Transfinite combinators and admissible ordinals

I’m back with another topic concerning ordinals!

The transfinite combinator calculus

The transfinite combinator calculus is an extension of normal combinator calculus. We define several related concepts as follows:

  • If an expression \(x\) is beta-reducible, then \(\operatorname{cof}(x)=\operatorname{cof}(\operatorname{reduce}(x))\) and \(x[\alpha]=\operatorname{reduce}(x)[\alpha]\).
  • If \(y\) is a limit, then \(\operatorname{cof}(xy)=\operatorname{cof}(y)\) and \((xy)[\alpha]=x(y[\alpha])\).
  • If \(x\) is a limit, \(y\) is not a limit, and \(xy\) is not beta-reducible, then \(\operatorname{cof}(xy)=\operatorname{cof}(x)\) and \((xy)[\alpha]=(x[\alpha])y\).
  • \(x\) is a limit iff \(\operatorname{cof}(x)\) is defined.
The base of the transfinite combinator calculus are \(S\) and \(K\) which are both not limits. Depending on our application we may choose various infinite combinators to add.

Ordinals in transfinite combinator calculus

I’ve decided for ordinals to be first-class objects due to their importance. (I tried to make one where ordinals are represented by extensions of Church numerals and is rather inelegant and has several problems.) When an ordinal needs to be applied to something else it behaves like its Church numeral. Formally, we can say that \((\alpha+1)xy=\alpha x(xy)\), and the cofinality and fundamental sequences agree with the normal ones. (We assume a system of fundamental sequences exists for all ordinals.) A limit of ordinals agrees with the regular limit. We then introduce two ordinal-based combinators:

  • \(Ax\) for ordinals is the successor ordinal, and otherwise is the Church numeral successor.
  • \(Exy\) if \(y\) is an ordinal \(<\operatorname{cof}(x)\) and \(x\) is a limit is \(x[y]\), otherwise it is undefined.
  • \(Qxy\) checks for equality between ordinals if both are ordinals, otherwise applies the Church numeral comparison operator to its arguments.
  • \(Px\) is the predecessor ordinal if \(x\) is a successor. If \(x\) is a limit ordinal, it returns \(x\). Otherwise, it applies the Church numeral predecessor operator to \(x\).

An OCF for admissible ordinals

We use the transfinite combinator calculus to construuct an OCF for admissible orrdinals. Like Bachmann’s, ours is somewhat cumbersome as it depends on fundamental sequences for all limit ordinals.

For every ordinal \(\alpha\), define a transfinite combinator calculus system \(\mathrm{TCC}_\alpha\) as follows:

  • The combinator \(S\) is not a limit, and satisfies \(Sxyz=xz(yz)\).
  • The combinator \(K\) is not a limit, and satisfies \(Kxy=x\).
  • The combinators \(A\), \(E\), \(Q\), and \(P\) are not limits.
  • The combinator \(\theta_\alpha\) is a limit. If either \(x\) or \(y\) is not an ordinal, its behavior is undefined. If \(y\ge\alpha\), it returns 0. Otherwise, it returns \(\theta_x(y)\).

Then let \[\Omega_\nu=\begin{cases}1 &\mbox{if } \nu=0\\ \aleph_\nu &\mbox{otherwise}\end{cases}\] And then \(\theta_\nu(0)=\omega_\nu\), and otherwise, \(\theta_\nu(\alpha)\) is the supremum of ordinals in \(\Omega_{\nu+1}\) that can be calculated in \(\mathrm{TCC}_\alpha\).

Properties

Obviously \(\theta_\nu(\alpha)\) has cardinality \(\aleph_\nu\). It seems like it takes only admissible values. (Hence why I made ordinals first-class objects, otherwise there is no clear way to stop it from taking the value \(\omega_{\omega_1^\mathrm{CK}}^\mathrm{CK}\).)

  • \(\theta_0(n)=\omega_n^\mathrm{CK}\) for finite \(n\).
  • \(\theta_0(\omega)=\omega_{\omega+1}^\mathrm{CK}\). (It skips over the non-admissible \(\omega_\omega^\mathrm{CK}\).)
  • \(\theta_0(\Omega_1)\) is the first admissible after the fisrt fixed-point of \(\alpha\mapsto\omega_\alpha^\mathrm{CK}\).
DISCLAIMER: These values are incorrect.

Leave a Comment

Your email address will not be published. Required fields are marked *