Cite

This course has no released version yet: the text at this address may change. Cite a released version when one exists.

/courses/cohen-genericity/posner-robinson
@misc{cohen-genericity-posner-robinson,
  title = {{The Posner–Robinson theorem}},
  howpublished = {Lesson 4 of \emph{Cohen genericity}, Computability Theory course notes},
  url = {/courses/cohen-genericity/posner-robinson},
  note = {Unreleased working version}
}

To cite a single statement, use the # link that appears next to its heading.

Lesson 4 of 5

The Posner–Robinson theorem

Coding a set into a 1-generic set so that it can be decoded with any non-computable set; the Posner–Robinson join theorem and Friedberg's jump inversion theorem.

35 min2 questions
Before this lesson:1-generic sets
Cohen genericity · lesson 4 of 5
  1. Genericity and Baire category
  2. Weakly 1-generic sets
  3. 1-generic sets
  4. The Posner–Robinson theorem
  5. Arithmetic genericity

Genericity is usually a tool for building sets with few computational abilities. Combined with coding, it also produces sets with prescribed strength. The idea is to build a 1-generic set GG that carries a given set ZZ in its bits, in such a way that the coding can be read back with the help of an arbitrary non-computable set AA. The resulting theorem, due in this form to Jockusch and Shore, has two classical consequences: the Posner–Robinson theorem, which says that every non-computable set joins with some set to compute the jump of the latter, and Friedberg’s jump inversion theorem, admitted in the course on Turing degrees, which says that every degree above 0′\obf' is a jump.

Theorem 4.1 (Jockusch–Shore).#

Let AA and ZZ be sets, with AA not computable. There is a 1-generic set GG such that Z≤TA⊕GZ \leq_T A \oplus G, Z≤TG⊕∅′Z \leq_T G \oplus \emptyset', and G≤TA⊕Z⊕∅′G \leq_T A \oplus Z \oplus \emptyset'.

4.1 The construction

We may assume that AA is not c.e. Indeed, if AA is c.e., then its complement is not c.e., since AA is not computable; and the complement is Turing equivalent to AA, so it can replace AA in the statement. Fix an enumeration (We)(W_e) of the c.e. sets of strings.

We build strings σ0⪯σ1⪯⋯\sigma_0 \preceq \sigma_1 \preceq \cdots, starting with σ0=ε\sigma_0 = \varepsilon, and let GG be their limit. At step ee, given σe\sigma_e, write i=Z(e)i = Z(e) and consider the set

De={m:some extension of σe i 0m 1 belongs to We}.D_e = \{ m : \text{some extension of } \sigma_e\, i\, 0^m\, 1 \text{ belongs to } W_e \}.

It is c.e. Since AA is not c.e., A≠DeA \neq D_e; so there is mm with either m∈De∖Am \in D_e \setminus A or m∈A∖Dem \in A \setminus D_e. Let mm be the least such number.

  • If m∈De∖Am \in D_e \setminus A, let σe+1\sigma_{e+1} be the first extension of σe i 0m 1\sigma_e\, i\, 0^m\, 1 enumerated into WeW_e. Then GG meets WeW_e.
  • If m∈A∖Dem \in A \setminus D_e, let σe+1=σe i 0m 1\sigma_{e+1} = \sigma_e\, i\, 0^m\, 1. Then no extension of σe+1\sigma_{e+1} belongs to WeW_e: σe+1\sigma_{e+1} avoids WeW_e.
Proof of Theorem 4.1.

GG is 1-generic. At each step ee, GG meets WeW_e or has a prefix avoiding it, so GG meets We∪We⊥W_e \cup W_e^\perp.

G≤TA⊕Z⊕∅′G \leq_T A \oplus Z \oplus \emptyset'. The bit Z(e)Z(e) is read from ZZ. The membership m∈Dem \in D_e is a c.e. question, answered by ∅′\emptyset', and m∈Am \in A is answered by AA; so the least mm is found. In the first case, the first extension enumerated into WeW_e is found by a computable search. So the sequence (σe)(\sigma_e) is computable in A⊕Z⊕∅′A \oplus Z \oplus \emptyset'.

Z≤TA⊕GZ \leq_T A \oplus G. We compute the sequence (σe)(\sigma_e) from A⊕GA \oplus G, by induction. Suppose σe\sigma_e is known. The bit of GG following σe\sigma_e is Z(e)Z(e). Then read the number mm of zeros that follow, up to the next 11; by construction it is the number chosen at step ee. If m∈Am \in A, step ee was in the second case, and σe+1=σe Z(e) 0m 1\sigma_{e+1} = \sigma_e\, Z(e)\, 0^m\, 1. If m∉Am \notin A, step ee was in the first case, so m∈Dem \in D_e, and σe+1\sigma_{e+1} is the first extension of σe Z(e) 0m 1\sigma_e\, Z(e)\, 0^m\, 1 enumerated into WeW_e, found by a computable search, which terminates. Hence A⊕GA \oplus G computes every bit Z(e)Z(e).

Z≤TG⊕∅′Z \leq_T G \oplus \emptyset'. The same decoding works with ∅′\emptyset' in place of AA: given σe\sigma_e, Z(e)Z(e) and mm are read from GG as above, and ∅′\emptyset' decides whether m∈Dem \in D_e, a c.e. question about (e,σe,Z(e),m)(e, \sigma_e, Z(e), m). If m∈Dem \in D_e, step ee was in the first case, since in the second case m∉Dem \notin D_e; otherwise it was in the second case. In both cases σe+1\sigma_{e+1} is then obtained as before.

The decoding from A⊕GA \oplus G uses AA for a single question at each step: is mm in AA? The construction chose mm precisely so that the answer distinguishes the two cases, which is possible because AA is not c.e. A c.e. set AA could coincide with DeD_e, leaving no such mm.

4.2 Consequences

Corollary 4.2 (Posner–Robinson).#

For every non-computable set AA, there is a 1-generic set GG with A⊕G≡TG′≡TG⊕∅′A \oplus G \equiv_T G' \equiv_T G \oplus \emptyset'.

Proof of Corollary 4.2.

Apply Theorem 4.1 with Z=A⊕∅′Z = A \oplus \emptyset': GG is 1-generic, A⊕∅′≤TA⊕GA \oplus \emptyset' \leq_T A \oplus G, and A⊕∅′≤TG⊕∅′A \oplus \emptyset' \leq_T G \oplus \emptyset'. Since GG is 1-generic, G′≡TG⊕∅′G' \equiv_T G \oplus \emptyset' (Theorem 6.3 of “Low and high sets”, with G⊕∅′≤TG′G \oplus \emptyset' \leq_T G' for every set). From ∅′≤TA⊕G\emptyset' \leq_T A \oplus G we get G⊕∅′≤TA⊕GG \oplus \emptyset' \leq_T A \oplus G. From A≤TG⊕∅′A \leq_T G \oplus \emptyset' we get A⊕G≤TG⊕∅′A \oplus G \leq_T G \oplus \emptyset'. Hence A⊕G≡TG⊕∅′≡TG′A \oplus G \equiv_T G \oplus \emptyset' \equiv_T G'.

So every non-computable set, however weak, bridges the gap between some set and its jump. For a low non-computable AA, this gives a set GG with A⊕G≡TG′≥T∅′A \oplus G \equiv_T G' \geq_T \emptyset' while AA alone is far from ∅′\emptyset'.

Corollary 4.3 (Friedberg’s jump inversion theorem).#

For every set Z≥T∅′Z \geq_T \emptyset', there is a 1-generic set GG with G′≡TZG' \equiv_T Z, and moreover G⊕∅′≡TZG \oplus \emptyset' \equiv_T Z.

Proof of Corollary 4.3.

Apply Theorem 4.1 with A=∅′A = \emptyset', which is not computable: GG is 1-generic, Z≤T∅′⊕GZ \leq_T \emptyset' \oplus G and G≤T∅′⊕Z⊕∅′≡TZG \leq_T \emptyset' \oplus Z \oplus \emptyset' \equiv_T Z. Hence G⊕∅′≡TZG \oplus \emptyset' \equiv_T Z, and G′≤TG⊕∅′G' \leq_T G \oplus \emptyset' since GG is 1-generic (Theorem 6.3 of “Low and high sets”). The converse G⊕∅′≤TG′G \oplus \emptyset' \leq_T G' holds for every set, so G′≡TZG' \equiv_T Z.

Corollary 4.4.#

There is a high set that computes no DNC function; in particular, a high set that is not Turing complete.

Proof of Corollary 4.4.

By Corollary 4.3 with Z=∅′′Z = \emptyset'', there is a 1-generic GG with G′≡T∅′′G' \equiv_T \emptyset'', so GG is high. By Theorem 3.4, GG computes no DNC function; since ∅′\emptyset' computes one, GG does not compute ∅′\emptyset'.

This gives a second construction of high incomplete sets, after Corollary 6.7 of “Low and high sets”; the set obtained here is moreover 1-generic and computable in ∅′′\emptyset''.

4.3 Exercises

Exercise 4.5 ★.

Show that every set Z≥T∅′Z \geq_T \emptyset' is the join of ∅′\emptyset' with a set GG such that G̸≥T∅′G \not\geq_T \emptyset'.

Solution

Take GG as in Corollary 4.3: Z≡TG⊕∅′Z \equiv_T G \oplus \emptyset', and GG is 1-generic, so it computes no DNC function, hence not ∅′\emptyset'.

Exercise 4.6 ★.

Let A≤T∅′A \leq_T \emptyset' be non-computable. Show that there is a low 1-generic set GG with A⊕G≡T∅′A \oplus G \equiv_T \emptyset'.

Solution

Apply Theorem 4.1 with Z=∅′Z = \emptyset'. Then GG is 1-generic, ∅′≤TA⊕G\emptyset' \leq_T A \oplus G, and G≤TA⊕∅′≡T∅′G \leq_T A \oplus \emptyset' \equiv_T \emptyset'. Hence A⊕G≤T∅′A \oplus G \leq_T \emptyset', so A⊕G≡T∅′A \oplus G \equiv_T \emptyset'. Since GG is 1-generic and Δ20\Delta^0_2, G′≤TG⊕∅′≡T∅′G' \leq_T G \oplus \emptyset' \equiv_T \emptyset' by Theorem 6.3 of “Low and high sets”, so GG is low.

4.4 Notes and references

Posner and Robinson proved their join theorem in 1981, answering a question on the degrees below 0′\obf'. The general form with an arbitrary coded set ZZ and the use of 1-generic sets is due to Jockusch and Shore (1984). Friedberg’s jump inversion theorem dates from 1957.

  • D. B. Posner and R. W. Robinson, Degrees joining to 0′, Journal of Symbolic Logic 46 (1981), 714–722.
  • C. G. Jockusch and R. A. Shore, Pseudo-jump operators II: transfinite iterations, hierarchies and minimal covers, Journal of Symbolic Logic 49 (1984), 1205–1236.
  • R. M. Friedberg, A criterion for completeness of degrees of unsolvability, Journal of Symbolic Logic 22 (1957), 159–160.