Is the injectivity scheme present in the below axiomatic exposition of a first-order set theory provable in $\text{ZF}$?

Extensionality: $\forall A,B \ [\forall x \ (x \in A \leftrightarrow x \in B) \to A=B]$

Empty: $\exists O \ \forall x \ (x \not \in O)$

Union: $\forall A \ \exists U \ \forall x \ (x \in U \leftrightarrow \exists y \in A \ (x \in y))$

Intersection: $\forall A \ \exists I \ \forall x (x \in I \leftrightarrow \exists y \in A (x \in y) \ \wedge \forall y \in A (x \in y))$

Injectivity: If $\phi(x,A)$ is a formula in which the symbols $``x", ``A"$ occur free and only free and in which the symbol $``B"$ doesn't occur, and $\phi(x,B)$ is the formula obtained from $\phi(x,A)$ by merely replacing all occurrences of the symbol $``A"$ by the symbol $``B"$, then all closures of:

$ \forall wf(A) \ [\forall x \ (\phi(x,A) \to wf(x))] \wedge \\ \forall A,B \ \big{(} \forall x (\phi(x,A) \leftrightarrow \phi(x,B) ) \to A=B\big{)} \\ \to \forall A \ \exists B \ \forall x \ \big{(}x \in B \leftrightarrow \phi(x,A) \big{)} $

are axioms.

where $wf$ stands for the predicate "is well founded", defined as:

$$ wf(x) \iff \not \exists c \ ( x \in c \wedge \forall m \in c \ (\exists n (n \in m \wedge n \in c))) $$

Infinity: $\exists N \ [\emptyset \in N \wedge \ \forall x \in N (x \cup \{x\} \in N)]$

The point of this question is that the above theory (if consistent) then it proves all axioms of $\text{ZF-Foundation}$, and so it interprets the whole of $\text{ZFC}$, Injectivity would prove injective replacement of elements of well-founded sets by well-founded sets, Pairing trivially follow, and Separation follows from having intersection with injective replacement, of course Power trivially follows from Injectivity. However, I couldn't manage to prove the injectivity schema in $\text{ZF-Foundation}$. Which raises suspicion about its consistency with the other axioms.

There is another related aside question, should this prove to be consistent relative to $\text{ZF}$, what should we regard the non-well founded objects that must exist in this theory? Are they sets? if they are not sets? then why did they play a rule in the construction of well-founded set? In a more general sense: If it proves very useful to have non-well founded objects to construct well-founded sets, then how are we to consider those non-well founded objects, would they be considered as "sets"? or still as another kind of objects?

Afternote: The Injectivity axiom schema is inconsistent. Andreas Blass has shown that we can have a set of all ordinals $\geq 2$ and singletons of elements a set $A$, for any arbitrary set $A$, now take $A=\emptyset$, and we get the set of all ordinals $\geq 2$, take its union and then we can get the set of all ordinals, which is inconsistent. I don't think this can be easily fixed.

  • $\begingroup$ Can you elaborate on your after note? I'm not sure that I get what it's saying. $\endgroup$ Mar 5, 2018 at 17:07

1 Answer 1


No, some instances of the injectivity scheme are refutable in ZF. For example, let $\phi(x,A)$ say "either $x$ is an ordinal $\geq2$ or $x=\{y\}$ for some $y\in A$. (To verify the second hypothesis of injectivity for this $\phi$, use that an ordinal $\geq2$ is never a singleton.)

  • $\begingroup$ Perhaps more simply, we could take $\phi(x, A)$ to be $x \not \in A$. Then the Injectivity axiom gives us complements. Taking the complement of the Empty set gives us a universal set, and having Separation available as well gives us the paradoxical Russell set. $\endgroup$ Mar 5, 2018 at 19:01
  • $\begingroup$ @SridharRamesh This violates the first hypothesis of Injectivity. $\endgroup$ Mar 5, 2018 at 20:03
  • $\begingroup$ True. Also, come to think of it, even in Andreas's answer, there is the problem that not all {y} may be well-founded. Still, one imagines some minor modification of this works. $\endgroup$ Mar 6, 2018 at 7:46
  • $\begingroup$ @SridharRamesh Note that my answer only claims refutability in ZF, where all sets are well-founded. In weaker theories like the one suggested in the question, where we don't have the axiom of foundation, it immediately follows that Injectivity isn't provable, but I haven't said that it's refutable there. $\endgroup$ Mar 6, 2018 at 11:34
  • $\begingroup$ @SridharRamesh all $\{y\}$ for each $y \in A$ would be well founded as long as $A$ is well founded, so take $\phi(x,A)$ to be the formula $(x \text{ is an ordinal } \geq 2 \lor \exists y \in A (x=\{y\}))$ then both hypothesis in the antecedent of injectivity would be satisfied for that formula. There is no need for foundation to prove that (I suppose). $\endgroup$ Mar 6, 2018 at 13:17

Your Answer

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

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