epsilon > 0,\ \exists N_1>0$ such that $d(p_n,p)< \epsilon /2$.

$(2)$ $q_n \rightarrow q$ for given $\epsilon > 0,\ \exists N_2>0$ such that $d(q_n,q)< \epsilon /2$.

Let $N> max(N_1,N_2)$. So $\forall n \geq N$, $$d(p_n,q_n) \leq d(p_n,p) + d(p,q) + d(q_n,q)\ (*)$$ $$< \epsilon /2 + d(p,q) + \epsilon /2$$ $$\Rightarrow d(p_n,q_n) - d(p,q) < \epsilon\ (1)$$ Similarly, mult. $(*)$ by $-1$ and rearrange to get $$d(p,q) - d(p_n,q_n) > \epsilon\ (2)$$ Thus, by $(1)$ and $(2)$ we get $$|d(p_n,q_n) - d(p,q)| < \epsilon$$ Hence, $d(p_n,q_n) \rightarrow d(p,q)$. $□$

Now, how can I generalize it to any metric space $X$? For some reason, I think it cannot be done. Thank you!

  • 1
    $\begingroup$ Show that $d(p_n,q_n)$ is a Cauchy sequence of real numbers. $\endgroup$ – uniquesolution Mar 15 '17 at 16:10

You don't want to use the fact that $p_n,q_n$ converges in your proof because it doesn't hold unless $X$ is a complete metric space. So instead, try to show that if $p_n,q_n$ are Cauchy in $X$ then $d(p_n,q_n)$ is Cauchy in $\mathbb{R}$ and then use the fact that $\mathbb{R}$ is complete so every Cauchy sequence in $\mathbb{R}$ converges. To do that, try to estimate

$$ \left| d(p_n,q_n) - d(p_m,q_m) \right| $$

in terms of $d(p_n,p_m), d(q_n,q_m)$ using the triangle inequality and then show that this is Cauchy because $d(p_n,p_m), d(q_n,q_m)$ are Cauchy.

  • $\begingroup$ Thank you for the tip! Here is my attempt (again!): Proof. Let $\{p_n\}$ and $\{q_n\}$ be Cauchy in $X$. Then, $(1)$ $p_n \rightarrow p$ for given $\epsilon > 0,\ \exists N_1>0$ such that $d(p_n,p_m)< \epsilon /2,\ \forall n, m> N_1$. $(2)$ $q_n \rightarrow q$ for given $\epsilon > 0,\ \exists N_2>0$ such that $d(q_n,q_m)< \epsilon /2,\ \forall n, m>N_2$. Let $N> max(N_1,N_2)$. So $\forall n \geq N$, $$d(p_n,q_n) \leq d(p_n,p_m) + d(p_m,q_m) + d(q_m,q_n)\ (*)$$ $$< \epsilon /2 + d(p_m,q_m) + \epsilon /2$$ $$\Rightarrow \boxed{d(p_n,q_n) - d(p_m,q_m) < \epsilon}\ (*)_1$$ (cont.) $\endgroup$ – Curious Math Student Mar 15 '17 at 17:16
  • $\begingroup$ (cont.) Similarly, multiply $(*)$ by $-1$ and get $$-d(p_n,q_n) \geq -d(p_n,p_m) - d(p_m,q_m) - d(q_m,q_n)$$ $$ d(p_m,q_m) - d(p_n,q_n) > -\epsilon /2 - \epsilon /2$$ $$\Rightarrow \boxed{d(p_m,q_m) - d(p_n,q_n) > \epsilon}\ (*)_2$$ By $(*)_1$ and $(*)_2$ we get $$|d(p_n,q_n) - d(p_m,q_m)| < \epsilon$$ Thus, we have shown that $d(p_n,q_n)$ is Cauchy... in $\mathbb{R}$. Hence, by the completion of $\mathbb{R}$, $d(p_n,q_n)$ converges. QED. Is my attempt okay? $\endgroup$ – Curious Math Student Mar 15 '17 at 17:18
  • 1
    $\begingroup$ @CuriousMathStudent: Yes, but you shouldn't write $p_n \to p$ because it is not true (and you don't use it) and similarly for $q_n \to q$. Also, it should be $-\varepsilon$ in the second starred equation. $\endgroup$ – levap Mar 15 '17 at 19:01
  • 1
    $\begingroup$ Oh right! I forgot being Cauchy doesn't imply convergence unless we are in a complete metric space. Thank you for very much for your help! $\endgroup$ – Curious Math Student Mar 16 '17 at 2:08

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.