Let $(y_n)$ be an unbounded sequence of natural numbers with $y_{n+1}>y_n\forall n\in\mathbb{N}$. Let $(x_n)$ be another sequence, and suppose that $$\lim\frac{x_{n+1}-x_n}{y_{n+1}-y_n}$$ exists. Prove that $$\lim\frac{x_n}{y_n}=\lim\frac{x_{n+1}-x_n}{y_{n+1}-y_n}$$

Theres also a hint that we could use from a previously used question in class which is here: Proving Two Limits Are Equal to Each Other

My issue is that I don't see a correlation between the two. Could some explain please?

  • 2
    $\begingroup$ See Stolz–Cesàro theorem. $\endgroup$ – Git Gud Oct 13 '14 at 17:51
  • $\begingroup$ That looks exactly like what I'm dealing with here but I still don't see the connection between the two problems $\endgroup$ – Jeff Oct 13 '14 at 17:55
  • 1
    $\begingroup$ It's the same thing, I spot no difference. $\endgroup$ – Git Gud Oct 13 '14 at 17:59
  • $\begingroup$ Actually I looked at a page different from the Wikipedia page and I see the connection, thanks I appreciate it $\endgroup$ – Jeff Oct 13 '14 at 17:59

So it seems that you want to use the following result to prove Stolz-Cesaro theorem:

Let $P_{ij}$, where $P_{ij}$ is a real number and $i, j$ natural number, allow the three statements to hold:

  • $P_{ij}\ge 0$ for all $i, j$

  • $\lim_{i\to\infty}P_{ij}=0$ for all $j$

  • $\sum_{j=1}^i P_{ij} =1$ for all $i$

Let $(x_i)$ be a convergent sequence and let a sequence $(y_i)$ be defined by

$y_i=\sum_{j=1}^i P_{ij}x_j$

Then $(y_i)$ is a convergent sequence and $\lim y_i=\lim x_i$

I will add that this is similar to Silverman–Toeplitz theorem. (Well, the assumption $\sup\limits_i \sum\limits_{j=0}^n P_{ij}<+\infty$ is missing, but it seems that you implicitly assume that the sum $y_i$ exists for each convergent sequence $(x_j)$. Silverman-Toeplitz theorem is stronger in the sense that it allows also negative values - which is why absolute value is added in some place - and it gives sufficient and necessary condition. But still it is close to the result you've included here, so it's probably worth mentioning.)

What about proving Stolz-Cesaro theorem in this form:

Theorem. Let us assume that $\lim\limits_{n\to\infty} \frac{a_n}{b_n}=l$, where $b_n\ge0$ and $\sum\limits_{n=1}^\infty b_n=+\infty$. Then $$\lim\limits_{n\to\infty}\frac{\sum_{k=1}^n a_k}{\sum_{k=1}^n b_k}=l.$$

As mentioned here, this is simply a different form of Stolz-Cesaro theorem.

Proof. Let us use the previous result with $x_j=\frac{a_j}{b_j}$ and $$P_{ij}= \begin{cases} \frac{b_j}{b_1+\dots+b_i} & i\ge j, \\ 0 & i<j. \end{cases} $$

The we have $$y_i = \frac{a_1}{b_1} \cdot \frac{b_1}{b_1+\dots+b_i} + \frac{a_2}{b_2} \cdot \frac{b_2}{b_1+\dots+b_i} + \dots + \frac{a_i}{b_i} \cdot \frac{b_i}{b_1+\dots+b_i} = \frac{a_1+\dots+a_i}{b_1+\dots+b_i}.$$

So if we show that $P_i$ fulfills the above condition we get that $$\lim\limits_{i\to\infty} \frac{a_1+\dots+a_i}{b_1+\dots+b_i} = \lim\limits_{i\to\infty} y_i = \lim\limits_{j\to\infty} x_j = \lim\limits_{j\to\infty} \frac{a_j}{b_j}.$$

This proof might seem somewhat unnatural on the first sight. Maybe it helps thinking about the case $b_n=1$ (i.e., $y_n=n$) first.


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.