I've been working on a project for a wiki that I'm a member of. It is the Sequence of the Day for September 2. You can see my progress at https://oeis.org/wiki/Template:Sequence_of_the_Day_for_September_2 .

In short I ask, when you consider tetrations with rational heights, and compare them to exponentials with rational powers, what is the equivalent form of tetration to the exponential $n^{1/n}$?

I was wondering if anyone here knew the answer and would share it with me. If so, I would like to rephrase my project.

  • $\begingroup$ Powers have the property that $(a^b)^c = a^{bc}$, so a rational power like $1/b$ has to have the effect of an inverse. Unfortunately I don't think tetrations have a similar property, so you might not be able to draw a meaningful parallel. $\endgroup$ – trutheality Jun 25 '11 at 3:16
  • 3
    $\begingroup$ Possible duplicate of How to evaluate fractional tetrations? $\endgroup$ – user334732 May 11 '17 at 12:36

For this answer, I use the tetration notation ${^b} a = a^{a^{...^a}}$ with height $b$.

You could just look at the sequence ${^{1/n}}n$ (assuming you have a definition of tetration for noninteger heights), but I imagine this is not what you are looking for because fractional exponentiation and fractional tetration are not good analogues to eachother. Many other answers have addressed this problem: e.g.: this, this, and this.

Perhaps a more interesting analogue is to look at the fact that $n^{1/n} = \sqrt[n]{n}$, so the sequence $a_n = n^{1/n}$ represents the unique positive number such that $(a_n)^n = n$. One could similarly define a sequence $b_n$ with tetration-inverse, so $b_n$ is the unique positive solution to the equation ${^n}(b_n) = n$. Fortunately for us, this sequence doesn't depend on how to compute tetration for non-integer heights so we don't have to do anything complicated here.

Is this sequence $b_n$ interesting? Probably not very much, but we can show one interesting fact: $$ \lim_{n\rightarrow\infty} b_n = e^{\frac{1}e} $$ Proof: Let $c_1 = \limsup b_n$ and $c_2 = \liminf b_n$. Suppose $c_1 > e^{\frac{1}e}$. Then there exists $\epsilon > 0$ such that for $n$ sufficiently large, $b_n > e^{\frac1e} + \epsilon$. This would imply $$ n = {^n}(b_n) > {^n}(e^{\frac1e} + \epsilon) $$ for all $n$ sufficiently large. This is impossible however, since for $\alpha > e^{\frac1e}$, ${^n} \alpha$ grows faster than $n$. Thus we can conclude $c_1 \le e^{\frac1e}$. Similarly, suppose $c_2 < e^{\frac1e}$. Then we would have for $n$ sufficiently large $b_n < e^{\frac1e}$, which would in turn yield $$ n = {^n} (b_n) < {^n}(e^{\frac1e}) < e $$ for all $n$ sufficiently large. (It is fairly straightforward to show that ${^n}(e^{\frac1e})$ is increasing and converges to $e$). This is a contradiction, and hence $c_2 \ge e^{\frac1e}$. Since $c_2 \le c_1$, we conclude that $c_1 = c_2 = e^{\frac1e}$. Therefore, $\lim b_n$ converges to $e^{\frac1e}$.


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.