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.

I am a little confused about the following concept from Real Analysis:

suppose $\{a_n\}$ is a sequence in $\mathbb{R}$ which is Cauchy. then it is convergent, and it is shown as follows:

first we show it is bounded (this I understand how to do). then, from the Bolzano-Weierstrass theorem(every bounded sequence of real numbers contains a convergent subsequence) we deduce that there exists a convergent subsequence $\{a_{n_k}\}$ with limit $a$ (say).

we now claim that $\lim_{n \to \infty} a_n = a$.

what we need to show then is that for any given $\varepsilon > 0$ we can find $N > 0$ such that $$\tag{1} n \geq N \Rightarrow |a_n - a| < \varepsilon\,. $$ by the triangle inequality we know that the following inequality is true, for all $n, n_k \in \mathbb{N}$: $$\tag{2} |a_n - a| \leq |a_n - a_{n_k}| + |a_{n_k} - a| \,. $$ i know that the usual argument is now to say we control the first term using the Cauchy property of $\{a_n\}$ and the second term using the convergence of $\{a_{n_k}\}$ to $a$.

and here is where I am struggeling: why are we allowed to constrain the $a_{n_k}$? this is not clear to me, given the statement ${(1)}$. it looks like we have to make a $second$ choice, namely for the right hand side of ${(2)}$ to be $< \varepsilon$, the $n_k$ must be large also.

in particular, this choice cannot (at least as far as i understand) be resolved in a straightforward way (such as simply taking the larger of two constants) because the subsequence is a sequecen indexed by the subscript $k$, not the subscript $n$ any more - so choosing $K$ so that $|a_{n_k} - a| < \varepsilon/2$ whenever $k \geq K$ cannot be combined with choosing $N$ so that $|a_n - a_{n_k}| < \varepsilon/2$ whenever $n \geq N$.

I added this last paragraph because I sense this is where my misunderstanding is coming from - what am I missing? Thanks a lot for your help!!

share|improve this question
add comment

3 Answers 3

up vote 0 down vote accepted

The basic idea is that to pick an $N$ for which $|a_m - a| \leq \epsilon$ for all $m \geq N$, you

  1. Use the convergence of a subsequnce to find a $k$ such that $|a_{n_m}-a| \leq \epsilon$ for all $m \geq k$

  2. Use the cauchy sequence property to find an $l$ such that $|a_p - a_q| \leq \epsilon$ for all $p,q \geq l$

  3. Pick $K$ such that $K \geq k$ and $n_K \geq l$, and let $N = n_K$.

You then have for an arbitrary $m \geq N$ $$ |a - a_m| = |a + (a_N - a_m) - a_N| \leq |a - a_N| + |a_N - a_m| \leq \underbrace{|a - a_{n_K}|}_{\leq \epsilon\text{ by (1)}} + \underbrace{|a_N - a_m|}_{\leq \epsilon\text{ by (2)}} \leq 2\epsilon $$

Now, for cleanliness, you might want to go back and replace $\epsilon$ with $\frac{\epsilon}{2}$ everywhere ;-)

share|improve this answer
add comment

You can make the two choices consistently.

First choose $N$ large enough so that $|a_n - a_m|< \epsilon/2$ for all $n,m \ge N$. Then choose $k$ such that $n_k \ge N$ and $|a_{n_k} - a| < \epsilon/2$. This can be done because any sufficiently large $k$ has both of these properties. Then letting $m = n_k$, the right hand side of your inequality (2) is $< \epsilon$.

share|improve this answer
add comment

The idea, if you want $|a_n - a| < \epsilon$, is to choose an $N_1$ such that for any $n_k > N_1$, $|a_{n_k} - a| < \frac{\epsilon}{2}$. This can be done by convergence.

Separately, you want to find an $N_2$ such that for any two $n, m > N_2$, $|a_n - a_m| < \frac{\epsilon}{2}$. This can be done by Cauchy. If you now choose $n_k, n > \max(N_1, N_2)$, then by the triangle inequality (letting $a_{n_k}$ take the place of $a_m$), $$ |a_n - a| \leq |a_n - a_{n_k}| + |a_{n_k} - a| < |a_n - a_{n_k}| + \frac{\epsilon}{2} < 2\frac{\epsilon}{2} = \epsilon $$ where the first strict inequality is valid since $n_k > N_1$, and the second one is valid since $n, n_k > N_2$.

share|improve this answer
add comment

Your Answer

 
discard

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.