On homeomorphisms of Cantor space that induce

arXiv:1908.05381v1 [math.LO] 15 Aug 2019

only the trivial Turing automorphism

Bjørn Kjos-Hanssen∗
August 16, 2019

Abstract
To determine whether there is a nontrivial automorphism of the Turing degrees
remains a major open problem of computability theory. Past results have limited
how nontrivial automorphisms could possibly be, and ruled out that an automor-
phism might be induced by a function on integers.
A homeomorphism F of the Cantor space is said to induce a uniform map mod
finite if for each a ∈ ω there is a b ∈ ω such that for all X and Y, if X(n) = Y(n) for
all n ≥ b then F(X)(m) = F(Y)(m) for all m ≥ a.
We show that if such an F induces an automorphism π of the Turing degrees,
then π is trivial. This generalizes, and provides an easier proof of, the past result
where F was assumed to be induced by a permutation of ω.

Dedicated to the celebration of the work of Theodore A. Slaman and W. Hugh Woodin

Contents
1 Introduction 1

2 Excluding permutations by recursion 2

3 Excluding maps on Cantor space mod finite 3

1 Introduction
Let DT denote the set of Turing degrees and let ≤ denote its ordering. This article gives
a partial answer to the following famous question.
Question 1. Does there exist a nontrivial automorphism of DT ?
Definition 2. A bijection π : DT → DT is an automorphism of DT if for all x, y ∈ DT ,
x ≤ y iff π(x) ≤ π(y). If moreover there exists an x with π(x) , x then π is nontrivial.
∗ This work was partially supported by a grant from the Simons Foundation (#315188 to Bjørn Kjos-

Hanssen) and by the Institute for Mathematical Sciences, National University of Singapore.

1
 Question 1 has a long history. Already in 1977, Jockusch and Solovay [2] showed
that each jump-preserving automorphism of the Turing degrees is the identity above
0(4) . Nerode and Shore 1980 [4] showed that each automorphism (not necessarily jump-
preserving) is equal to the identity on some cone {a : a ≥ b}. Slaman and Woodin [5]
showed that each automorphism is equal to the identity on the cone above 0′′ . and
[5, 6] that Aut(D) is countable.
An obstacle to reducing the base of the cone to 0′ and ultimately 0 is that Turing
reducibility is Σ03 , but not Π02 or Σ02 in the sense of descriptive set theory.
In the other direction, S. Barry Cooper [1] claimed to construct a nontrivial auto-
morphism, induced by a discontinuous function on ωω , itself induced by a function on
ω<ω . That claim was not independently verified. In [3] we attacked the problem by
ruling out a certain simple but natural possibility: automorphisms induced by permu-
tations of finite objects. We showed that no permutation of ω represents a nontrivial
automorphism of the Turing degrees. That proof was too complicated, in a way, and
did not extend from DT down to Dm . Here we give a more direct proof using the shift
map n 7→ n + 1. Our proof here will generalize to a certain class of homeomorphisms,
distinct from the class of such homeomorphisms that the result in [3] generalizes to.

2 Excluding permutations by recursion

Lemma 3. Suppose θ : ω → ω is a bijection such that for all computable f : ω → ω,
θ−1 ◦ f ◦ θ is computable. Then θ is computable.
Proof. Let f (n) = n + 1 and k(n) = (θ−1 ◦ f ◦ θ)(n). Then for any m, k(θ−1 (m)) =
θ−1 ( f (m)) and so we compute θ−1 by recursion:

θ−1 (m + 1) = k(θ−1 (m)) 

Lemma 4. Suppose σ ∈ 2<ω , g : ω → ω, and Φ is a Turing functional, satisfying

(∀n)(∀τ  σ)(∃ρ  τ)(Φρ (n) ↓ and ρ ◦ g(n) = Φρ (n)). (1)

For any ρ  σ and n, if Φρ (n) ↓, then g(n) < |ρ|.

Proof. If instead g(n) ≥ |ρ| then ρ(g(n)) is undefined. So let τ  ρ, τ(g(n)) = 1 − Φρ(n).
This τ violates (1). 
Lemma 5. If g : ω → ω is injective and Φ is a Turing functional such that

{B : B ◦ g = ΦB }

is nonmeager, then g is computable.

Proof. By assumption, it is not the case that

(∀σ)(∃n)(∃τ  σ)(∀ρ  τ)(Φρ (n) ↓→ ρ ◦ g(n) , Φρ (n)).

So we have

(∃σ)(∀n)(∀τ  σ)(∃ρ  τ)(Φρ (n) ↓ and ρ ◦ g(n) = Φρ (n)).

2
Pick such a σ: then Φ cannot make a mistake above σ, and we can always extend to
get the right answer.
As finite data we assume we know the values of n and g(n) for which g(n) < |σ|.
We compute the value g(n) as follows.
Check the finite database of {(k, g(k)) : g(k) < |σ|}, and output g(n) is found. Oth-
erwise we know g(n) ≥ |σ|.
By dovetailing computations, find a ρ0  σ such that Φρ0 (n). By Lemma 4 we
have that g(n) < |ρ0 |. Thus, g(n) ∈ I where I is the closed interval [|σ|, |ρ0 | − 1]. Let
a ∈ I, b ∈ I, a < b. It suffices to show how to eliminate either a or b as a candidate for
being equal to g(n).
Let τ ≻ σ be such that τ(a) , τ(b) and let ρ  τ be such that Φρ (n) ↓. Then mark
as eliminated whichever c ∈ {a, b} makes ρ(c) , Φρ (n). Thus we one-by-one eliminate
all a ∈ I until only one candidate remains. 
Theorem 6. No permutation of the integers can induce a nontrivial automorphism of
Dr , for any reducibility ≤r between ≤1 and ≤T .
Proof. Suppose θ : ω → ω is a permutation (bijection) and consider any injective
recursive f . For any B, let A = B ◦ θ−1 ◦ f (so x ∈ A iff f (x) ∈ B ◦ θ−1 ). We
have A ≤1 B ◦ θ−1 and so by assumption A ◦ θ ≤T B ◦ θ−1 ◦ θ = B. This gives
(B ◦ θ−1 ◦ f ) ◦ θ = ΦB for some Turing functional Φ. Let g = θ−1 ◦ f ◦ θ. Since for each
B and f there exists such a Φ, for each f there must be some Φ such that the Gδ set
{B : B ◦ g = ΦB }
is nonmeager. By Lemma 5, g is computable. By Lemma 3, θ is computable.
But this means that for any A, A ◦ θ ≤1 A, so that the represented automorphism π
is everywhere-decreasing: π(x) ≤1 x. Applying this to π−1 we get π(x) ≡1 x. 

3 Excluding maps on Cantor space mod finite

Making computability-theoretic uniformity assumptions is an easy way to rule out cer-
tain possible Turing automorphisms, but we will not discuss that further as we are more
interested in uniformity of a simpler, or purely combinatorial, kind.
Definition 7. Let [a, ∞) = {n ∈ ω : a ≤ n}. A map Φ uniformly induces a map on 2ω
mod finite if for each a there is a b such that for all A, B, if A ↾ [b, ∞) = B ↾ [b, ∞)
then F(A) ↾ [a, ∞) = F(B) ↾ [a, ∞].
Some continuous maps F : 2ω → 2ω induce maps F̃ : 2ω / =∗ → 2ω / =∗ but do not
have the uniform property:
Example 8. Let us code an alphabet of size 4 into 2ω . Then ΦX will look for the first 2
in X. When it appears, if X(0) = 0, then output 2; if X(0) = 1, then output 3. In other
words,

X X(n) + X(0) if n is the minimal m such that X(m) = 2,


Φ (n) = 
X(n)
 otherwise
Then ΦX =∗ X for all X, so A =∗ B implies ΦA =+ ΦB , but not uniformly.

3
Lemma 9. If F(A) = A ◦ f for a permutation f : ω → ω then F induces a uniform
map mod finite.
Proof. Let a ∈ ω. Let b = min{ f (n) : n ≥ a}. Suppose X(m) = Y(m) for all m ≥ b.
Then X ◦ f (n) = Y ◦ f (n) for all n ≥ a. 
Lemma 10. Suppose Θ : 2ω → 2ω is a homeomorphism such that for all computable
f : ω → ω, the function Θ−1 ◦ f ∗ ◦ Θ is computable. Then Θ is computable.
Proof. Let πn : ω → ω be the constant n function. For π∗n : 2ω → 2ω , note that
π∗n (A)(u) = (A ◦ πn )(u) = A(n), so π∗n (A) ∈ {0ω , 1ω }. Let f (n) = n + 1. Note that

(π∗n ◦ f ∗ )(A)(u) = π∗n ( f ∗ (A))(u) = (( f ∗ (A)) ◦ πn )(u) = ( f ∗ (A))(n)

= A( f (n)) = A(n + 1) = (A ◦ πn+1 )(u) = (π∗n+1 (A))(u)

so π∗n ◦ f ∗ = π∗n+1 . Also, f ◦ πn = πn+1 . Let

Φ = Θ−1 ◦ f ∗ ◦ Θ.

By assumption, Φ is computable. Then

Θ ◦ Φ = f∗ ◦ Θ

π∗n ◦ Θ ◦ Φ = π∗n ◦ f ∗ ◦ Θ
π∗n ◦ Θ ◦ Φ = π∗n+1 ◦ Θ
(π∗n ◦ Θ) ◦ Φ = π∗n+1 ◦ Θ
Since homeomorphisms have finite use, π∗0 ◦ Θ is just a finite amount of information,
and so we can recursively compute π∗n+1 ◦ Θ this way. 
Remark 11. In terms of Odifreddi’s notation where σn is the nth truth table, Θ(A) = {n :
A |= σt(n) } for some (not a priori computable) t, and π∗0 ◦Θ(A) = {n : A |= σt(0) } ∈ {∅, ω}.
For a continuous map Ψ : 2ω → 2ω , let the lower-case version ψ : ω → ω pick out the
associated truth tables. That is,

Ψ(A) = {n : A |= σψ(n) }.

Let Φ(A) = {n : A |= σφ(n) }. Let Σ(A) = {n : A |= σn }. Then Φ(A) = φ∗ (Σ(A)), since

Φ(A)(n) = 1 ↔ A |= σφ(n) ,

and
φ∗ (Σ(A))(n) = 1 ↔ Σ(A)(φ(n)) = 1 ↔ A |= σφ(n) .
The Turing functional Σ is left-invertible as we can effectively pick out a list of truth
tables that simply return the answer to “n ∈ A?” So we have Σ−1 Σ(A) = A. However,
Σ is not onto (not every list of answers is a coherent list of answers to truth table

4
questions) so Σ−1 is not total and we do not in general have A = ΣΣ−1 A. Fortunately,
the domain of Σ−1 is a Π01 class given by a computable tree with no dead ends. Now

(π∗n ◦ Θ) ◦ Φ = π∗n+1 ◦ Θ

becomes
π∗n θ∗ Σφ∗ Σ = π∗n+1 θ∗ Σ
π∗n θ∗ Σφ∗ = π∗n+1 θ∗
Now π∗n θ∗ = π∗θ(n) , so
π∗θ(n) Σφ∗ = π∗θ(n+1)
We may express this in oracle notation with ΞnA = π∗n (A), ΞnA (u) = A(n), as
A ΦA
ΞΘ Θ
n+1 = Ξn

A ΦA
ΞΘ Θ
n+1 (u) = Ξn (u)
A
ΘA (n + 1) = ΘΦ (n)
Theorem 12. Let F be a homeomorphism of 2ω which is induced by a uniform map
mod finite, Θ. Then F is computable.
Proof. The proof follows that of Theorem 6, with Lemma 10 playing the role of
Lemma 3. Our new assumption of uniformity mod finite makes the proof of a lemma
corresponding to Lemma 5 go through. 
Example 13 (The inductive procedure in Lemma 10.). Suppose Φ is the truth table
reduction given by ΦA (n) = A(2n) · A(2n + 1). Suppose ΘA (0) = A(2) → A(3). Then
A
ΘA (1) = ΘΦ (0) = ΦA (2) → ΦA (3)
= A(4)A(5) → A(6)A(7).

Next,
A
ΘA (2) = ΘΦ (1)
= ΦA (4)ΦA (5) → ΦA (6)ΦA (7)
= A(8)A(9)A(10)A(11) → A(12)A(13)A(14)A(15).

Example 14 (Another example of Lemma 10). Let Θ be given by





 h1, 1i if hA(2n), A(2n + 1)i = h0, 0i,

h0, 1i if hA(2n), A(2n + 1)i = h0, 1i,


hΘA (2n), ΘA(2n + 1)i = 





 h1, 0i if hA(2n), A(2n + 1)i = h1, 0i,


h0, 0i if hA(2n), A(2n + 1)i = h1, 1i,

= h¬A(2n + 1), ¬A(2n)i

5
which satisfies Θ = Θ−1 . Then Φ = Θ−1 ◦ f ∗ ◦ Θ applied to the Thue-Morse sequence
t is given by:

t = 0110 1001 1001 0110 . . .

→Θ 0110 1001 1001 0110 . . .
→f∗ 1101 0011 0010 1101 . . .
→Θ−1 0001 1100 1110 0001 . . .

Now ΘA (0) = ¬A(1). And ΘA (1) = ¬ΦA (1). In particular:

Θt (0) = ¬t(1) = 0,
t
t
Θ (1) = ΘΦ (0) = ¬Φt (1) = 1,
t
Θt (2) = ΘΦ (1) = ¬Φt (0) = 1

We computed this using Φ above but we see that it is also correct for Θ.
Actually ΘA (n) = ¬A(n + (−1)n mod 2 ) so f ∗ (ΘA )(n) = ¬A(n + 1 + (−1)n+1 mod 2 ) and
n mod 2
 
Θ−1 ( f ∗ (ΘA ))(n) = A [n + (−1)n mod 2 ] + 1 + (−1)[n+(−1) ]+1 mod 2

To simplify this, note that

[n + (−1)n mod 2 ] + 1 mod 2 = n mod 2

so  
Θ−1 ( f ∗ (ΘA ))(n) = A [n + (−1)n mod 2 ] + 1 + (−1)n mod 2
= A(n + 3)[n even], A(n − 1)[n odd]
and ΘA (n) is the negation of A(n + 1)[n even], A(n − 1)[n odd]. And then the claim is
A
that ΘA (n + 1) is ΘΦ (n).
Example 15. Let Θ(A) = A + K mod 2, where K is your favorite non-computable
set. Then Θ−1 ( f ∗ (Θ(A))) is the sequence A(1) + K(1) + K(0), A(2) + K(2) + K(1), . . .
If this were a computable operator (sequence of truth tables) then so is the sequence
K(1) + K(0), K(2) + K(1), . . . But then K would be computable: start by knowing K(0).
Then K(1) = (K(1) + K(0)) + K(0), and so on inductively.
Remark 16. Woodin mentioned on June 6, 2019 that he and Slaman may have shown,
in unpublished work from the 1990s, that each automorphism of Da , the degrees of
arithmetical reducibility, is represented by a continuous function outright. This gives
some extra interest in a possible future Da version of our results.

6
References
[1] S. Barry Cooper. The Turing universe is not rigid. University of Leeds Pure
Mathematics Preprint Series 1997, no. 16 (revised February 1998).
[2] Carl G. Jockusch, Jr. and Robert M. Solovay. Fixed points of jump preserving
automorphisms of degrees. Israel J. Math., 26(1):91–94, 1977.
[3] Bjørn Kjos-Hanssen. Permutations of the integers induce only the trivial automor-
phism of the Turing degrees. Bull. Symb. Log., 24(2):165–174, 2018.
[4] Anil Nerode and Richard A. Shore. Reducibility orderings: theories, definability
and automorphisms. Ann. Math. Logic, 18(1):61–89, 1980.
[5] Theodore A. Slaman. Global properties of the Turing degrees and the Turing jump.
In Computational prospects of infinity. Part I. Tutorials, volume 14 of Lect. Notes
Ser. Inst. Math. Sci. Natl. Univ. Singap., pages 83–101. World Sci. Publ., Hacken-
sack, NJ, 2008.
[6] Theodore A. Slaman and Hugh Woodin. Definabil-
ity in degree structures. Online draft, July 2005.
URL:https://math.berkeley.edu/˜slaman/talks/sw.pdf.