0

Let $b>1$.

Define $log_b^*(n)$ as taking the logaritm base $b$ as many times a possible such that $log_b^*(n) > 1$.

Now define $g(n) = ln_b(n) \ln_b^{[2]}(n) \ln_b^{[3]}(n) ... log_b^*(n)$

where $ln_b^{[k]}$ means taking log base $b$ a number of $k$ times.

Now let $z$ be a complex number and

$$f(z,b) = \sum_{n>b}^{\infty} g(n) z^n$$

$$h(z,b) = \sum_{n>b}^{\infty} \frac{ z^n}{g(n)}$$

What are the analytic continuations of $f(z,b)$ and $h(z,b)$ with respect $z$ ? Or other continuations if there is a natural boundary ?

So basically : How to extend $f(z,b)$ and $h(z,b)$ to $|z|>1$ ?

related, especially for understanding different $b$ :

When does this iterated logarithm series converge / diverge?

https://math.stackexchange.com/a/5072686/39261

And maybe this is handy too :

https://math.stackexchange.com/a/5072816/39261

I was wondering if the Borel transform, Borel transform and summation etc is all we need ?

mick
  • 17,886
  • The convergence radius is probably $1$ in each case, so you can't continue in a neighbourhood of $|z| = 1$. Perhaps, you can continue them in some part of the plane but not all of it. – Cactus Jun 12 '25 at 13:08
  • @Cactus I know the taylor radius is 1. A perhaps statement without motivation is not much tbh. Why do you believe that ? Are you familiar with analytic continuation or other continuations ? – mick Jun 14 '25 at 11:01
  • turning $g(n)$ into a nice interpolation taylor series and then using polylog to sum it will probably work. – mick Jun 16 '25 at 07:16

0 Answers0