Rigid many-one degrees contain infinite antichains of $1$-degrees

TL;DR

This paper demonstrates that many-one degrees of certain sets contain infinite antichains of 1-degrees, revealing a rigidity principle that applies broadly to random and generic sets, and answering Odifreddi's question positively in a measure-theoretic sense.

Contribution

It introduces the concept of m-rigidity and proves that m-rigid sets have many-one degrees containing infinite antichains of 1-degrees, extending previous results and providing a measure-theoretic perspective.

Findings

m-rigid sets have many-one degrees with infinite 1-antichains

every 1-generic set is m-rigid, forming a comeager family

Martin-Löf random reals are m-rigid, making Odifreddi's question true with probability 1

Abstract

Odifreddi asked whether every non-irreducible many-one degree must contain an infinite antichain of one-one degrees. Positive answers are known for computably enumerable many-one degrees (Degtev) and, more recently, for many-one degrees admitting a $\Delta^0_2$ representative (Batyrshin). In this note we isolate a rigidity principle behind these phenomena. Call a set $A\subseteq\omega$ \emph{$m$-rigid} if every total computable $m$-autoreduction of $A$ is eventually the identity. We prove that if $A$ is $m$-rigid, then its many-one degree $\deg_m(A)$ contains an infinite antichain of $1$-degrees. The proof uses a uniform duplication construction: for each computable parameter $S$ we define $B_S\equiv_m A$ so that any injective reduction $B_S\le_1 B_T$ induces an $m$-autoreduction of $A$ and therefore forces $S\subseteq^{*}T$. Choosing an almost-inclusion infinite antichain of computable sets yields the desired infinite $1$-antichain inside $\deg_m(A)$. As applications, Jockusch's rigidity theorem implies that every $1$-generic set is $m$-rigid, giving a comeager family of positive instances. Moreover, $m$-rigidity holds with Lebesgue measure $1$ (indeed, every Martin-L\"of random real is $m$-rigid). Consequently, Odifreddi's Question~5 has a positive answer \emph{with probability $1$} for a fair-coin random $A\in 2^\omega$; any counterexample (if it exists) is confined to a null set (and, by genericity, also to a meager set).