3

I was trying to solve some problem from a question here on MSE by first trying to find something about simplified version, and, if I calculated correctly I obtained:

$$\lim_{n \to + \infty} (\sqrt{a})^{3^n} \cdot \prod_{k=2}^{n} (\dfrac {a}{k})^{3^{n-k+1}}=1$$

Even if my calculations are not right this is interesting in itself and what I would like to know is does such an $a$ exists? That is, even though I obtained this result and am trying to find closed form for $a$ (or an approximation) I really do not know even if there exists an $a$ that would satisfy this limit problem.

Does it exists? How to calculate it if it exists?

Shalom
  • 949
  • Could you please link the question from "here on MSE"? How can we tell if you calculated correctly to obtain the formula you ask about, without knowing the original question? – amWhy Mar 28 '18 at 00:09
  • Just looking I would expect the limit to be $0$. Once $n$ gets huge compared to $a$ most of those $\frac ak$ terms will be tiny and overwhelm the $a^{3^n/2}$ – Ross Millikan Mar 28 '18 at 00:17
  • @RossMillikan This is $+ \infty \cdot 0$ limit, It could be that such an $a$ exists. – Shalom Mar 28 '18 at 00:19
  • @Shalom: I am suggesting that the $0$ is much stronger than the $+\infty$. I am not certain. – Ross Millikan Mar 28 '18 at 00:21
  • @RossMillikan I understand what you´re saying, but I believe that there is such an $a$. – Shalom Mar 28 '18 at 00:22

1 Answers1

1

For any $a \gt 1$ the limit is $0$, not $1$.

Lets start with the numerator of the product. $$\prod_{k=2}^na^{3^{n-k+1}}=a^{\sum_{k=2}^n3^{n-k+1}}\\ \sum_{k=2}^n3^{n-k+1}=\frac{3^n-3^1}{3-1}\approx \frac {3^n}2 \\ \text {and all the } a\text{s including the leading factor give us }a^{3^n} $$ We will loosely bound the denominator from below, remembering that $n \gg a$ $$\prod_{k=2}^nk^{3^{n-k+1}}\gt \prod_{k=a^3}^nk^{3^{n-k+1}}\gt \prod_{k=a^3}^n(a^3)^{3^{n-k+1}}=(a^3)^{\frac{3^n-3^{a^3}}{3-1}}=a^{\frac 323^n}\cdot a^{-3^{a^3}}$$ The trailing $a^{-3^{a^3}}$ is just a constant. For any $a \gt 1$ the terms depending on $n$ will be less than $a^{\frac 123^n}$, which can be made as small as we want.

Ross Millikan
  • 374,822