## FANDOM

10,302 Pages

A fast-growing hierarchy (FGH) is a certain hierarchy mapping ordinals $$\alpha < \mu$$ to functions $$f_\alpha: \mathbb{N} \rightarrow \mathbb{N}$$. For large ordinals $$\alpha$$, $$f_\alpha$$ grows very rapidly. Due to its simple and clear definition, as well as its origins in professional mathematics, FGH is a popular benchmark for large number functions.

If you are unfamiliar with using ordinals in functions, you may want to read the introduction to the fast-growing hierarchy article.

## Definition

A fast-growing hierarchy consists of an ordinal $$\mu$$ and a fundamental sequence system $$S: \mu \cap \text{Lim} \rightarrow (\mathbb{Z}_0 \rightarrow \mu)$$, where $$S(\alpha)(n)$$ is denoted $$\alpha[n]$$. The semantics are as follows:

• $$f_0(n) = n + 1$$
• $$f_{\alpha+1}(n) = f^n_\alpha(n)$$, where $$f^n$$ denotes function iteration
• $$f_\alpha(n) = f_{\alpha[n]}(n)$$ if and only if $$\alpha$$ is a limit ordinal

The general case where $$f_0$$ is any increasing function forms a fast iteration hierarchy.

For a fast-growing hierarchy to be useful to googologists, it is also expected to satisfy the property that for all $$\alpha < \beta < \mu$$, $$f_\alpha$$ is eventually dominated by $$f_\beta$$.

## Systems of fundamental sequences

### Wainer hierarchy

Definitions of the fast-growing hierarchy and choices of fundamental sequence systems vary between authors, so it is generally problematic to speak of "the" fast-growing hierarchy. The most well-known FGH, however, is the Wainer hierarchy, which has $$\mu = \varepsilon_0 + 1$$ and an FS system defined as follows:

• $$\omega[n] = n$$
• $$\omega^{\alpha + 1}[n] = \omega^\alpha n$$
• $$\omega^{\alpha}[n] = \omega^{\alpha[n]}$$ if and only if $$\alpha$$ is a limit ordinal
• $$(\omega^{\alpha_1} + \omega^{\alpha_2} + \cdots + \omega^{\alpha_{k - 1}} + \omega^{\alpha_k})[n] = \omega^{\alpha_1} + \omega^{\alpha_2} + \cdots + \omega^{\alpha_{k - 1}} + \omega^{\alpha_k}[n]$$ where $$\alpha_1 \geq \alpha_2 \geq \cdots \geq \alpha_{k - 1} \geq \alpha_k$$
• $$\varepsilon_0[0] = 0$$ (alternatively $$1$$)
• $$\varepsilon_0[n + 1] = \omega^{\varepsilon_0[n]}$$ = $$\omega\uparrow\uparrow (n-1)$$ (alternatively $$\omega\uparrow\uparrow n$$)

For example, the fundamental sequence for $$\omega^\omega$$ is $$1, \omega, \omega^2, \omega^3, \ldots$$. When authors refer to "the fast-growing hierarchy" without clarification, the Wainer hierarchy is usually meant.

### Veblen's hierarchy

Every non-zero ordinal $$\alpha$$ can be uniquely written in Veblen's variation of Cantor's normal form:

$$\alpha=\varphi_{\beta_1}(\gamma_1) + \varphi_{\beta_2}(\gamma_2) + \cdots + \varphi_{\beta_k}(\gamma_k)$$, where $$\varphi_{\beta}(\gamma)$$ is a function of Veblen's hierarchy, $$\varphi_{\beta_1}(\gamma_1) \ge \varphi_{\beta_2}(\gamma_2) \ge \cdots \ge \varphi_{\beta_k}(\gamma_k)$$ and each $$\gamma_m < \varphi_{\beta_m}(\gamma_m)$$,

For limit ordinals $$\alpha<\Gamma_0$$, written in Veblen's variation of Cantor's normal form, the fundamental sequences for the Veblen's hierarchy are defined as follows:

• $$(\varphi_{\beta_1}(\gamma_1) + \varphi_{\beta_2}(\gamma_2) + \cdots + \varphi_{\beta_k}(\gamma_k))[n]=\varphi_{\beta_1}(\gamma_1) + \cdots + \varphi_{\beta_{k-1}}(\gamma_{k-1}) + (\varphi_{\beta_k}(\gamma_k) [n])$$
• $$\varphi_0(\gamma)=\omega^{\gamma}$$ and $$\varphi_0(\gamma+1) [n] = \omega^{\gamma} \cdot n$$
• $$\varphi_{\beta+1}(0)[n]=\varphi_{\beta}^n(0)$$, where $$\varphi^n$$ denotes function iteration
• $$\varphi_{\beta+1}(\gamma+1)[n]=\varphi_{\beta}^n(\varphi_{\beta+1}(\gamma)+1)$$
• $$\varphi_{\beta}(\gamma) [n] = \varphi_{\beta}(\gamma [n])$$ for a limit ordinal $$\gamma<\varphi_\beta(\gamma)$$
• $$\varphi_{\beta}(0) [n] = \varphi_{\beta [n]}(0)$$ for a limit ordinal $$\beta<\varphi_\beta(0)$$
• $$\varphi_{\beta}(\gamma+1) [n] = \varphi_{\beta [n]}(\varphi_{\beta}(\gamma)+1)$$ for a limit ordinal $$\beta$$

Veblen's function can be presented as a two-argument function $$\varphi_\beta(\gamma)=\varphi(\beta,\gamma)$$.

Note: $$\varphi(0,\gamma)=\omega^\gamma$$, $$\varphi(1,\gamma)=\varepsilon_\gamma$$, $$\varphi(2,\gamma)=\zeta_\gamma$$ and $$\varphi(3,\gamma)=\eta_\gamma$$

### Buchholz hierarchy

In An Independence Result for ($$\Pi_1^1$$-CA) + BI, Wilfried Buchholz discusses an ordinal hierarchy where $$\mu = \psi_0(\varepsilon_{\Omega_\omega + 1})$$, where $$\psi$$ is Buchholz's ordinal collapsing function and $$\psi_0(\varepsilon_{\Omega_\omega + 1})$$ is the TFB ordinal.

### Other hierarchies

We can easily replace the Wainer hierarchy with a different one, such as a hierarchy that works for much larger ordinals such as the small Veblen ordinal or the Takeuti-Feferman-Buchholz ordinal. As long as fundamental sequences are defined, it is possible to extend FGH up to an arbitrary countable ordinal. However, it is impossible to create an effective system that works for all countable ordinals, since a system of fundamental sequences up to $$\omega_1$$ would be nonconstructive.

Selecting fundamental sequences is not an easy problem, since some selections can lead to pathological hierarchies where $$\alpha < \beta$$ does not necessarily imply $$f_\alpha <^* f_\beta$$ (where $$<^*$$ means eventual domination). In a 1976 paper, Diana Schmidt proved a theorem that is useful for identifying and guarding against pathological hierarchies.[1] Given a fundamental sequence system, define $$P(\beta + 1) = \beta$$ and $$\forall \beta \in \text{Lim} : P(\beta) = \beta[0]$$. A fundamental sequence system is built-up if, for all $$\alpha$$ and $$n$$, $$\alpha[n] = P^k(\alpha[n+1])$$ for some $$1 \leq k < \omega$$. Schmidt showed that a built-up fundamental sequence system guarantees that all $$f_\alpha$$ are monotonically increasing and that $$\alpha < \beta \Rightarrow f_\alpha <^* f_\beta$$.

It is possible to define the fast-growing hierarchy for all recursive ordinals, and even for nonrecursive countable ordinals. However, the definitions will necessarily be nonrecursive, making analysis far more complicated. To our knowledge, there has been no research into the creation of nonrecursive fast-growing hierarchies.

## Approximations

Below are some functions in the Wainer hierarchy and Veblen's hierarchy compared to other googological notations.

There are a few things to note:

\begin{eqnarray*} f_0(n) &=& n + 1 \\ f_1(n) &=& f_0^n(n) = ( \cdots ((n + 1) + 1) + \cdots + 1) = n + n = 2n \\ f_2(n) &=& f_1^n(n) = 2(2(\ldots 2(2n))) = 2^n n > 2 \uparrow n \\ f_3(n) &\ge& 2^nn((2^{2^nn})\uparrow\uparrow (n-1)) \ge 2\uparrow\uparrow n\\ f_4(n) &\ge& f_3(n)\uparrow\uparrow\uparrow n \ge 2\uparrow\uparrow\uparrow n \\ f_m(n) &\ge& f_{m-1}(n)\uparrow^{m-1}n \ge 2\uparrow^{m-1} n \\ f_\omega(n) &>\ge& f_\omega(n-1)\uparrow^{n-1}n \ge 2\uparrow^{n-1} n = Ack(n) \\ f_{\omega+1}(n) &>& \lbrace n,n,1,2 \rbrace \\ f_{\omega+2}(n) &>& \lbrace n,n,2,2 \rbrace \\ f_{\omega+m}(n) &>& \lbrace n,n,m,2 \rbrace \\ f_{\omega2}(n) &>& \lbrace n,n,n,2 \rbrace \\ f_{\omega3}(n) &>& \lbrace n,n,n,3 \rbrace \\ f_{\omega m}(n) &>& \lbrace n,n,n,m \rbrace \\ f_{\omega^2}(n) &>& \lbrace n,n,n,n \rbrace \\ f_{\omega^3}(n) &>& \lbrace n,n,n,n,n \rbrace \\ f_{\omega^m}(n) &>& \lbrace n,m+2 [2] 2 \rbrace \\ f_{\omega^{\omega}}(n) &>& \lbrace n,n+2 [2] 2 \rbrace > \lbrace n,n [2] 2 \rbrace \\ f_{\omega^{\omega}+1}(n) &>& \lbrace n,n,2 [2] 2 \rbrace \\ f_{\omega^{\omega}+2}(n) &>& \lbrace n,n,3 [2] 2 \rbrace \\ f_{\omega^{\omega}+m}(n) &>& \lbrace n,n,m+1 [2] 2 \rbrace \\ f_{\omega^{\omega}+\omega}(n) &>& \lbrace n,n,n+1 [2] 2 \rbrace > \lbrace n,n,n [2] 2 \rbrace \\ f_{\omega^{\omega}+\omega+1}(n) &>& \lbrace n,n,1,2 [2] 2 \rbrace \\ f_{\omega^{\omega}+\omega2}(n) &>& \lbrace n,n,n,2 [2] 2 \rbrace \\ f_{\omega^{\omega}+\omega^2}(n) &>& \lbrace n,n,n,n [2] 2 \rbrace \\ f_{{\omega^{\omega}}2}(n) &>& \lbrace n,n [2] 3 \rbrace \\ f_{{\omega^{\omega}}3}(n) &>& \lbrace n,n [2] 4 \rbrace \\ f_{{\omega^{\omega}}m}(n) &>& \lbrace n,n [2] m+1 \rbrace \\ f_{\omega^{\omega+1}}(n) &>& \lbrace n,n [2] n+1 \rbrace > \lbrace n,n [2] n \rbrace \\ f_{\omega^{\omega+2}}(n) &>& \lbrace n,n [2] n,n \rbrace \\ f_{\omega^{\omega+3}}(n) &>& \lbrace n,n,n [2] n,n,n \rbrace \\ f_{\omega^{\omega+m}}(n) &>& \lbrace n,m [2] 1 [2] 2 \rbrace \\ f_{\omega^{\omega2}}(n) &>& \lbrace n,n [2] 1 [2] 2 \rbrace = \lbrace n,2 [3] 2 \rbrace \\ f_{\omega^{\omega3}}(n) &>& \lbrace n,n [2] 1 [2] 1 [2] 2 \rbrace = \lbrace n,3 [3] 2 \rbrace \\ f_{\omega^{\omega m}}(n) &>& \lbrace n,m [3] 2 \rbrace \\ f_{\omega^{\omega^2}}(n) &>& \lbrace n,n [3] 2 \rbrace \\ f_{\omega^{\omega^3}}(n) &>& \lbrace n,n [4] 2 \rbrace \\ f_{\omega^{\omega^m}}(n) &>& \lbrace n,n [m+1] 2 \rbrace \\ f_{\omega^{\omega^\omega}}(n) &>& \lbrace n,n [n+1] 2 \rbrace = \lbrace n,n [1,2] 2 \rbrace \\ f_{^4{\omega}}(n) &>& \lbrace n,n [1 [2] 2] 2 \rbrace \\ f_{^5{\omega}}(n) &>& \lbrace n,n [1 [1,2] 2] 2 \rbrace \\ f_{^6{\omega}}(n) &>& \lbrace n,n [1 [1 [2] 2] 2] 2 \rbrace \\ f_{\varepsilon_0}(n) &>& \lbrace n,n [ [1]] 2 \rbrace \\ f_{\varepsilon_02}(n) &>& \lbrace n,n [ [1]] 3 \rbrace \\ f_{\varepsilon_0m}(n) &>& \lbrace n,n [ [1]] m+1 \rbrace \\ f_{\varepsilon_0\omega}(n) &>& \lbrace n,n [ [1]] n+1 \rbrace \\ f_{\varepsilon_0{\omega^{\omega}}}(n) &>& \lbrace n,n [ [1]] 1 [2] 2 \rbrace \\ f_{\varepsilon_0{\omega^{\omega^{\omega}}}}(n) &>& \lbrace n,n [ [1]] 1 [1,2] 2 \rbrace \\ f_{\varepsilon_0{\omega^{\omega^{\omega^{\omega}}}}}(n) &>& \lbrace n,n [ [1]] 1 [1 [2] 2] 2 \rbrace \\ f_{\varepsilon_0^2}(n) &>& \lbrace n,n [ [1]] 1 [ [1]] 2 \rbrace \\ f_{\varepsilon_0^3}(n) &>& \lbrace n,n [ [1]] 1 [ [1]] 1 [ [1]] 2 \rbrace \\ f_{\varepsilon_0^{\omega}}(n) &>& \lbrace n,n [ [2]] 2 \rbrace \\ f_{\varepsilon_0^{\omega^{\omega}}}(n) &>& \lbrace n,n [ [1,2]] 2 \rbrace \\ f_{\varepsilon_0^{\omega^{\omega^{\omega}}}}(n) &>& \lbrace n,n [[1 [2] 2]] 2 \rbrace \\ f_{\varepsilon_0^{\varepsilon_0}}(n) &>& \lbrace n,n [[1 [ [1]] 2]] 2 \rbrace \\ f_{\varepsilon_0^{\varepsilon_0^{\varepsilon_0}}}(n) &>& \lbrace n,n [[1 [ [1]] 1 [ [1]] 2]] 2 \rbrace \\ f_{\varepsilon_0^{\varepsilon_0^{\varepsilon_0^{\varepsilon_0}}}}(n) &>& \lbrace n,n [[1 [[1 [ [1]] 2]] 2]] 2 \rbrace \\ f_{\varepsilon_1}(n) &>& \lbrace n,n [[[1]]] 2 \rbrace \\ f_{\varepsilon_2}(n) &>& \lbrace n,n [[[ [1]]]] 2 \rbrace \\ f_{\varepsilon_{\omega}}(n) &>& \lbrace n,n [1\backslash1,2] 2 \rbrace \\ f_{\varepsilon_{\omega^2}}(n) &>& \lbrace n,n [1\backslash1,1,2] 2 \rbrace \\ f_{\varepsilon_{\omega^{\omega}}}(n) &>& \lbrace n,n [1\backslash1 [2] 2] 2 \rbrace \\ f_{\varepsilon_{\omega^{\omega^{\omega}}}}(n) &>& \lbrace n,n [1\backslash1 [1,2] 2] 2 \rbrace \\ f_{\varepsilon_{\varepsilon_0}}(n) &>& \lbrace n,n [1\backslash1 [ [1]] 2] 2 \rbrace \\ f_{\varepsilon_{\varepsilon_{\varepsilon_0}}}(n) &>& \lbrace n,n [1\backslash1 [1\backslash1 [ [1]] 2] 2] 2 \rbrace \\ f_{\zeta_0}(n) &>& \lbrace n,n [1\backslash1\backslash2] 2 \rbrace \\ f_{\zeta_0^{\zeta_0}}(n) &>& \lbrace n,n [1 [1\backslash1\backslash2] 2\backslash1\backslash2] 2 \rbrace \\ f_{\varepsilon_{\zeta_0+1}}(n) &>& \lbrace n,n [1\backslash2\backslash2] 2 \rbrace \\ f_{\varepsilon_{\zeta_0+2}}(n) &>& \lbrace n,n [1\backslash3\backslash2] 2 \rbrace \\ f_{\varepsilon_{\varepsilon_{\zeta_0+1}}} &>& \lbrace n,n [1\backslash1 [1\backslash2\backslash2] 2\backslash2] 2 \rbrace \\ f_{\zeta_1}(n) &>& \lbrace n,n [1\backslash1\backslash3] 2 \rbrace \\ f_{\zeta_2}(n) &>& \lbrace n,n [1\backslash1\backslash4] 2 \rbrace \\ f_{\zeta_{\zeta_0}}(n) &>& \lbrace n,n [1\backslash1\backslash1 [1\backslash1\backslash2] 2] 2 \rbrace \\ f_{\eta_0}(n) &>& \lbrace n,n [1\backslash1\backslash1\backslash2] 2 \rbrace \\ f_{\varphi(4,0)}(n) &>& \lbrace n,n [1\backslash1\backslash1\backslash1\backslash2] 2 \rbrace \\ f_{\varphi(\omega,0)}(n) &>& \lbrace n,n [1 [2]\backslash2] 2 \rbrace \\ f_{\varphi(\varphi(\omega,0),0)}(n) &>& \lbrace n,n [1 [1 [1 [2]\backslash2]\backslash2] 2] 2 \rbrace \\ f_{\Gamma_0}(n) &>& \lbrace n,n [1/2] 2 \rbrace \\ f_{\varphi(1,0,0,0)}(n) &>& \lbrace n,n [1 [1\neg4] 2] 2 \rbrace \\ f_{\vartheta(\Omega^{\omega})}(n) &>& \lbrace n,n [1 [1\neg1,2] 2] 2 \rbrace \\ f_{\vartheta(\Omega^{\Omega})}(n) &>& \lbrace n,n [1 [1\neg1\neg2] 2] 2 \rbrace \\ f_{\vartheta(\Omega^{\Omega^{\Omega}})}(n) &>& \lbrace n,n [1 [1 [1\backslash_33] 2] 2] 2 \rbrace \\ f_{\vartheta(\vartheta_1(1))}(n) &>& \lbrace n,n [1 [1\sim3] 2] 2 \rbrace \\ f_{\vartheta(\vartheta_1(2))}(n) &>& \lbrace n,n [1 [1\sim1\sim2] 2] 2 \rbrace \\ f_{\vartheta(\vartheta_1(\omega))}(n) &>& \lbrace n,n [1 [1 [2/_32] 2] 2] 2 \rbrace \\ f_{\vartheta(\vartheta_1(\Omega))}(n) &>& \lbrace n,n [1 [1 [1/2/_32] 2] 2] 2 \rbrace \\ f_{\vartheta(\Omega_2)}(n) &>& \lbrace n,n [1 [1 [1\sim2/_32] 2] 2] 2 \rbrace \\ f_{\vartheta(\Omega_3)}(n) &>& \lbrace n,n [1 [1 [1 [1/_32/_42] 2] 2] 2] 2 \rbrace \\ f_{\vartheta(\Omega_{\omega})}(n) &>& \lbrace n,n [1\bullet2] 2 \rbrace \\ f_{\vartheta(\Omega_{\varepsilon_0})}(n) &>& \lbrace n,n [1 [2/_{1 [1\backslash2] 2}2] 2] 2 \rbrace \\ f_{\vartheta(\Omega_{\Gamma_0})}(n) &>& \lbrace n,n [1 [2/_{1 [1/2] 2}2] 2] 2 \rbrace \\ f_{\vartheta(\Omega_{\vartheta(\Omega_2)})}(n) &>& \lbrace n,n [1 [2/_{1 [1 [1 [1\sim2/_32] 2] 2] 2}2] 2] 2 \rbrace \\ f_{\vartheta(\Omega_{\vartheta(\Omega_3)})}(n) &>& \lbrace n,n [1 [2/_{1 [1 [1 [1 [1/_32/_42] 2] 2] 2] 2}2] 2] 2 \rbrace \\ f_{\vartheta(\Omega_{\vartheta(\Omega_{\omega})})}(n) &>& \lbrace n,n [1 [2/_{1 [1\bullet2] 2}2] 2] 2 \rbrace \\ f_{\vartheta(\Omega_{\vartheta(\Omega_{\vartheta(\Omega_{\omega})})})}(n) &>& \lbrace n,n [1 [2/_{1 [2/_{1 [1\bullet2] 2}2] 2}2] 2] 2 \rbrace \end{eqnarray*}

## Extended Grzegorczyk hierarchy

The Grzegorczyk hierarchy is a hierarchy of functions (specifically - it contains all and only the primitive recursive functions) classified by growth rate. Although the 'extended Grzegorczyk hierarchy' can sometimes be an alternate name for the fast-growing hierarchy, it may also be used as a way of strictly classifying functions based on their growth rates, and a system of fundamental sequences.

## Specific numbers

• 160 is an integer equal to f2(5), and also the number of possible knight moves in 6×6 minichess.
• 212 is the first number n, such that f2(n) is larger than the first noncanonical -illion.
• The isotope radium-212 is the only radium nuclide with negative mass excess. No heavier elements have any isotopes with negative mass excess.
• 384 is an integer equal to f2(6), 8!! and 12!!!!, and also the number of days in some years in the Hebrew calendar.
• 896 is an integer equal to f2(7), and also the number of possible rook moves in chess.
• 1,651 is the smallest number n, for which f2(n) is larger than a googolding. Its prime factorization is 13 × 127.
• 4,608 is an integer equal to f2(9), and also an Achilles number and the number of possible rook moves in quatrochess.
• 491,520 is an integer equal to f2(15). Additionally, it has an unrelated property: As it is equal to 6C4 × 224−1, it also appears in UEFA European Championship-related combinatorics.