I know it's very petty question in real-analysis but I want to make the proof finely logical.

There is 3 steps of the proof, we all know well, that cauchy sequence is convergent sequence. The first is proving cauchy sequence is bounded. The second is that by Bolzano-Weierstrass theorem, there exists a cluster point. The third is showing the original cauchy sequence is convergent to the cluster point.

I think there is a logical gap in my text book, but i cannot make it fine.

Let $L$ be the cluster point of $\{a_n\}$ which is cauchy sequence. Then there exists a subsequence $\{a_{n_k}\}$ convergent to $L$.

Given $\epsilon>0$, $\exists N_1 \in \mathbb{N}, \forall k : [k>N_1 \rightarrow |a_{n_k} - L| < \epsilon/2$.

By cauchy condition, $\exists N_2 \in \mathbb{N}, \forall n,m : [n,m > N_2 \rightarrow |a_m - a_n| < \epsilon/2$.

$\textbf{N := max\{N_1, N_2\}}$, then for all $n>N$ $$|a_n-L| = |a_n - a_{n_{N+1}} + a_{n_{N+1}} -L| \leq |a_n - a_{n_{N+1}}| + |a_{n_{N+1}} -L| < \epsilon.$$

I think the bold text is the logical gap in the proof. If we set $N$ like the above proof, there is no verification that $n_{N+1} > N_2$. So, if we make this proof well, additory condition is required such as

$\textbf{N :=max\{N_1, N_2\}}$, then for all $n>N$ $\textbf{and $n_k>N$} $.

However the definition of convergence has only $n>N$ assumption to make it sense the inequality that $|a_n -L|<\epsilon$.

How do we set the index numbering to make the proof have no logical gap?

  • 1
    $\begingroup$ $n_j\ge j$ is part of the definition of "subsequence". $\endgroup$ Aug 9, 2015 at 23:30

1 Answer 1



$$n_k \ge k, \forall k \in \mathbb N$$

Where $(n_k)$ is a strictly increasing sequence of naturals.

(inductive step: $n_{k+1} > n_k \implies n_{k+1} \ge n_k + 1 \ge k + 1$)


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.