δφ is a googological system introduced by a Japanese Googologist Jason, and is the second system defined by shifting definition. It has not been formalised.

It is intended to perform as an ordinal function associated to Veblen function through shifting definition. A restricted system of δφ consists of the constant term $$0$$, an associative $$2$$-ary function $$+$$, and variadic functions $$\delta \varphi(x_1,\ldots,x_k)$$ and $$\delta \varphi([d]x_1,\ldots,x_k)$$. The term $$d$$ in the latter expression plays a role to indicate a definition, which is coded into an ordinal, of an ordinal function.

## Explanation

When $$x_1 = 0$$, then $$x_1$$ is often omitted. For example, $$\delta \varphi([d])$$ is a shorthand of $$\delta \varphi([d]0)$$, and $$\delta \varphi([d],x_2)$$ is a shorthand of $$\delta \varphi([d]0,x_2)$$. Moreover, the expressions $$\delta \varphi(x_1,x_2)$$ and $$\delta \varphi([d]0,x_2)$$ in the restricted system themselves are shorthands of $$\delta \varphi(x_1,x_2)$$ and $$\delta \varphi(0,[d]0,[0,0]x_2)$$, and hence the full system of δφ is much more complicated.

We explain how it is intended to work. As we noted above, the $$2$$-ary function $$+$$ does not play a role of the addition. Therefore we denote by $$+_{\delta \varphi}$$ the $$2$$-ary function $$+$$ in order to distinguish it from the addition.

The $$1$$-ary $$\delta \varphi(x_1)$$ coincides with $$\omega^{x_1} = \varphi(0,x_1)$$, and the $$2$$-ary function $$x +_{\delta \varphi} y$$ coincides with $$x + y$$ as long as $$y$$ is smaller than $$\varepsilon_0$$. On the other hand, the $$2$$-ary function $$\delta \varphi(x_1,x_2)$$ behaves in a tricky way, The initial value $$\delta \varphi(1,0)$$, which we will denote by $$A$$, coincides with $$\varphi(1,0) = \varepsilon_0$$. However, $$A +_{\delta \varphi} A$$ is intended to be $$\varphi(2,0) = \zeta_0$$, which is much larger than $$A + A = \varepsilon_0 \times 2$$. The ordinal $$\varepsilon_0 \times 2$$ is expressed as $$A +_{\delta \varphi} \delta \varphi([A]0)$$. Similarly, we have $$A +_{\delta \varphi} \delta \varphi([A]0) +_{\delta \varphi} \delta \varphi([A]0) = \varepsilon_0 \times 3$$, and $$+_{\delta \varphi} \delta \varphi([A]0)$$ plays a role of $$+ \varepsilon_0$$. Although the expression $$\delta \varphi([A]0)$$ itself is not a normal expression in this system, it is harmless to regard it as $$\varepsilon_0$$ as long as we interpret occurrences of $$\delta \varphi([A]0)$$ in a normal expression.

The next significant expression is $$A +_{\delta \varphi} \delta \varphi(\delta \varphi([A]0) +_{\delta \varphi} \delta \varphi(0))$$. By $$\delta \varphi(0) = \omega^0 = 1$$, $$x +_{\delta \varphi} y = x + y$$ as long as $$y$$ is smaller than $$\varepsilon_0 = A$$, and $$\delta \varphi(x_1) = \omega^{x_1}$$, it is the sum of $$A$$ and $$\omega^{A+1} = A \times \omega$$ with respect to $$+_{\delta \varphi}$$. The value is intended to coincide with $$A + \omega^{A+1} = A \times \omega$$, and hence we do not have to care about the difference of $$+_{\delta \varphi}$$ and $$+$$ in this realm. Indeed, we have $$A +_{\delta \varphi} y = A + y$$ for any normal expression $$A +_{\delta \varphi} y$$ smaller than $$A +_{\delta \varphi} A$$.

The $$1$$-ary function $$\delta \varphi([A]x_1)$$ plays a role of $$\varphi(1,x_1)$$ although the expression itself is not normal. For example, we have $$A +_{\delta \varphi} \delta \varphi([A] \delta \varphi([A]0)) = A + \varepsilon_{\varepsilon_0} = \varepsilon_{\varepsilon_0}$$. The limit of ordinals expressed by $$0$$, a single occurrence of $$A$$, $$+_{\delta \varphi}$$, $$\delta \varphi(x_1)$$, and $$\delta \varphi([A],x_1)$$ is $$\zeta_0$$, which is expressed as $$A +_{\varphi \delta} A$$.

Similarly, $$\delta \varphi([A +_{\delta \varphi} A]x_1)$$ plays a role of $$\varphi(2,x_1)$$, and $$A +_{\varphi \delta} A +_{\varphi \delta} A$$ is intended to coincide with $$\varphi(3,0) = \eta_0$$. Continuing a similar computation, we obtain the following analysis:

• $$\delta \varphi(A +_{\delta \varphi} \delta \varphi(0))$$, which is often abbreviated to $$A \times \omega$$ although it does not coincides with $$\varepsilon_0 \times \omega$$, is intended to coincide with $$\varphi(\omega,0)$$, and $$\delta \varphi([\delta \varphi(A +_{\delta \varphi} \delta \varphi(0))]x_1)$$ plays a role of $$\varphi(\omega,x_1)$$.
• $$\delta \varphi(\delta \varphi(A +_{\delta \varphi} \delta \varphi(0)))$$, which is often abbreviated to $$A^{\omega}$$ although it does not coincides with $$\varepsilon_0^{\omega}$$, is intended to coincide with the small Veblen ordinal, and $$\delta \varphi([\delta \varphi(\delta \varphi(A +_{\delta \varphi} \delta \varphi(0)))]x_1)$$ plays a role of the enumeration of fixed points of multivariable Veblen functions.
• $$\delta \varphi(\delta \varphi(A +_{\delta \varphi} A))$$, which is often abbreviated to $$A^A$$ although it does not coincides with $$\varepsilon_0^{\varepsilon_0}$$, is intended to coincide with the large Veblen ordinal, and $$\delta \varphi([\delta \varphi(\delta \varphi(A +_{\delta \varphi} A))]x_1)$$ plays a role of the enumeration of fixed points of transfinte-variable Veblen functions.
• $$\delta \varphi([A]A +_{\delta \varphi} \delta \varphi(0))$$, which might be abbreviated to $$\varepsilon_{A+1}$$ although it does not coincides with $$\varepsilon_{\varepsilon_0+1}$$, is intended to coincide with the Bachmann-Howard ordinal.

We should recall that we have not used $$\delta \varphi(1,1)$$ in the computation above. It is intended to coincide with $$\psi(\Omega_2^{\Omega_2})$$ with respect to a certain OCF, and is expanded as $$\delta \varphi([\varphi([\cdots \varphi([A]A +_{\delta \varphi} \delta \varphi(0)) \cdots]A +_{\delta \varphi} \delta \varphi(0))]A +_{\delta \varphi} \delta \varphi(0))$$.

## Issue

In this section, we argue on an issue on δφ. In order to formalise δφ, we need to define values such as $$\delta \varphi([A]0)$$, although it is not a normal expression. According to Jason, $$\delta \varphi([A],0)$$ is expected to be smaller than $$A$$, and $$A +_{\delta \varphi} \delta \varphi([A],0)$$ is expected to coincide with $$\varepsilon_0 \times 2$$, which is greater than $$A +_{\delta \varphi} \beta$$ for any ordinal $$\beta$$ below $$\varepsilon_0$$. On the other hand, $$A$$ is intended to "correspond" to $$\varepsilon_0$$, and every ordinal below $$A$$ can be expressed by $$0$$, $$+_{\delta \varphi}$$, and $$\delta \varphi(x_1)$$. It implies that there is no ordinal $$\alpha$$ such that $$\alpha$$ is smaller than $$A$$ but $$A +_{\delta \varphi} \alpha$$ differs from $$A +_{\delta \varphi} \beta$$ for any ordinal $$\beta$$ below $$\varepsilon_0$$. In order to avoid such an obvious contradiction, we need to justify the equalities above in terms of ordinal types in the following way:

Let $$C$$ denote the set of ordinals which can be expressed by $$0$$, the usual addition $$+$$, and variadic functions $$\delta \varphi(x_1,\ldots,x_k)$$ and $$\delta \varphi([d],x_1,\ldots,x_k)$$ with respect to a certain restriction on the normality of expressions so that every ordinal in $$C$$ admits a unique normal expression. For each $$\alpha \in C$$, we denote by $$o(\alpha$$\) the ordinal type of the strict well-ordered set $$(C \cap \alpha,\in) = (\{\beta \in C \mid \beta < \alpha\},\in)$$. Forgetting the properties of $$\delta \varphi$$ such as $$A = \varepsilon_0$$ explained in the previous section, assume the following alternative conditions:

1. $$\delta \varphi(x_1)$$ coincides with $$\omega^{x_1}$$ for any $$x_1$$.
2. $$\delta \varphi(1,0)$$ is a sufficiently large ordinal such as $$\Omega$$.
3. The normality of an expression of ordinals below $$\varepsilon_0$$ coincides with the normality in Cantor normal form.
4. There is no ordinal $$\alpha \in C$$ satisfying $$\varepsilon_0 \leq \alpha < A$$.

Then we have $$o(A) = \varepsilon_0$$. It does not contradict if we define $$\delta \varphi([A]0)$$ as $$\varepsilon_0$$ as long as it is not a normal expression. In particular, the properties $$o(A) = \varepsilon_0$$, $$\delta \varphi([A]0) < A$$, $$o(A + \delta \varphi([A]0)) = \varepsilon_0 \times 2$$, and $$o(A + A) = \zeta_0$$ are consistent. Since $$o$$ does not necessarily commute with $$+$$ as we expect $$o(A+A) = \zeta_0 \neq \varepsilon_0 \times 2 = o(A) + o(A)$$, the usual addition plays the role of $$+_{\delta \varphi}$$ without a modification. As a result, the equalities $$\alpha = \beta$$ between expressions $$\alpha$$ in δφ and actual ordinals $$\beta$$ above should be regarded as short hands of the equalities $$o(\alpha) = \beta$$.

## Sources

1. The user page of Jason in Japanese Googology Wiki.
3. Jason, δφのいろいろ, Japanese Googology WIki user blog.

By Aeton: Okojo numbers · N-growing hierarchy
By 新井 (Arai): Arai's $$\psi$$
By バシク (BashicuHyudora): Primitive sequence number · Pair sequence number · Bashicu matrix system 1/2/3/4
By ふぃっしゅ (Fish): Fish numbers (Fish number 1 · Fish number 2 · Fish number 3 · Fish number 4 · Fish number 5 · Fish number 6 · Fish number 7 · S map · SS map · s(n) map · m(n) map · m(m,n) map) · Bashicu matrix system 1/2/3/4 computation programmes · TR function (I0 function)
By じぇいそん (Jason): Irrational arrow notation · δOCF · δφ · ε function
By 甘露東風 (Kanrokoti): KumaKuma ψ function
By 小林銅蟲 (Kobayashi Doom): Sushi Kokuu Hen
By koteitan: Bashicu matrix system 2.3
By mrna: 段階配列表記 · 降下段階配列表記 · 多変数段階配列表記 · SSAN · S-σ
By Naruyoko Naruyo: Y sequence computation programme · ω-Y sequence computation programme
By Nayuta Ito: N primitive · Flan numbers · Large Number Lying on the Boundary of the Rule of Touhou Large Number 4
By p進大好きbot: Large Number Garden Number
By たろう (Taro): Taro's multivariable Ackermann function
By ゆきと (Yukito): Hyper primitive sequence system · Y sequence · YY sequence · Y function · ω-Y sequence