In recursion theory, α recursion theory is a generalisation of recursion theory to subsets of admissible ordinals α {\displaystyle \alpha } . An admissible set is closed under Σ 1 ( L α ) {\displaystyle \Sigma _{1}(L_{\alpha })} functions, where L ξ {\displaystyle L_{\xi }} denotes a rank of Godel's constructible hierarchy. α {\displaystyle \alpha } is an admissible ordinal if L α {\displaystyle L_{\alpha }} is a model of Kripke–Platek set theory. In what follows α {\displaystyle \alpha } is considered to be fixed.
The objects of study in α {\displaystyle \alpha } recursion are subsets of α {\displaystyle \alpha } . These sets are said to have some properties:
There are also some similar definitions for functions mapping α {\displaystyle \alpha } to α {\displaystyle \alpha } :[3]
Additional connections between recursion theory and α recursion theory can be drawn, although explicit definitions may not have yet been written to formalize them:
We say R is a reduction procedure if it is α {\displaystyle \alpha } recursively enumerable and every member of R is of the form ⟨ H , J , K ⟩ {\displaystyle \langle H,J,K\rangle } where H, J, K are all α-finite.
A is said to be α-recursive in B if there exist R 0 , R 1 {\displaystyle R_{0},R_{1}} reduction procedures such that:
If A is recursive in B this is written A ≤ α B {\displaystyle \scriptstyle A\leq _{\alpha }B} . By this definition A is recursive in ∅ {\displaystyle \scriptstyle \varnothing } (the empty set) if and only if A is recursive. However A being recursive in B is not equivalent to A being Σ 1 ( L α [ B ] ) {\displaystyle \Sigma _{1}(L_{\alpha }[B])} .
We say A is regular if ∀ β ∈ α : A ∩ β ∈ L α {\displaystyle \forall \beta \in \alpha :A\cap \beta \in L_{\alpha }} or in other words if every initial portion of A is α-finite.
Shore's splitting theorem: Let A be α {\displaystyle \alpha } recursively enumerable and regular. There exist α {\displaystyle \alpha } recursively enumerable B 0 , B 1 {\displaystyle B_{0},B_{1}} such that A = B 0 ∪ B 1 ∧ B 0 ∩ B 1 = ∅ ∧ A ≰ α B i ( i < 2 ) . {\displaystyle A=B_{0}\cup B_{1}\wedge B_{0}\cap B_{1}=\varnothing \wedge A\not \leq _{\alpha }B_{i}(i<2).}
Shore's density theorem: Let A, C be α-regular recursively enumerable sets such that A < α C {\displaystyle \scriptstyle A<_{\alpha }C} then there exists a regular α-recursively enumerable set B such that A < α B < α C {\displaystyle \scriptstyle A<_{\alpha }B<_{\alpha }C} .
Barwise has proved that the sets Σ 1 {\displaystyle \Sigma _{1}} -definable on L α + {\displaystyle L_{\alpha ^{+}}} are exactly the sets Π 1 1 {\displaystyle \Pi _{1}^{1}} -definable on L α {\displaystyle L_{\alpha }} , where α + {\displaystyle \alpha ^{+}} denotes the next admissible ordinal above α {\displaystyle \alpha } , and Σ {\displaystyle \Sigma } is from the Levy hierarchy.[5]
There is a generalization of limit computability to partial α → α {\displaystyle \alpha \to \alpha } functions.[6]
A computational interpretation of α {\displaystyle \alpha } -recursion exists, using " α {\displaystyle \alpha } -Turing machines" with a two-symbol tape of length α {\displaystyle \alpha } , that at limit computation steps take the limit inferior of cell contents, state, and head position. For admissible α {\displaystyle \alpha } , a set A ⊆ α {\displaystyle A\subseteq \alpha } is α {\displaystyle \alpha } -recursive iff it is computable by an α {\displaystyle \alpha } -Turing machine, and A {\displaystyle A} is α {\displaystyle \alpha } -recursively-enumerable iff A {\displaystyle A} is the range of a function computable by an α {\displaystyle \alpha } -Turing machine. [7]
A problem in α-recursion theory which is open (as of 2019) is the embedding conjecture for admissible ordinals, which is whether for all admissible α {\displaystyle \alpha } , the automorphisms of the α {\displaystyle \alpha } -enumeration degrees embed into the automorphisms of the α {\displaystyle \alpha } -enumeration degrees.[8]
Some results in α {\displaystyle \alpha } -recursion can be translated into similar results about second-order arithmetic. This is because of the relationship L {\displaystyle L} has with the ramified analytic hierarchy, an analog of L {\displaystyle L} for the language of second-order arithmetic, that consists of sets of integers.[9]
In fact, when dealing with first-order logic only, the correspondence can be close enough that for some results on L ω = HF {\displaystyle L_{\omega }={\textrm {HF}}} , the arithmetical and Levy hierarchies can become interchangeable. For example, a set of natural numbers is definable by a Σ 1 0 {\displaystyle \Sigma _{1}^{0}} formula iff it's Σ 1 {\displaystyle \Sigma _{1}} -definable on L ω {\displaystyle L_{\omega }} , where Σ 1 {\displaystyle \Sigma _{1}} is a level of the Levy hierarchy.[10] More generally, definability of a subset of ω over HF with a Σ n {\displaystyle \Sigma _{n}} formula coincides with its arithmetical definability using a Σ n 0 {\displaystyle \Sigma _{n}^{0}} formula.[11]