2
$\begingroup$

Wikipedia states that, in $ZF$, the Axiom of Choice ($AC$) implies the Hahn-Banach theorem, but that the Hahn-Banach theorem does not imply $AC$. It also states that in $ZF$, the Hahn-Banach theorem implies the paradoxical decomposition of the sphere. My question is simply this:

In $ZF$, does the Hahn-Banach theorem imply some weak choice principle, and if so, what is it? Also in $ZF$, is this weak choice principle (if it exists) equivalent to Hahn-Banach?

$\endgroup$
4
  • 1
    $\begingroup$ Have you seen this thread mathoverflow.net/questions/45844/hahn-banach-without-choice ? Does it answer your question (maybe this thread should be marked as a duplicate in this case)? $\endgroup$ Jan 9, 2015 at 16:34
  • 1
    $\begingroup$ What do you mean when you write a choice principle? $\endgroup$
    – Asaf Karagila
    Jan 9, 2015 at 16:34
  • $\begingroup$ You should obtain a copy of Herrlich "The Axiom of Choice", I think it might answer a lot of your questions before you even ask them here. $\endgroup$
    – Asaf Karagila
    Jan 9, 2015 at 16:35
  • $\begingroup$ @JohannesHahn: It gets me closer, but not quite there.... $\endgroup$ Jan 9, 2015 at 17:15

1 Answer 1

7
$\begingroup$

Over $ZF$, the Hahn-Banach theorem is equivalent to the statement that every Boolean algebra $B$ admits a real-valued measure, i.e., a non-negative real-valued function $\mu$ on $B$ such that $\mu(0)=0$, $\mu(1)=1$ and $\mu(a \vee b)=\mu(a)+\mu(b)$ whenever $a \wedge b = 0$. This statement can be seen as a choice principle via, e.g., Zorn's lemma, which can be used to build such a measure. But it is also possible to construct this measure using just the Boolean Prime Ideal theorem ($BPI$), which is known to be strictly weaker than $AC$. More about this can be found in the usual literature, Howard & Rubin "Consequences of the Axiom of Choice".

$\endgroup$
2
  • $\begingroup$ Does this real-valued measure extend to a probability measure on $B$? Also, what about the equivalence of Hahn-Banach to $WKL_0$ in reverse mathematics? $\endgroup$ Jan 12, 2015 at 12:24
  • $\begingroup$ It's clear that such a measure must take values within $0$ and $1$, but in order to even state countable additivity you would need $B$ to have countable joins and meets, which is not necessarily going to happen. Regarding $WKL_0$, you need to specify which base theory and which statement of Hahn-Banach theorem you are considering. $ZF$ is too strong since it proves Weak König's lemma, and the general formulation of Hahn-Banach theorem as in the above equivalence involves arbitrary vector spaces not necessarily definable in second order arithmetic. $\endgroup$
    – godelian
    Jan 12, 2015 at 13:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.