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/n-genericity
@misc{cohen-genericity-n-genericity,
  title = {{Arithmetic genericity}},
  howpublished = {Lesson 5 of \emph{Cohen genericity}, Computability Theory course notes},
  url = {/courses/cohen-genericity/n-genericity},
  note = {Unreleased working version}
}

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

Lesson 5 of 5

Arithmetic genericity

n-generic sets meet or avoid every Σ⁰ₙ set of strings; their n-th jump is computed from the set and the n-th jump of the empty set; they decide every Σ⁰ₙ property by a prefix; the hierarchy of genericity notions.

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

The notions of the previous lessons relativize. A set that is 1-generic relative to ∅(n−1)\emptyset^{(n-1)} meets or avoids every Σn0\Sigma^0_n set of strings, and this is enough to control its first nn jumps and to decide every Σn0\Sigma^0_n property by a finite prefix. This lesson proves these facts, compares the notions of genericity introduced in the course, and builds nn-generic sets of the smallest possible complexity.

Recall from Definition 3.1 that GG is nn-generic, for n≥1n \geq 1, if it is 1-generic relative to ∅(n−1)\emptyset^{(n-1)}; by Post’s theorem, the ∅(n−1)\emptyset^{(n-1)}-c.e. sets of strings are exactly the Σn0\Sigma^0_n ones. Every set is considered 00-generic.

5.1 Comparison of the notions

Proposition 5.1.#

For every n≥1n \geq 1, every nn-generic set is weakly nn-generic, and every weakly (n+1)(n+1)-generic set is nn-generic.

Proof of Proposition 5.1.

If WW is a dense Σn0\Sigma^0_n set of strings, no string avoids WW, so W⊥=∅W^\perp = \emptyset, and an nn-generic set meets W∪W⊥=WW \cup W^\perp = W.

Let GG be weakly (n+1)(n+1)-generic and WW a Σn0\Sigma^0_n set of strings. The set W⊥={σ:∀τ⪰σ τ∉W}W^\perp = \{ \sigma : \forall \tau \succeq \sigma\ \tau \notin W \} is Πn0\Pi^0_n. So W∪W⊥W \cup W^\perp is the union of a Σn0\Sigma^0_n and a Πn0\Pi^0_n set, hence Δn+10\Delta^0_{n+1}, in particular Σn+10\Sigma^0_{n+1}. It is dense. So GG meets it.

This gives the chain

weakly 1-generic⇐1-generic⇐weakly 2-generic⇐2-generic⇐⋯\text{weakly 1-generic} \Leftarrow \text{1-generic} \Leftarrow \text{weakly 2-generic} \Leftarrow \text{2-generic} \Leftarrow \cdots

None of these implications can be reversed. For the implications from weak (n+1)(n+1)-genericity to nn-genericity, this follows from Theorem 5.5 below together with the relativization of Corollary 2.4: there is an nn-generic set computable in ∅(n)\emptyset^{(n)}, and no weakly (n+1)(n+1)-generic set is computable in ∅(n)\emptyset^{(n)}. For the implications from nn-genericity to weak nn-genericity, the separation is due to Kurtz and is admitted here. For n=1n = 1, it follows from two results also admitted: every hyperimmune degree contains a weakly 1-generic set (Kurtz), so some weakly 1-generic set has c.e. degree, while no 1-generic set has c.e. degree (Jockusch).

5.2 The jumps of an n-generic set

The lowness property of 1-generic sets relativizes.

Lemma 5.2.#

If GG is 1-generic relative to AA, then (G⊕A)′≤TG⊕A′(G \oplus A)' \leq_T G \oplus A'.

Proof of Lemma 5.2.

For each ee, let We={σ:Φeσ⊕A(e) ⁣ ⁣↓}W_e = \{ \sigma : \Phi^{\sigma \oplus A}_e(e)\converge \}, where the computation with oracle σ⊕A\sigma \oplus A only queries the GG-side below ∣σ∣|\sigma|. The sets WeW_e are AA-c.e., uniformly in ee. Since GG is 1-generic relative to AA, it has a prefix in WeW_e or a prefix avoiding WeW_e. Both questions are decided by A′A', uniformly in ee and in the prefix: "σ∈We\sigma \in W_e" is Σ10(A)\Sigma^0_1(A), and ”σ\sigma avoids WeW_e” is Π10(A)\Pi^0_1(A). With oracle G⊕A′G \oplus A', search through the prefixes G ⁣↾kG \uh k for one that is in WeW_e or avoids WeW_e; the search terminates. In the first case ΦeG⊕A(e)\Phi^{G \oplus A}_e(e) converges; in the second case, it diverges, since a convergent computation would have its GG-use below some k′≥∣G ⁣↾k∣k' \geq |G \uh k|, and G ⁣↾k′G \uh k' would be an extension of G ⁣↾kG \uh k in WeW_e.

Theorem 5.3.#

Let n≥1n \geq 1 and let GG be nn-generic. Then G(k)≡TG⊕∅(k)G^{(k)} \equiv_T G \oplus \emptyset^{(k)} for every k≤nk \leq n.

Proof of Theorem 5.3.

By induction on kk. For k=0k = 0 there is nothing to prove. Let k<nk < n and suppose G(k)≡TG⊕∅(k)G^{(k)} \equiv_T G \oplus \emptyset^{(k)}. The jump is invariant under Turing equivalence, so G(k+1)≡T(G⊕∅(k))′G^{(k+1)} \equiv_T (G \oplus \emptyset^{(k)})'. Since k≤n−1k \leq n - 1, GG is 1-generic relative to ∅(k)\emptyset^{(k)}: indeed it is 1-generic relative to ∅(n−1)\emptyset^{(n-1)}, and every ∅(k)\emptyset^{(k)}-c.e. set is ∅(n−1)\emptyset^{(n-1)}-c.e. By Lemma 5.2 with A=∅(k)A = \emptyset^{(k)}, (G⊕∅(k))′≤TG⊕∅(k+1)(G \oplus \emptyset^{(k)})' \leq_T G \oplus \emptyset^{(k+1)}. Conversely, G≤T(G⊕∅(k))′G \leq_T (G \oplus \emptyset^{(k)})' and ∅(k+1)≤T(G⊕∅(k))′\emptyset^{(k+1)} \leq_T (G \oplus \emptyset^{(k)})', since the jump of a set computes the jump of every set below it. Hence G(k+1)≡TG⊕∅(k+1)G^{(k+1)} \equiv_T G \oplus \emptyset^{(k+1)}.

The hypothesis is used at the step from kk to k+1k+1 only through genericity relative to ∅(k)\emptyset^{(k)}, which is why nn-genericity controls exactly the first nn jumps.

5.3 Deciding Σ⁰ₙ properties

A Σn0\Sigma^0_n property of sets is a property P(X)P(X) defined by a Σn0\Sigma^0_n formula of arithmetic with a set parameter XX. By the relativized form of Post’s theorem, there is an index ee such that P(X)P(X) holds if and only if ΦeX(n−1)(0) ⁣ ⁣↓\Phi^{X^{(n-1)}}_e(0)\converge, for every set XX. For (n−1)(n-1)-generic sets, the oracle X(n−1)X^{(n-1)} can be replaced by X⊕∅(n−1)X \oplus \emptyset^{(n-1)}, and 1-genericity relative to ∅(n−1)\emptyset^{(n-1)} decides the resulting Σ10(∅(n−1))\Sigma^0_1(\emptyset^{(n-1)}) property.

Theorem 5.4.#

Let n≥1n \geq 1, let PP be a Σn0\Sigma^0_n property, and let GG be nn-generic. There is a prefix σ\sigma of GG such that either every (n−1)(n-1)-generic set extending σ\sigma satisfies PP, or no (n−1)(n-1)-generic set extending σ\sigma satisfies PP. In particular, P(G)P(G) holds if and only if PP holds for every (n−1)(n-1)-generic set extending some prefix of GG.

Proof of Theorem 5.4.

Let ee be such that P(X)  ⟺  ΦeX(n−1)(0) ⁣ ⁣↓P(X) \iff \Phi^{X^{(n-1)}}_e(0)\converge for every XX. The reduction X(n−1)≤TX⊕∅(n−1)X^{(n-1)} \leq_T X \oplus \emptyset^{(n-1)} of Theorem 5.3, for (n−1)(n-1)-generic sets XX, is given by a procedure that does not depend on XX: the proof of Lemma 5.2 describes it uniformly, and the induction composes these procedures. So there is an index dd such that P(X)  ⟺  ΦdX⊕∅(n−1)(0) ⁣ ⁣↓P(X) \iff \Phi^{X \oplus \emptyset^{(n-1)}}_d(0)\converge for every (n−1)(n-1)-generic set XX. For n=1n = 1 we take d=ed = e, and the equivalence holds for every set.

Let W={τ:Φdτ⊕∅(n−1)(0) ⁣ ⁣↓}W = \{ \tau : \Phi^{\tau \oplus \emptyset^{(n-1)}}_d(0)\converge \}, a Σn0\Sigma^0_n set of strings. Since GG is nn-generic, it has a prefix σ∈W\sigma \in W or a prefix σ\sigma avoiding WW. In the first case, every set X≻σX \succ \sigma satisfies ΦdX⊕∅(n−1)(0) ⁣ ⁣↓\Phi^{X \oplus \emptyset^{(n-1)}}_d(0)\converge, so every (n−1)(n-1)-generic X≻σX \succ \sigma satisfies PP. In the second case, no X≻σX \succ \sigma satisfies ΦdX⊕∅(n−1)(0) ⁣ ⁣↓\Phi^{X \oplus \emptyset^{(n-1)}}_d(0)\converge, since a convergent computation would place a prefix of XX extending σ\sigma in WW; so no (n−1)(n-1)-generic X≻σX \succ \sigma satisfies PP. Since GG is itself (n−1)(n-1)-generic and extends σ\sigma, P(G)P(G) holds exactly in the first case.

Taking negations, the same holds for Πn0\Pi^0_n properties. The restriction to (n−1)(n-1)-generic extensions cannot be removed for n≥2n \geq 2: the Σ20\Sigma^0_2 property ”XX is finite” fails for every 1-generic set, but every string has a finite extension, so no string has only infinite extensions.

5.4 n-generic sets below the n-th jump

Theorem 5.5.#

For every n≥1n \geq 1, there is an nn-generic set G≤T∅(n)G \leq_T \emptyset^{(n)}. Such a set satisfies G(n)≡T∅(n)G^{(n)} \equiv_T \emptyset^{(n)}.

Proof of Theorem 5.5.

Let (We)(W_e) be an enumeration of the Σn0\Sigma^0_n sets of strings, uniform in the sense that {(e,τ):τ∈We}\{ (e, \tau) : \tau \in W_e \} is Σn0\Sigma^0_n; for instance WeW_e is the ee-th set of strings c.e. relative to ∅(n−1)\emptyset^{(n-1)}. Build σ0=ε⪯σ1⪯⋯\sigma_0 = \varepsilon \preceq \sigma_1 \preceq \cdots with oracle ∅(n)\emptyset^{(n)}. At step ee, ask whether σe\sigma_e has an extension in WeW_e; this is a Σn0\Sigma^0_n question, answered by ∅(n)\emptyset^{(n)}. If yes, search for such an extension τ\tau, using ∅(n)\emptyset^{(n)} to decide membership in WeW_e, and let σe+1=τ 0\sigma_{e+1} = \tau\, 0 for the first one found. If no, σe\sigma_e avoids WeW_e, and σe+1=σe 0\sigma_{e+1} = \sigma_e\, 0. The lengths increase strictly, so the limit GG is a set; it meets or avoids every WeW_e, so it is nn-generic, and G≤T∅(n)G \leq_T \emptyset^{(n)}. By Theorem 5.3, G(n)≡TG⊕∅(n)≡T∅(n)G^{(n)} \equiv_T G \oplus \emptyset^{(n)} \equiv_T \emptyset^{(n)}.

For n=1n = 1, this is the existence of low 1-generic sets, already seen in the course on Turing degrees. A set GG with G(n)≡T∅(n)G^{(n)} \equiv_T \emptyset^{(n)} is called lown\mathrm{low}_n.

5.5 Exercises

Exercise 5.6 ★.

A set GG is GLn\mathrm{GL}_n if G(n)≡T(G⊕∅′)(n−1)G^{(n)} \equiv_T (G \oplus \emptyset')^{(n-1)}. Show that every nn-generic set is GLn\mathrm{GL}_n.

Solution

Since G⊕∅′≤TG′G \oplus \emptyset' \leq_T G', we have (G⊕∅′)(n−1)≤TG(n)(G \oplus \emptyset')^{(n-1)} \leq_T G^{(n)} for every set. Conversely, G≤T(G⊕∅′)(n−1)G \leq_T (G \oplus \emptyset')^{(n-1)} and ∅(n)=(∅′)(n−1)≤T(G⊕∅′)(n−1)\emptyset^{(n)} = (\emptyset')^{(n-1)} \leq_T (G \oplus \emptyset')^{(n-1)}, so by Theorem 5.3, G(n)≡TG⊕∅(n)≤T(G⊕∅′)(n−1)G^{(n)} \equiv_T G \oplus \emptyset^{(n)} \leq_T (G \oplus \emptyset')^{(n-1)}.

Exercise 5.7 ★★.

Let GG be 2-generic. Show that GG is not computable in ∅′\emptyset', and that GG is not high, that is, ∅′′̸≤TG′\emptyset'' \not\leq_T G'.

Hint

For a Turing functional Φe\Phi_e, consider the set of strings τ\tau such that Φeτ⊕∅′(x) ⁣ ⁣↓=0\Phi^{\tau \oplus \emptyset'}_e(x)\converge = 0 for some x∈∅′′x \in \emptyset''.

Solution

By Proposition 5.1, a 2-generic set is weakly 2-generic, so it is not computable in ∅′\emptyset' by Corollary 2.4.

By Theorem 5.3, G′≡TG⊕∅′G' \equiv_T G \oplus \emptyset', so it suffices to show ΦeG⊕∅′≠∅′′\Phi^{G \oplus \emptyset'}_e \neq \emptyset'' for every ee. Let

Ue={τ:∃x (Φeτ⊕∅′(x) ⁣ ⁣↓=0 and x∈∅′′)}.U_e = \{ \tau : \exists x\ (\Phi^{\tau \oplus \emptyset'}_e(x)\converge = 0 \text{ and } x \in \emptyset'') \}.

The first condition is Σ10(∅′)\Sigma^0_1(\emptyset'), hence Σ20\Sigma^0_2, and ∅′′\emptyset'' is Σ20\Sigma^0_2, so UeU_e is Σ20\Sigma^0_2. If GG meets UeU_e, then ΦeG⊕∅′(x)=0\Phi^{G \oplus \emptyset'}_e(x) = 0 for some x∈∅′′x \in \emptyset'', so ΦeG⊕∅′≠∅′′\Phi^{G \oplus \emptyset'}_e \neq \emptyset''. Otherwise some prefix σ\sigma of GG avoids UeU_e. Suppose, for a contradiction, that ΦeG⊕∅′=∅′′\Phi^{G \oplus \emptyset'}_e = \emptyset''. Then

x∉∅′′  ⟺  ∃τ⪰σ Φeτ⊕∅′(x) ⁣ ⁣↓=0.x \notin \emptyset'' \iff \exists \tau \succeq \sigma\ \Phi^{\tau \oplus \emptyset'}_e(x)\converge = 0.

Indeed, if x∉∅′′x \notin \emptyset'', a long enough prefix of GG gives the value 00; if x∈∅′′x \in \emptyset'', no extension of σ\sigma gives 00, since σ\sigma avoids UeU_e. The right-hand side is Σ10(∅′)\Sigma^0_1(\emptyset'), so the complement of ∅′′\emptyset'' is Σ20\Sigma^0_2, and ∅′′\emptyset'' is Δ20\Delta^0_2, that is, computable in ∅′\emptyset'. This contradicts the hierarchy theorem.

The argument shows more generally: if GG is 1-generic relative to ZZ and CC is c.e. in ZZ but not computable in ZZ, then C̸≤TG⊕ZC \not\leq_T G \oplus Z.

5.6 Notes and references

The arithmetic forcing relation was defined by Feferman (1965), and nn-generic sets were studied systematically by Jockusch (1980), who proved most of the results of this lesson. Kurtz (1981) introduced weak nn-genericity and separated it from nn-genericity. Jockusch showed that 1-generic degrees are never c.e.; Kurtz showed that every hyperimmune degree contains a weakly 1-generic set.

  • S. Feferman, Some applications of the notions of forcing and generic sets, Fundamenta Mathematicae 56 (1965), 325–345.
  • C. G. Jockusch, Degrees of generic sets, in Recursion Theory: its Generalisations and Applications, London Mathematical Society Lecture Note Series 45, Cambridge University Press, 1980, 110–139.
  • S. A. Kurtz, Randomness and genericity in the degrees of unsolvability, PhD thesis, University of Illinois at Urbana-Champaign, 1981.