Transitive Closure of a Well-Founded Relation is Well-Founded (without Axiom of Choice)

1.2k Views Asked by At

I am interested in proving the titular claim:

Transitive Closure of a Well-Founded Relation is Well-Founded (without Axiom of Choice)

My approach:

Let $R$ be a well-founded relation. We construct the transitive closure $R^t$ as the union of the range of $F:\mathbb{N}\to \mathrm{fld}(R)\times\mathrm{fld}(R)$ defined (inductively) by $$ F(0)=R\qquad\text{and}\qquad F(n+1)=F(n)\circ R $$ for every $n\in\mathbb{N}$. That is to say, $R^t\overset{\mathrm{def}}{=}\bigcup\mathrm{ran}(F)=\bigcup_{i=0}^\infty F(i)$...

My thought process:

I can show that for each natural number $n$, $\bigcup_{i=0}^nF(i)$ is well-founded given that $R$ is well-founded (by induction on $n$), but that doesn't imply that $R^t$ is well-founded. It seems I need some sort of "transfinite induction" argument where if all the "finite" predecessors $\bigcup_{i=0}^nF(i)$ are well-founded, then $\bigcup_{i=0}^\infty F(i)$ is well-founded (but I don't think we necessarily have a well-ordered set to make such an argument).

Any hints/advice is much appreciated!

(The reason I want to work in ZF is because Enderton's Elements of Set Theory pg 244 gives a proof using AC, but remarks as an exercise to give a proof without it)

3

There are 3 best solutions below

6
On BEST ANSWER

Suppose that $R$ is well-founded, and let $A$ be any nonempty subset of its field. We wish to show that $A$ contains an $R^t$-minimal element.

Let $B$ consist of all elements of $A$, plus every element of every finite $R$-sequence that connects two elements of $A$.

Because $R$ is well-founded, $B$ contains an $R$-minimal element $b$. The elements that were added to $B$ as intermediate elements of $R$-sequences cannot, by construction, be $R$-minimal, so we must have $b\in A$.

I claim that $b$ is $R^t$-minimal in $A$ -- because it it were not, it would sit at the top of an $R$-sequence that witnesses $\langle a,b\rangle\in R^t$, and the penultimate element of that sequence would be a predecessor of $b$ in $B$, contradicting $b$ being $R$-minimal in $B$.

Thus, $R^t$ is well founded.

2
On

The main use of choice in this proof would be to show there is no decreasing chains. But in the absence of choice we need to return to the minimal elements definitions. For simplicity, denote by $S$ the field of $R$ and $R^t$.

So suppose that $A\subseteq S$ is a non-empty set. It has some $R$-minimal element $a$, now show that from this assumption that $a$ is $R$-minimal, it is impossible that there is any $b\in A$ for which $\langle b,a\rangle\in R^t$. To do so, recall that if $\langle x,y\rangle\in R^t$, then there is a finite chain of elements such that each "link" in the chain belongs to $R$.

4
On

I don't think it is necessarily true in the absence of choice.

Assume that countable choice fails -- i.e. there is a family of nonempty sets $(A_n)_{n\in\omega}$ which has no choice function. Without loss of generality assume the $A_n$s are pairwise disjoint and also disjoint from $\omega$.

Now let $$R=\{\langle a,n\rangle\mid n\in\omega, a\in A_{n+1}\}\cup\{\langle n,a\rangle\mid n\in\omega, a\in A_n\}$$

Then $R$ is well-founded, because an infinitely descending $R$-chain would pick an element from each of cofinitely many of the $A_n$s.

On the other hand $R^t$ is not well-founded, due to the descending chain $$ \cdots 3 \mathrel{R^t} 2 \mathrel{R^t} 1 \mathrel{R^t} 0 $$