1 Introduction
Assuming the Axiom of Choice, the infinite cardinal numbers can be constructed inductively in many ways, one of which is the following:
An interesting concept is what results of this definition when one replaces the use of arbitrary wellorderings at each stage by only those which are appropriately computable. Let us be more precise. Recall Gödel’s constructible hierarchy given by
$L_0 = \varnothing $
,
$L_{\alpha +1} =$
all sets definable over
$L_\alpha $
from finitely many elements of
$L_\alpha $
, and
$L_\lambda = \bigcup _{\alpha <\lambda }L_\alpha $
at limit stages. A subset of an ordinal
$\alpha $
is
$\alpha $
-recursively enumerable (
$\alpha $
-r.e.) if it is
$\Sigma _1$
-definable over
$L_\alpha $
with parameters. Note that this definition makes sense for arbitrary
$\alpha $
.
Definition 1.1. The effective cardinal numbers are defined inductively as follows:
We have chosen to define effective cardinals in terms of recursive enumerability and not plain recursiveness (where a set is recursive if it is both r.e. and co-r.e.); the two notions will coincide in all cases of interest, but the computations involved will be simpler this way. Recall that an ordinal
$\alpha $
is said to be admissible if
$L_\alpha $
is a model of Kripke–Platek set theory, the result of removing from Zermelo–Fraenkel set theory the axioms of Powerset and Replacement and adding the axioms of Separation and Collection, both restricted to formulae in which only bounded quantifiers appear. For simplicity, we will assume that
${\mathsf {KP}}$
includes the schema of foundation for all formulae. For an ordinal
$\alpha $
, we denote by
$\alpha ^+$
the smallest admissible ordinal greater than
$\alpha $
. According to our definition, all admissible ordinals are effective cardinals, though the converse is not true.
Definition 1.2. Let
$\alpha $
,
$\beta $
be ordinals.
$\eta _\alpha $
is said to be
$\beta $
-Gandy if
$\alpha ^+\leq \eta _{\alpha +\beta }$
.
Our definition is a generalization of that of a Gandy ordinal (which coincides with that of a
$1$
-Gandy ordinal). These ordinals were first studied by H. Friedman (unpublished) and Gostanian [Reference Gostanian9] (and indirectly by Solovay [Reference Solovay22]) and named by Abramson and Sacks [Reference Abramson and Sacks1].
A question of interest is that of the admissibility of
$\eta _\alpha $
; another one is the related question of whether
$\eta _\alpha $
is
$\beta $
-Gandy, for a given
$\beta $
. In this article, we shall present the main properties of effective cardinals and present an application to the reverse mathematics of determinacy. For context, we mention the following result, which is commonly known:
Theorem 1.3 (Aczel, Richter; Gostanian; Grilliot; Solovay)
The following are equivalent over
${\mathsf {KPl}}:$
-
(1) There is an ordinal which is not $1$
-Gandy. -
(2) $\Sigma ^0_2$
-determinacy.
Proof Sketch
Theorem 1.3 is obtained by combining several well-known results: first, by
${\mathsf {KPl}}$
, we have access to Shoenfield’s absoluteness, so we may assume
$V = L$
in both directions. By work of Gostanian [Reference Gostanian9], the existence of an ordinal which is not
$1$
-Gandy is equivalent to the existence of an ordinal which is
$\Pi _1^{+}$
-reflecting (in the terminology of [Reference Aguilera3]). This is equivalent to
$\Sigma ^1_1$
-reflection by Aczel–Richter [Reference Aczel, Richter, Fenstad and Hinman2]. By Aczel–Richter [Reference Aczel, Richter, Fenstad and Hinman2], this ordinal, if it exists, is equal to the closure ordinal of
$\Sigma ^1_1$
-inductive definitions. By a theorem of Grilliot (unpublished, but see [Reference Aguilera and Kerchove4] for a proof), this ordinal, if it exists, is also the closure ordinal of monotone
$\Sigma ^1_1$
-inductive definitions. A theorem of Solovay (unpublished, but see [Reference Kechris12] or [Reference Moschovakis17]) asserts that if this ordinal
$\sigma ^1_1$
exists, then all
$\Sigma ^0_2$
games won by Player I have a winning strategy in
$L_{\sigma ^1_1}$
. As noted by Welch [Reference Welch24], the argument also shows that all other games are won by Player II, as witnessed by a strategy in any admissible set containing
$L_{\sigma ^1_1}$
. (In fact, they appear in a strict initial segment of such an admissible set; see [Reference Aguilera and Lubarsky6].) For the converse, if
$\Sigma ^0_2$
-determinacy holds, then a theorem of Tanaka [Reference Tanaka23] asserts that all monotone
$\Sigma ^1_1$
inductive definitions reach a fixpoint. By Grilliot’s theorem, all non-monotone
$\Sigma ^1_1$
inductive definitions have a fixpoint, so that there is a
$\Sigma ^1_1$
-reflecting ordinal, by Aczel-Richter [Reference Aczel, Richter, Fenstad and Hinman2] and thus an ordinal which is not
$1$
-Gandy, by Gostanian [Reference Gostanian9].
The ordinal of Theorem 1.3 has many equivalent characterizations, some of which we have mentioned; we refer the reader to [Reference Aguilera and Lubarsky5] for a compilation of some others.
We shall prove the following analogue of Theorem 1.3:
Theorem 1.4. The following are equivalent over
${\mathsf {KPl}}:$
-
(1) There is an ordinal which is not $\omega $
-Gandy. -
(2) $\Sigma ^0_4$
-determinacy.
As an immediate consequence, we obtain the following boldface result in terms of a relativized form of Gandiness. Below, the notion of an ordinal
$\xi $
being
$\alpha $
-Gandy relative to
$x\in \mathbb {R}$
is defined just like before, except that Definitions 1.1 and 1.2 are modified to speak of the
$L[x]$
-hierarchy instead of the L-hierarchy.
Corollary 1.5. The following are equivalent over
${\mathsf {KPl}}:$
-
(1) For each $x\in \mathbb {R}$
, there is an ordinal which is not
$\omega $
-Gandy relative to x. -
(2) $\boldsymbol \Sigma ^0_4$
-determinacy.
We do not know if there is any kind of analogue of Theorem 1.3 and Theorem 1.4 for
$\Sigma ^0_3$
sets, though the existence of an ordinal which is not
$2$
-Gandy implies the consistency of
$\Sigma ^0_3$
-determinacy and indeed of Z
$_2$
.
Let us finish this introduction by briefly recalling the history of
$\Sigma ^0_4$
-determinacy.
The first proof of
$\Sigma ^0_4$
-determinacy was obtained as a consequence of Martin’s [Reference Martin13] proof of
$\Sigma ^1_1$
-determinacy, which assumed the existence of a measurable cardinal. The first proof that did not require assumptions beyond
${\mathsf {ZFC}}$
was that of Paris [Reference Paris18]. Martin’s [Reference Martin14] proof of Borel Determinacy showed that
$\Sigma ^0_4$
-determinacy is indeed provable in Z
$_3$
and further work of his (see [Reference Martin15]) showed this cannot be improved to Z
$_2$
. Martin’s unprovability result built on work of Friedman [Reference Friedman8] whereby
$\Sigma ^0_5$
-determinacy is not provable in Z
$_2$
. A strengthening of Martin’s unprovability result due to Montalbán and Shore [Reference Montalbán and Shore16] showed that Z
$_2$
cannot even prove that all Boolean combinations of
$\Sigma ^0_3$
games are determined. However, Martin has shown (see [Reference Martin15]) that all wellfounded models of Z
$_2$
satisfy
$\Delta ^0_4$
-determinacy. A previous characterization of
$\Sigma ^0_4$
-determinacy in the spirit of Reverse Mathematics was carried out by Hachtman [Reference Hachtman10]; the proof of ours will rely on Hachtman’s result.
2 Basic properties
We shall use basic properties of the constructible universe L without much detail. We refer the reader to [Reference Barwise7], [Reference Jech11], [Reference Simpson21], or [Reference Sacks20] for background. In talking about formulae which define sets over structures of the form
$L_\alpha $
, it will be convenient to recall the existence of Gödel’s pairing function, which allows multiple elements of
$L_\alpha $
to be coded by tuples
$\langle x,y,z\rangle $
.
$L_\alpha $
is closed under Gödel’s pairing function whenever
$\alpha $
is a multiplicatively indecomposable ordinal.
We have mentioned this before, but let us state it explicitly.
Lemma 2.1. Suppose
$\alpha $
is admissible. Then,
$\alpha $
is an effective cardinal.
Proof. This is simply because if
$\prec $
is a wellorder (in the real world) which belongs to
$L_\alpha $
, then
$L_\alpha $
can recursively construct an isomorphism from
$\prec $
to an ordinal, so if
$\eta _\beta <\alpha $
, then
$\eta _{\beta +1}\leq \alpha $
.
Lemma 2.2. Let
$\alpha $
be a recursively inaccessible ordinal. Then
$\alpha = \eta _\alpha $
.
Proof. This is immediate from the previous lemma.
Below, we say that an ordinal
$\gamma $
is
$\beta $
-r.e. if there is a
$\beta $
-r.e. wellordering of
$\beta $
of length
$\gamma $
.
Lemma 2.3. Let
$\alpha $
,
$\beta $
, and
$\gamma $
be ordinals. Suppose that
$\beta <\eta _\alpha $
and
$\gamma $
is
$\beta $
-r.e. Then,
$\gamma < \eta _{\alpha +1}$
.
Proof. Let
$\phi $
be the
$\Sigma _1$
formula which defines over
$L_\beta $
a wellordering of length
$\gamma $
. Then, the formula
is
$\Sigma _0$
(thus
$\Sigma _1$
) over
$L_{\eta _\alpha }$
with parameters in
$L_{\eta _\alpha }$
.
Lemma 2.4. Let
$\alpha $
be an ordinal. Then,
$\eta _\alpha $
is a limit ordinal and is closed under addition, multiplication, exponentiation with base
$\omega $
, the Veblen functions, etc.
Proof. This is proved by induction. The limit stages are immediate. Supposing
$\eta _\alpha $
satisfies those closure properties and
$\beta ,\gamma $
are
$\eta _\alpha $
-r.e., we can use the fact that
$\eta _\alpha $
is multiplicatively indecomposable (thus closed under coding of tuples) to combine the definitions of
$\beta ,\gamma $
into
$\eta _\alpha $
-r.e. wellorderings of length
$\beta +\gamma $
and
$\beta \cdot \gamma $
. Closure under the Veblen functions is a bit tricker and will not be used below, but it can be done by lifting the proof that
$\omega _1^{ck}$
is closed under the Veblen functions (see, e.g., [Reference Rathjen, Weiermann, Cooper and Sorbi19]).
Lemma 2.4 will be very useful and used without explicit mention in the future. Its main consequence is that
$L_{\eta _\alpha }$
is closed under the Cantor pairing function. Thus, we can replace mention of finite sequences of parameters in
$L_{\eta _\alpha }$
by single parameters which can additionally be assumed to be ordinals. Many consequences of admissibility will hold in
$L_{\eta _\alpha }$
, except possibly those that crucially depend on
$\Sigma _0$
-collection.
Lemma 2.5. Let
$\alpha $
be an ordinal. Then, every illfounded
$\eta _\alpha $
-r.e. linear order has an infinite descending chain which is
$\Pi _2$
-definable over
$L_{\eta _{\alpha +1}}$
.
Proof. Let
$\prec $
be such a linear order. Using the closure properties provided by Lemma 2.4, we may construct inside of
$L_{\eta _{\alpha +1}}$
an isomorphism
for some
$\beta _x$
, for each x in the wellfounded part of
$\prec $
. The construction of
$i_x$
is by transfinite recursion on
$\prec \upharpoonright x$
and requires
$\beta _x$
stages, so it belongs to
$L_{\eta _\alpha + \beta _x}$
. Since
$L_{\eta _{\alpha +1}}$
is additively indecomposable,
$i_x \in L_{\eta _{\alpha +1}}$
. The function f given by
$f(x) = \beta _x$
is partial
$\Delta _1$
over
$L_{\eta _\alpha +1}$
, with domain a subset of
$L_{\eta _\alpha }$
which need not be
$\Delta _1$
over
$L_{\eta _\alpha +1}$
. Consider the set
Thus, A is
$\Sigma _1$
-definable over
$L_{\eta _\alpha +1}$
. A is precisely the wellfounded part of
$\prec $
, for if
$x \in \text {wfp}(\prec )\setminus A$
, then
$\prec \upharpoonright x$
cannot be isomorphic to an ordinal smaller than
$\eta _{\alpha +1}$
(otherwise
$f(x)$
would be defined), so
$\prec \upharpoonright x$
is an
$\eta _{\alpha }$
-r.e. wellorder of length at least
$\eta _{\alpha +1}$
, which contradicts the definition. Using A we can replicate the proof of König’s lemma: by induction on
$i\in \mathbb {N}$
, define
$x_{i+1}$
to be the
$<_L$
-least
$x \not \in A$
such that
$x \prec x_i$
. This defines an infinite descending chain through
$\prec $
.
As for the complexity of this sequence s, we have
$x_i <\eta _\alpha $
for each i, and the restriction of
$<_L$
to
$L_{\eta _\alpha }$
is
$\Sigma _1$
over
$L_{\eta _\alpha }$
and hence an element of
$L_{\eta _\alpha +1}$
. Therefore, s is the unique
$\omega $
-sequence all of whose elements
$x_i$
have the following properties:
-
(1) $x_i \not \in A$
; -
(2) $x_{i+1} \prec x_{i}$
; -
(3) for all $y\in \eta _\alpha $
, if
$y \prec x_i$
and
$y\not \in A$
then
$x_{i+1} <_L y$
.
The first condition is
$\Pi _1$
over
$L_{\eta _{\alpha +1}}$
; the second is
$\Delta _0$
; and the third is
$\Pi _2$
, since
$x_{i+1}<_L y$
is equivalent to
$L_{\eta _\alpha }\models x_{i+1}<_L y$
(we cannot conclude that the third condition is
$\Sigma _1$
, because
$\eta _{\alpha +1}$
may not be admissible).
We remark the following consequence of the proof of the previous lemma:
Lemma 2.6. Let
$\alpha $
be an ordinal and let
$\prec $
be an illfounded
$\eta _\alpha $
-r.e. linear order. Then, either
$\prec $
has an infinite descending sequence in
$L_{\eta _\alpha +1}$
, or the wellfounded part of
$\prec $
is isomorphic to
$\eta _{\alpha +1}$
.
Proof. This follows from the preceding argument, using the observation that if the range of f is bounded below
$\eta _{\alpha +1}$
, say
$f[\eta _\alpha ] \subset \beta < \eta _{\alpha +1}.$
Then, A is
$\Sigma _1$
-definable over
$L_\beta $
.
Below, an ordinal
$\alpha $
is locally countable if every infinite
$\beta <\alpha $
has a bijection with
$\omega $
in
$L_\alpha $
.
Lemma 2.7. Let
$\alpha $
be an ordinal. Suppose
$\eta _\alpha $
is inadmissible. If
$\eta _\alpha $
is locally countable, then
$\eta _{\alpha +1}$
is admissible.
Proof. Suppose first that
$\eta _\alpha $
is a successor effective cardinal and let
$\beta $
be its predecessor. Since
$\eta _\alpha $
is locally countable, there is a real
$x \in L_{\eta _\alpha }$
coding a wellordering of
$\mathbb {N}$
of length
$\beta $
. Let
$\gamma $
be least such that
$x \in L_{\gamma +1}$
. Then, for each
$\xi $
,
Since
$\eta _\alpha $
is additively indecomposable, we have
The proof that
$\omega _1^{ck}$
is the least admissible ordinal (see [Reference Barwise7]) relativizes and shows that
$\omega _1^x$
, the supremum of x-recursive ordinals, is least such that
$L_{\omega _1^x}[x]$
is admissible. Thus,
Therefore, there are
$\eta _\alpha $
-r.e. wellorderings of lengths cofinal below
$\eta _\alpha ^+$
and the result follows.
The previous argument applies also to the case that
$\eta _\alpha $
is a limit effective cardinal, but there is a largest admissible smaller than
$\eta _\alpha $
; the remaining case is that
$\eta _\alpha $
is an inadmissible limit of admissibles. This case will not be necessary below, but it is good to note. To obtain it, follow the argument of Gostanian [Reference Gostanian9, Theorem 2.1] replacing
${\mathsf {KP}}$
by
${\mathsf {KPl}}$
throughout. This shows that if
$\alpha $
is a locally countable limit of admissibles, then the
$\alpha $
-recursive wellorders are cofinal below
$\alpha ^+$
unless
$\alpha $
is
$\Sigma ^1_1$
-reflecting. This is enough to yield the result, since
$\Sigma ^1_1$
-reflecting ordinals are all admissible.
3
$\Sigma ^0_4$
-determinacy
In this section we prove the main theorem.
Theorem 3.1. The following are equivalent over
$\mathsf {KP}$
:
-
(1) $\Sigma ^0_4$
-determinacy
$;$
-
(2) there is an ordinal which is not $\omega $
-Gandy.
To prove the corresponding result for
$\Sigma ^0_2$
-determinacy (Theorem 1.3), one first uses Solovay’s theorem on the complexity of winning strategies for
$\Sigma ^0_2$
games and positive
$\Sigma ^1_1$
induction, and then uses the results of Aczel–Richter, Gostanian, and Grilliot to relate this to 1-Gandy ordinals. The proof of Theorem 3.1 follows a similar outline: first we appeal to a theorem on the complexity of winning strategies for
$\Sigma ^0_4$
games, and then we relate that to
$\omega $
-Gandy ordinals. The result we will require is due to Hachtman and we state it now. Here, recall that the rank of a wellfounded relation B is defined inductively as the strict supremum of the ranks
$\rho (a)$
of its elements, where
$\rho (a) = \sup \{\rho (b) + 1: bBa\}$
.
Theorem 3.2 (Hachtman [Reference Hachtman10])
The least ordinal
$\theta $
such that every
$\Sigma ^0_4$
game has a winning strategy definable over
$L_\theta $
is the least ordinal such that
$L_\theta $
satisfies “
$\mathbb {R}$
exists and every wellfounded tree has a rank.”
Recall that by Gödel’s condensation lemma, a structure of the form
$L_\theta $
satisfies “
$\mathbb {R}$
exists” if and only if it is not locally countable. Thus, we need to prove:
Theorem 3.3. Assume
${\mathsf {KP}}$
. Then, the existence of an ordinal
$\zeta $
which is not
$\omega $
-Gandy is equivalent to the existence of an ordinal
$\theta $
such that
$\theta $
is not locally countable and
$L_{\theta } \models $
“every wellfounded tree has a rank.” Moreover, letting
$\zeta $
and
$\theta $
be the least such ordinals, then
$\zeta = \eta _\zeta $
and
$\theta = \eta _{\zeta +\omega }$
.
We prove this theorem in the remainder of the section. We would like to thank R. Lubarsky for bringing the following fact to our attention at some point in the past.
Lemma 3.4. Let
$\alpha $
be a multiplicatively indecomposable ordinal and let
$\rho _\alpha $
be the supremum of ranks of wellfounded
$\alpha $
-r.e. trees. Then
$\rho _\alpha $
is the supremum of lengths of
$\alpha $
-r.e. wellorderings of
$\alpha $
.
Proof. If T is a wellfounded
$\alpha $
-r.e. tree, then its Kleene–Brouwer linearization
$\prec _T$
is an
$\alpha $
-r.e. wellorder. A simple induction shows that the length of
$\prec _T$
is at least the rank of T. Conversely, suppose
$\prec $
is an
$\alpha $
-r.e. wellorder and let
$T_\prec $
be the tree of all descending chains through
$\prec $
ordered by end-extension. Then this is a wellfounded
$\alpha $
-r.e. tree and—as before—a simple induction shows that its rank is at least the length of
$\prec $
.
We will use the preceding lemma without mention, alternating between talking about trees and linear orders as we deem convenient. Lemma 3.5 below is already half of the proof of the theorem.
Lemma 3.5. Suppose
$\eta _\xi $
is not
$\omega $
-Gandy. Then, letting
$\gamma = \eta _{\xi + \omega } < \xi ^+$
,
$\gamma $
is not locally countable and
Proof. It follows from Lemma 2.7 that for all
$i \in \mathbb {N}$
with
$i \neq 0$
,
$L_{\eta _{\xi + i}}$
is not locally countable. This implies in particular that for each i,
so
$L_{\gamma }$
is also not locally countable. It remains to show that
If
$\prec $
is a linear ordering in
$L_\gamma $
, then
$\prec $
is
$\eta _{\xi +i}$
-r.e. for some i, by Lemma 2.3. Thus, there are two possibilities: either
$\prec $
is wellfounded (in the real world) and thus isomorphic to some ordinal
${<}\eta _{\xi +i+1}$
. If so, then such an isomorphism can be constructed within
$L_{\eta _{\xi +i+1}}$
and belongs to
$L_\gamma $
. Otherwise,
$\prec $
is illfounded (in the real world) and thus has an infinite descending chain in
$L_{\eta _{\xi +i}+1}$
(by Lemma 2.5) and thus in
$L_\gamma $
.
For the rest of this section, we will denote by
$\theta $
the least ordinal such that
$L_\theta $
satisfies “
$\mathbb {R}$
exists and every wellfounded tree has a rank,” and we write
Observe that
$\zeta = \eta _\zeta $
.
Lemma 3.6.
$\theta = \eta _{\zeta +\omega }$
. Moreover,
$\eta _{\zeta +i}$
is inadmissible for all
$i\in \mathbb {N}$
.
Proof. Clearly
$\eta _{\zeta +\omega }$
is inadmissible. We prove that for each i,
$\eta _{\zeta +i}<\theta $
and
$\eta _{\zeta +i}$
is inadmissible. This implies that
$\eta _\zeta $
is not
$\omega $
-Gandy and that
$\eta _{\zeta +\omega }\leq \theta $
, so that
$\eta _{\zeta +\omega } = \theta $
by Lemma 3.5 and the minimality of
$\theta $
.
Inductively, suppose that
$\eta _{\zeta +i}<\theta $
and that either
$i = 0$
or
$\eta _{\zeta +i}$
is inadmissible.
Claim 3.7. Let
$\prec $
be an illfounded
$\eta _{\zeta +i}$
-r.e. linear order. Then, there is an infinite descending chain through
$\prec $
definable over
$L_{\eta _{\zeta +i}}$
.
Proof. Suppose otherwise and let
$\beta \neq \eta _{\zeta +i}$
be least such that there is an infinite descending chain through
$\prec $
in
$L_{\beta +1}$
and let d be such a chain. By the definition of
$\theta $
,
$\beta <\theta $
. We have
as witnessed by a bijection
$f: \eta _{\zeta +i} \to \zeta $
. Using the hypothesis that each
$\eta _{\zeta + j}$
is inadmissible for
$0 < j \leq i$
, we see that there is no admissible ordinal between
$\zeta $
and
$\eta _{\zeta +i}$
, and thus such a bijection is definable over
$L_{\eta _{\zeta +i}}$
. Then,
$f[d]$
is an
$\omega $
-sequence of ordinals countable in
$L_{\theta }$
which belongs to
$L_{\beta +1}$
and not to
$L_{\eta _{\zeta +i}+1}$
. By Gödel’s condensation lemma,
$\zeta $
is no longer a cardinal in
$L_{\beta +1}$
, which is a contradiction.
Claim 3.8.
$\eta _{\zeta +i+1}$
is inadmissible.
Proof. This follows from the previous claim, since every
$\eta _{\zeta +i}$
-r.e. linear order can be mapped in a
$\Delta _1^{L_{\eta _{\zeta +i+1}}}$
way to an infinite descending chain or to its order-type, and these order-types are cofinal in
$\eta _{\zeta +i+1}$
.
It follows from the first claim that the set of indices of
$\eta _{\zeta +i}$
-r.e. wellorders is definable over
$L_{\eta _{\zeta +i} +1}$
. Thus, there is a linear order in
$L_{\eta _{\zeta +i} + 2}$
(and thus in
$L_\theta $
) which is longer than all
$\eta _{\zeta +i}$
-r.e. wellorders (namely, the sum of all
$\eta _{\zeta +i}$
-r.e. wellorders) and thus has length at least
$\eta _{\zeta +i+1}$
. Therefore,
$\eta _{\zeta +i+1} < \theta $
.
We have seen that
$\theta = \eta _{\zeta +\omega }$
is a supremum of inadmissible effective cardinals. Thus,
$\zeta $
is not
$\omega $
-Gandy. This completes the proof of the theorem.
4 Closing remarks
We have exhibited a characterization of
$\Sigma ^0_4$
-determinacy in the spirit of Reverse Mathematics which is analogous to the Aczel–Richter–Gostanian–Grilliot–Solovay–Tanaka characterization of
$\Sigma ^0_2$
-determinacy. Some questions are left open by this, however. Tanaka [Reference Tanaka23] characterized
$\Sigma ^0_2$
-determinacy in terms of a theory of monotone inductions. Is there an interesting characterization for
$\Sigma ^0_4$
-determinacy in terms of inductive definability? Is there a characterization for
$\Sigma ^0_3$
-determinacy or
$\Sigma ^0_5$
-determinacy in terms of Gandy ordinals?
We believe that the notion of an effective cardinal is a natural one and foreshadow that there is more to be said about them.
Funding
This work was partially supported by FWO grant 3E017319 and FWF grant I-4513.


