2

The stolz caesaro theorem seems to be a discrete analogue of L'hopitals rule. I can understand l'hopitals rule via the taylor series, that is:

$$ \lim_{x \to a} \frac{P(x)}{Q(x)} = \frac{ P(a) + \frac{dP}{dx}|_a (x-a) + O ( (x-a)^2) }{ Q(a) + \frac{dQ}{dx}|_a (x-a) + O ( (x-a)^2)}$$

Now if $P(a) = Q(a) = 0$, the limit becomes ratio of first order term/ the ratio of values nearby to the point of interest. So, Is there a similar 'series' intuition for stolz caesaro theorem?

1 Answers1

1

It seems trivial compared to Taylor series or L'Hopital, because it follows directly from a much more general theorem: $ \def\lfrac#1#2{{\large\frac{#1}{#2}}} $

Theorem. Take any $a,b,c∈ℂ$ and sequences $f,g$ from $ℂ$ such that both $\sum_{k=0}^n g(k) → ∞$ and $\lfrac{f(n)}{g(n)} → c$ as $n→∞$. Then $\lfrac{ a + \sum_{k=0}^n f(k) }{ b + \sum_{k=0}^n g(k) } → c$ as $n→∞$.

Proof. For convenience let $[ε]$ denote $\{ z : z∈ℂ ∧ |z|<ε \}$. Take any $ε>0$. Let $m∈ℕ$ such that $\lfrac{f(n)}{g(n)} ∈ c+[ε]$ for every $n∈ℕ_{≥m}$. As $n→∞$, $\lfrac{ a + \sum_{k=0}^n f(k) }{ b + \sum_{k=0}^n g(k) }$ $= \lfrac{ a + \sum_{k=0}^{m-1} f(k) + \sum_{k=m}^n f(k) }{ b + \sum_{k=0}^{m-1} g(k) + \sum_{k=m}^n g(k) }$ $\overset{*}∈ \lfrac{ (1+o(1)) · \sum_{k=m}^n f(k) }{ (1+o(1)) · \sum_{k=m}^n g(k) }$ $⊆ (1+o(1)) · \lfrac{ \sum_{k=m}^n \big( (c+[ε])·g(k) \big) }{ \sum_{k=m}^n g(k) }$ $⊆ (1+o(1)) · (c+[ε])$, where the $*$-marked relation holds because $\sum_{k=m}^n f(k) → ∞$ and $\sum_{k=m}^n g(k) → ∞$. Since this holds for every $ε>0$, therefore $\lfrac{ a + \sum_{k=0}^n f(k) }{ b + \sum_{k=0}^n g(k) } → c$ as $n→∞$.
−−−
Note that in the above theorem statement we allow division by zero but the assertion of a limit means that eventually the value is defined (no division by zero). If you wish to avoid that issue, simply impose that $g(n)$ and $b + \sum_{k=0}^n g(k)$ are never zero for every $n∈ℕ$.

user21820
  • 57,693
  • 9
  • 98
  • 256
  • I'm having a hard time following all the symbolic mathematics, could you add in a preface for what you're exactly doing? – tryst with freedom Jan 24 '21 at 12:58
  • @Buraian: This is quite basic stuff. Point specifically to the first statement that you cannot understand, and we will go from there. (I edited to fix a minor error.) – user21820 Jan 24 '21 at 13:34
  • Ok, I learned some more about logical symbols and now I could read your answer. However I am still being tripped over the little 'o' notation that you've used, also is there a name for this theorem? @user21820 – tryst with freedom Feb 21 '21 at 15:11
  • @Buraian: I just made the theorem up on the spot after reading your question. Are you asking about the "$o(1)$"? The whole chain of comparisons is under the asymptotic condition "$n→∞$", and so each of those "$o(1)$"s means that for any $r∈ℝ^+$ we have that it is eventually bounded (as $n→∞$) by $r$ times $1$. – user21820 Feb 21 '21 at 15:21
  • What does r represent, I don't see it anywhere in the text @user21820 – tryst with freedom Feb 21 '21 at 15:22
  • It's impossible to fully understand what the little-o notation means without working through the formal definitions (say from wikipedia). However, you can get some intuition by looking at how simple limits are expressed in that notation. For example, $\lim_{x→0} f(x) = t$ iff $f(x) = t+o(1)$ as $x→0$. In the usage in my post, we have something of the form $U+V$ where $U$ is constant but $V→∞$, so clearly $U+V∈V·(1+o(1))$. – user21820 Feb 21 '21 at 15:23
  • $r$ is not supposed to be in the post; expand the wikipedia definitions for $o(1)$ and you will see that $r$ in my comment corresponds to the $ε$ in the wikipedia article. – user21820 Feb 21 '21 at 15:24
  • Hmm thank you, I'll return again when I learn more to work through the symbolic logic sentences – tryst with freedom Feb 21 '21 at 15:27
  • 1
    @Buraian: If you have difficulty with logic, feel free to ask about specific notation or concepts in this chat-room. By the way, although you closing your question as a duplicate is technically correct (since you found an earlier thread asking the same thing), do note that the top-voted answer at the duplicate thread has no semblance of a proof, and the next highest-voted answer fails to extend to $ℂ$. While it may feel easier to look for 'intuitive explanations', I strongly suggest you learn to gradually wean yourself off that. – user21820 Feb 21 '21 at 15:31
  • 1
    Thank you for the invite @user21820 , I'll hop in some time – tryst with freedom Feb 21 '21 at 15:34