# Transfinite Knuth-arrow hierarchy vs. fast-growing hierarchy

Suppose Knuth arrow notation (and hence the hyperoperation sequence) is extended to transfinite ordinal indices as follows:

Let μ be a large countable ordinal such that a fundamental sequence is assigned to every limit ordinal less than μ, and take these to be the same as in the definition of the fast-growing hierarchy of functions ($f_\alpha$). Now define
$$m\uparrow^\alpha n \ \ = \ \ \begin{cases} m n & \text{if }\alpha=0, \\ (m\uparrow^{\alpha-1})^n 1 & \text{if }\alpha \text{ is a successor ordinal}, \\ m\uparrow^{\alpha[n]}n & \text{if }\alpha \text{ is a limit ordinal}, \end{cases}$$ for all integers $m, n \ge 0$ and ordinal $\alpha < \mu$, where $(\alpha[n])_{n=0,1,2,\dots}$ is the fundamental sequence for $\alpha$.

It can be shown that for all $\alpha < \omega$, there is an $n_0$ (depending on $\alpha$) such that for all $n > n_0$, $$f_\alpha(n) \ \ \lt \ \ 2\uparrow^\alpha n \ \ \le \ \ f_{\alpha+1}(n).\tag{*}$$

Question: Does (*) hold for all $\alpha < \mu$? Or, is there a "cross-over point", i.e., a least ordinal $\beta (< \mu)$ such that for all $\alpha \ge \beta, \ \ \ f_\alpha(n) > 2\uparrow^\alpha n \$ for all sufficiently large $n$?

EDIT: Small cases $n = 0, 1, 2, 3, ...$ suggest that $\omega$ is actually the cross-over point. Proof?

Motivation: If (*) holds for all $\alpha < \mu$, it would seem that with minor adjustments virtually all of the points of interest concerning the fast-growing hierarchy ($f_\alpha)$ apply as well to the hierarchy of functions $(2\uparrow^\alpha)$.

NB: Arrow notation has been used for convenience. The discussion could be rephrased in terms of extending the hyperoperation sequence $H_\alpha(m,n)$, with a shift in the indexing to accomodate successor and addition.

Certainly this eventual inequality holds for all $\alpha<\omega$. Then note that

$$f_\omega(n)=f_{\omega[n]}(n)\ll2\uparrow^{\omega[n]}n=2\uparrow^\omega n$$

So the lower bound holds for $\omega$. Likewise does the upper bound somewhat trivially. However, it fails for $\omega+1$. Let $\omega[n]=n$. Note that

$$f_\omega(n+1)=f_{n+1}(n+1)\gg2\uparrow^n(n+1)\ge(2\uparrow^nn)+1=(2\uparrow^\omega n)+1$$

Hence by induction,

$$f_\omega^k(n+1)\gg((2\uparrow^\omega)^kn)+1>(2\uparrow^\omega)^k1$$

$$f_{\omega+1}(n)=f_\omega^n(n)\gg(2\uparrow^\omega)^n1=2\uparrow^{\omega+1}n$$