I know ZF is not finitely axiomatizable so a "minimal set of axioms for ZF" is actually a minimal set of metaxioms (or axiom schemata) that quantify (in natural language) over well-formed-formulas of first order logic (with equality symbol), like separation or replacement.

EDIT: I want to read a discussion about "minimal sets" of axioms for ZF(C) set theory from these usual axioms which have a name. By a "minimal list" I mean a non-redundant axiomatics. An example of a list, in the usual ZFC formulations, the "minimal" axioms would be (1) extensionality, (2) union, (3) pair, (4) infinity, (5) substitution, (6) choice. Separation and power come out with (6), the empty comes out via separation. Another list is Bourbaki's. A Professor of mine said me that "He [Bourbaki] has a very nice formulation that uses (1) extensionality, (2) pair, (3) parts, (4) infinity, (5) separation and union (in a single axiom, but I count two axioms for my purposes). It is well known that he did not use the axiom of choice (AC) because using Hilbert's epsilon (his famous "tau") allows him to demonstrate AC".

My purpose is didactic. I am seeking some lists like that and the discussions of preferring some above others.

  • $\begingroup$ So what is your question? $\endgroup$ – Asaf Karagila Oct 16 '18 at 12:02
  • 1
    $\begingroup$ @AsafKaragila The title I think ... $\endgroup$ – Noah Schweber Oct 16 '18 at 12:20
  • $\begingroup$ @Noah: Ah. Right. I got confused, because the title isn't part of the body. $\endgroup$ – Asaf Karagila Oct 16 '18 at 12:21
  • 1
    $\begingroup$ I don't understand the part "Separation and power come out with (6)". Huh? $\endgroup$ – Asaf Karagila Oct 17 '18 at 19:55
  • 1
    $\begingroup$ Also, you might be interested in math.stackexchange.com/questions/916072/… $\endgroup$ – Asaf Karagila Oct 17 '18 at 19:55

The earliest serious result in this direction I know is due to Levy. Levy showed that the seemingly-weak theory $$\mbox{$T_0=$ Z + $parameter$-$free$ replacement}$$ proves full replacement.

Levy, and later (and some unintentional overlap) Schindler and Schlicht, looked at axiomatizations of ZF (and related) which avoid using parameters in the separation/replacement schemes. (Levy's paper is in the somewhat-hard-to-get-hold-of Tarski symposium proceedings, and I don't think there's an online version.)

Moving forward, I think some of Harvey Friedman's work addresses this problem - see e.g. this paper of his.

On the other hand, I believe Mostowski showed that any c.e. axiomatization of ZF (or similar) has a proper c.e. sub-axiomatization (the c.e. requirement is actually WLOG; given a proper sub-axiomatization $Y$ of $X$, fix $y\in X\setminus Y$ and let $Y'=X\setminus\{y\}$). So this rules out the existence of any truly minimal axiomatization. But I can't find a source for this or prove it on my own at the moment, so take this with a grain of salt.

  • $\begingroup$ @Noah A short account of the argument for Levy's result is available at Ralf Schindler's page. $\endgroup$ – Andrés E. Caicedo Oct 16 '18 at 12:49
  • $\begingroup$ @AndrésE.Caicedo Hm, that proves weaker results than what I remembered, so it looks like my memory is wrong. $\endgroup$ – Noah Schweber Oct 16 '18 at 12:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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