Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Topological complexity of ideal limit points

Marek Balcerzak Institute of Mathematics, Lodz University of Technology, al. Politechniki 8, 93-590 Lodz, Poland marek.balcerzak@p.lodz.pl Szymon Gł\kab Institute of Mathematics, Lodz University of Technology, al. Politechniki 8, 93-590 Lodz, Poland szymon.glab@p.lodz.pl  and  Paolo Leonetti Department of Economics, Università degli Studi dell’Insubria, via Monte Generoso 71, 21100 Varese, Italy leonetti.paolo@gmail.com
Abstract.

Given an ideal \mathcal{I}caligraphic_I on the nonnegative integers ω𝜔\omegaitalic_ω and a Polish space X𝑋Xitalic_X, let ()\mathscr{L}(\mathcal{I})script_L ( caligraphic_I ) be the family of subsets SX𝑆𝑋S\subseteq Xitalic_S ⊆ italic_X such that S𝑆Sitalic_S is the set of \mathcal{I}caligraphic_I-limit points of some sequence taking values in X𝑋Xitalic_X. First, we show that ()\mathscr{L}(\mathcal{I})script_L ( caligraphic_I ) may attain arbitrarily large Borel complexity. Second, we prove that if \mathcal{I}caligraphic_I is a Gδσsubscript𝐺𝛿𝜎G_{\delta\sigma}italic_G start_POSTSUBSCRIPT italic_δ italic_σ end_POSTSUBSCRIPT-ideal then all elements of ()\mathscr{L}(\mathcal{I})script_L ( caligraphic_I ) are closed. Third, we show that if \mathcal{I}caligraphic_I is a simply coanalytic ideal and X𝑋Xitalic_X is first countable, then every element of ()\mathscr{L}(\mathcal{I})script_L ( caligraphic_I ) is simply analytic. Lastly, we studied certain structural properties and the topological complexity of minimal ideals \mathcal{I}caligraphic_I for which ()\mathscr{L}(\mathcal{I})script_L ( caligraphic_I ) contains a given set.

Key words and phrases:
Ideal limit points; Borel and analytic ideals; simply analytic set; Fubini product; pruned trees; Wadge reduction.
2010 Mathematics Subject Classification:
Primary: 40A35, 54H05. Secondary: 11B05, 28A05, 54A20.

1. Introduction

Let \mathcal{I}caligraphic_I be an ideal on the nonnegative integers ω𝜔\omegaitalic_ω, that is, a subset of 𝒫(ω)𝒫𝜔\mathcal{P}(\omega)caligraphic_P ( italic_ω ) closed under taking subsets and finite unions. Unless otherwise stated, it is assumed that \mathcal{I}caligraphic_I is admissible, namely, ω𝜔\omega\notin\mathcal{I}italic_ω ∉ caligraphic_I and that \mathcal{I}caligraphic_I contains the family FinFin\mathrm{Fin}roman_Fin of finite subsets of ω𝜔\omegaitalic_ω. Intuitively, the ideal \mathcal{I}caligraphic_I represents the family of “small” subsets of ω𝜔\omegaitalic_ω. An important example is the family of asymptotic density zero sets

𝒵:={Sω:limn|S[0,n]|n=0}.assign𝒵conditional-set𝑆𝜔subscript𝑛𝑆0𝑛𝑛0\mathcal{Z}:=\left\{S\subseteq\omega:\lim_{n\to\infty}\frac{|S\cap[0,n]|}{n}=0% \right\}.caligraphic_Z := { italic_S ⊆ italic_ω : roman_lim start_POSTSUBSCRIPT italic_n → ∞ end_POSTSUBSCRIPT divide start_ARG | italic_S ∩ [ 0 , italic_n ] | end_ARG start_ARG italic_n end_ARG = 0 } .

Define +:=𝒫(ω)assignsuperscript𝒫𝜔\mathcal{I}^{+}:=\mathcal{P}(\omega)\setminus\mathcal{I}caligraphic_I start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT := caligraphic_P ( italic_ω ) ∖ caligraphic_I. Ideals are regarded as subsets of the Cantor space {0,1}ωsuperscript01𝜔\{0,1\}^{\omega}{ 0 , 1 } start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT, hence we can speak about their topological complexity. For instance, FinFin\mathrm{Fin}roman_Fin is a Fσsubscript𝐹𝜎F_{\sigma}italic_F start_POSTSUBSCRIPT italic_σ end_POSTSUBSCRIPT-ideal, and 𝒵𝒵\mathcal{Z}caligraphic_Z is a Fσδsubscript𝐹𝜎𝛿F_{\sigma\delta}italic_F start_POSTSUBSCRIPT italic_σ italic_δ end_POSTSUBSCRIPT-ideal which is not Fσsubscript𝐹𝜎F_{\sigma}italic_F start_POSTSUBSCRIPT italic_σ end_POSTSUBSCRIPT.

Pick also a sequence 𝒙=(xn)𝒙subscript𝑥𝑛\bm{x}=(x_{n})bold_italic_x = ( italic_x start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT ) be taking values in a topological space X𝑋Xitalic_X. Then, we denote by Λ𝒙()subscriptΛ𝒙\Lambda_{\bm{x}}(\mathcal{I})roman_Λ start_POSTSUBSCRIPT bold_italic_x end_POSTSUBSCRIPT ( caligraphic_I ) the set of \mathcal{I}caligraphic_I-limit points of 𝒙𝒙\bm{x}bold_italic_x, that is, the set of all ηX𝜂𝑋\eta\in Xitalic_η ∈ italic_X for which there exists a subsequence (xnk)subscript𝑥subscript𝑛𝑘(x_{n_{k}})( italic_x start_POSTSUBSCRIPT italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT end_POSTSUBSCRIPT ) such that

limkxnk=η and {nk:kω}+.formulae-sequencesubscript𝑘subscript𝑥subscript𝑛𝑘𝜂 and conditional-setsubscript𝑛𝑘𝑘𝜔superscript\lim_{k\to\infty}x_{n_{k}}=\eta\quad\text{ and }\quad\{n_{k}:k\in\omega\}\in% \mathcal{I}^{+}.roman_lim start_POSTSUBSCRIPT italic_k → ∞ end_POSTSUBSCRIPT italic_x start_POSTSUBSCRIPT italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT end_POSTSUBSCRIPT = italic_η and { italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT : italic_k ∈ italic_ω } ∈ caligraphic_I start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT .

It is well known that, even in the case where 𝒙𝒙{\bm{x}}bold_italic_x is a real bounded sequence, it is possible that Λ𝒙(𝒵)subscriptΛ𝒙𝒵\Lambda_{\bm{x}}(\mathcal{Z})roman_Λ start_POSTSUBSCRIPT bold_italic_x end_POSTSUBSCRIPT ( caligraphic_Z ) is the empty set, see [13, Example 4]. The topological nature of the sets of \mathcal{I}caligraphic_I-limits points Λ𝒙()subscriptΛ𝒙\Lambda_{\bm{x}}(\mathcal{I})roman_Λ start_POSTSUBSCRIPT bold_italic_x end_POSTSUBSCRIPT ( caligraphic_I ) and their relationship with the slightly weaker variant of \mathcal{I}caligraphic_I-cluster points have been studied in [2], cf. also [16, 17].

In this work, we continue along this line of research. To this aim, we introduce our main definition:

Definition 1.1.

Let X𝑋Xitalic_X be a topological space and \mathcal{I}caligraphic_I be an ideal on ω𝜔\omegaitalic_ω. We denote by X()subscript𝑋\mathscr{L}_{X}(\mathcal{I})script_L start_POSTSUBSCRIPT italic_X end_POSTSUBSCRIPT ( caligraphic_I ) the family of sets of \mathcal{I}caligraphic_I-limit points of sequences 𝒙𝒙{\bm{x}}bold_italic_x taking values in X𝑋Xitalic_X together with the emptyset, that is,

X():={AX:A=Λ𝒙() for some sequence 𝒙Xω}{}.assignsubscript𝑋conditional-set𝐴𝑋𝐴subscriptΛ𝒙 for some sequence 𝒙superscript𝑋𝜔\mathscr{L}_{X}(\mathcal{I}):=\left\{A\subseteq X:A=\Lambda_{\bm{x}}(\mathcal{% I})\text{ for some sequence }{\bm{x}}\in X^{\omega}\right\}\cup\{\emptyset\}.script_L start_POSTSUBSCRIPT italic_X end_POSTSUBSCRIPT ( caligraphic_I ) := { italic_A ⊆ italic_X : italic_A = roman_Λ start_POSTSUBSCRIPT bold_italic_x end_POSTSUBSCRIPT ( caligraphic_I ) for some sequence bold_italic_x ∈ italic_X start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT } ∪ { ∅ } .

If the topological space X𝑋Xitalic_X is understood, we write simply ()\mathscr{L}(\mathcal{I})script_L ( caligraphic_I ).

A remark is in order about the the addition of {}\{\emptyset\}{ ∅ } in the above definition: it has been proved by Meza-Alcántara in [28, Section 2.7] that, if X=[0,1]𝑋01X=[0,1]italic_X = [ 0 , 1 ], then there exists a [0,1]01[0,1][ 0 , 1 ]-valued sequence 𝒙𝒙\bm{x}bold_italic_x such that Λ𝒙()=subscriptΛ𝒙\Lambda_{\bm{x}}(\mathcal{I})=\emptysetroman_Λ start_POSTSUBSCRIPT bold_italic_x end_POSTSUBSCRIPT ( caligraphic_I ) = ∅ if and only if there exists a function ϕ:ω[0,1]:italic-ϕ𝜔01\phi:\omega\to\mathbb{Q}\cap[0,1]italic_ϕ : italic_ω → blackboard_Q ∩ [ 0 , 1 ] such that ϕ1[A]superscriptitalic-ϕ1delimited-[]𝐴\phi^{-1}[A]\in\mathcal{I}italic_ϕ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_A ] ∈ caligraphic_I for every set A[0,1]𝐴01A\subseteq\mathbb{Q}\cap[0,1]italic_A ⊆ blackboard_Q ∩ [ 0 , 1 ] with at most finitely many limit points; cf. also [3, Proposition 6.4] and, more generally, [10] for analogues in compact uncountable spaces. On the other hand, if X𝑋Xitalic_X is not compact, it is easy to see that, for every ideal \mathcal{I}caligraphic_I, there is a sequence with no \mathcal{I}caligraphic_I-limit points. Thus, the addition of {}\{\emptyset\}{ ∅ } in the above definition avoids the repetition of known results in the literature and to add further subcases based on the topological structure of the underlying space X𝑋Xitalic_X.

We summarize in Theorem 1.2 and Theorem 1.3 below the known results from [2, 16, 17] about the families X()subscript𝑋\mathscr{L}_{X}(\mathcal{I})script_L start_POSTSUBSCRIPT italic_X end_POSTSUBSCRIPT ( caligraphic_I ).

For, given a topological space X𝑋Xitalic_X and an ordinal 1α<ω11𝛼subscript𝜔11\leq\alpha<\omega_{1}1 ≤ italic_α < italic_ω start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT, we use the standard Borel pointclasses notations Σα0(X)subscriptsuperscriptΣ0𝛼𝑋\Sigma^{0}_{\alpha}(X)roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT italic_α end_POSTSUBSCRIPT ( italic_X ) and Πα0(X)subscriptsuperscriptΠ0𝛼𝑋\Pi^{0}_{\alpha}(X)roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT italic_α end_POSTSUBSCRIPT ( italic_X ), so that Σ10(X)subscriptsuperscriptΣ01𝑋\Sigma^{0}_{1}(X)roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ( italic_X ) stands for the open sets of X𝑋Xitalic_X, Π10(X)subscriptsuperscriptΠ01𝑋\Pi^{0}_{1}(X)roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ( italic_X ) for the closed sets, Σ20(X)superscriptsubscriptΣ20𝑋\Sigma_{2}^{0}(X)roman_Σ start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT ( italic_X ) for the Fσsubscript𝐹𝜎F_{\sigma}italic_F start_POSTSUBSCRIPT italic_σ end_POSTSUBSCRIPT-sets, etc.; we denote by Δα0(X):=Σα0(X)Πα0(X)assignsubscriptsuperscriptΔ0𝛼𝑋subscriptsuperscriptΣ0𝛼𝑋subscriptsuperscriptΠ0𝛼𝑋\Delta^{0}_{\alpha}(X):=\Sigma^{0}_{\alpha}(X)\cap\Pi^{0}_{\alpha}(X)roman_Δ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT italic_α end_POSTSUBSCRIPT ( italic_X ) := roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT italic_α end_POSTSUBSCRIPT ( italic_X ) ∩ roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT italic_α end_POSTSUBSCRIPT ( italic_X ) the ambiguous classes; also, if X𝑋Xitalic_X is a Polish space, Σ11(X)subscriptsuperscriptΣ11𝑋\Sigma^{1}_{1}(X)roman_Σ start_POSTSUPERSCRIPT 1 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ( italic_X ) stands for the analytic sets, Π11(X)subscriptsuperscriptΠ11𝑋\Pi^{1}_{1}(X)roman_Π start_POSTSUPERSCRIPT 1 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ( italic_X ) for the coanalytic sets, etc., see e.g. [25, Section 11.B] or [31, Section 3.6]. Again, we suppress the reference to the underlying space X𝑋Xitalic_X if it is clear from the context.

Recall that an ideal \mathcal{I}caligraphic_I is a P-ideal if it is σ𝜎\sigmaitalic_σ-directed modulo finite sets, that is, for every sequence (An)subscript𝐴𝑛(A_{n})( italic_A start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT ) with values in \mathcal{I}caligraphic_I there exists A𝐴A\in\mathcal{I}italic_A ∈ caligraphic_I such that AnAsubscript𝐴𝑛𝐴A_{n}\setminus Aitalic_A start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT ∖ italic_A is finite for all nω𝑛𝜔n\in\omegaitalic_n ∈ italic_ω. Important examples include the Σ11subscriptsuperscriptΣ11\Sigma^{1}_{1}roman_Σ start_POSTSUPERSCRIPT 1 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT P-ideals, which are known to be necessarily Π30subscriptsuperscriptΠ03\Pi^{0}_{3}roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT. A topological space X𝑋Xitalic_X is discrete if it contains only isolated points.

Theorem 1.2.

Let X𝑋Xitalic_X be a nondiscrete first countable Hausdorff space. Then:

  1. (i)

    ()Π10subscriptsuperscriptΠ01\mathscr{L}(\mathcal{I})\subseteq\Pi^{0}_{1}script_L ( caligraphic_I ) ⊆ roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT if and only if \mathcal{I}caligraphic_I is a Σ20superscriptsubscriptΣ20\Sigma_{2}^{0}roman_Σ start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT ideal, provided that \mathcal{I}caligraphic_I is a Σ11subscriptsuperscriptΣ11\Sigma^{1}_{1}roman_Σ start_POSTSUPERSCRIPT 1 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT P-ideal.

  2. (ii)

    ()Σ20subscriptsuperscriptΣ02\mathscr{L}(\mathcal{I})\subseteq\Sigma^{0}_{2}script_L ( caligraphic_I ) ⊆ roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT, provided that \mathcal{I}caligraphic_I is a Σ11subscriptsuperscriptΣ11\Sigma^{1}_{1}roman_Σ start_POSTSUPERSCRIPT 1 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT P-ideal.

If, in addition, all closed subsets of X𝑋Xitalic_X are separable, then:

  1. (iii)

    ()=Π10superscriptsubscriptΠ10\mathscr{L}(\mathcal{I})=\Pi_{1}^{0}script_L ( caligraphic_I ) = roman_Π start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT, provided that \mathcal{I}caligraphic_I is a Σ20superscriptsubscriptΣ20\Sigma_{2}^{0}roman_Σ start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT ideal.

  2. (iv)

    ()=Σ20superscriptsubscriptΣ20\mathscr{L}(\mathcal{I})=\Sigma_{2}^{0}script_L ( caligraphic_I ) = roman_Σ start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT, provided that \mathcal{I}caligraphic_I is Σ11superscriptsubscriptΣ11\Sigma_{1}^{1}roman_Σ start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 1 end_POSTSUPERSCRIPT P-ideal which is not Σ20superscriptsubscriptΣ20\Sigma_{2}^{0}roman_Σ start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT.

Proof.

It follows by [2, Theorems 2.2, 2.5, 2.7, and 3.4]. ∎

To state the next result, we recall some further definitions. An ideal \mathcal{I}caligraphic_I has the hereditary Baire property if the restriction A:={SA:S}𝐴assignconditional-set𝑆𝐴𝑆\mathcal{I}\upharpoonright A:=\{S\cap A:S\in\mathcal{I}\}caligraphic_I ↾ italic_A := { italic_S ∩ italic_A : italic_S ∈ caligraphic_I } has the Baire property for every A+𝐴superscriptA\in\mathcal{I}^{+}italic_A ∈ caligraphic_I start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT. Note that all analytic ideals have the hereditary Baire property: indeed, the proof goes verbatim as in [24, Theorem 3.13], considering that analytic sets are closed under continuous preimages, and that they have the Baire property, see e.g. [31, Proposition 4.1.2 and Theorem 4.3.2]. In addition, there exist ideals with the Baire property but without the hereditary Baire property, see e.g. [32, Proposition 2.1].

Also, an ideal \mathcal{I}caligraphic_I on ω𝜔\omegaitalic_ω is said to be a P+superscript𝑃P^{+}italic_P start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT-ideal if, for every decreasing sequence (An)subscript𝐴𝑛(A_{n})( italic_A start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT ) with values in +superscript\mathcal{I}^{+}caligraphic_I start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT, there exists A+𝐴superscriptA\in\mathcal{I}^{+}italic_A ∈ caligraphic_I start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT such that AAn𝐴subscript𝐴𝑛A\setminus A_{n}italic_A ∖ italic_A start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT is finite for all nω𝑛𝜔n\in\omegaitalic_n ∈ italic_ω. It is known that all Fσsubscript𝐹𝜎F_{\sigma}italic_F start_POSTSUBSCRIPT italic_σ end_POSTSUBSCRIPT-ideals are P+superscript𝑃P^{+}italic_P start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT-ideals, see [23] and [22, Observation 2.2]. We remark that P+superscript𝑃P^{+}italic_P start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT-ideal may have arbitrarily high Borel complexity, as it has been proved in [21, p. 2031] and [22, Example 2.6], cf. also [14].

Moreover, an ideal \mathcal{I}caligraphic_I is called a Farah ideal if there exists a sequence (Kn)subscript𝐾𝑛(K_{n})( italic_K start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT ) of hereditary compact subsets in 𝒫(ω)𝒫𝜔\mathcal{P}(\omega)caligraphic_P ( italic_ω ) such that S𝑆S\in\mathcal{I}italic_S ∈ caligraphic_I if and only if for all nω𝑛𝜔n\in\omegaitalic_n ∈ italic_ω there exists kω𝑘𝜔k\in\omegaitalic_k ∈ italic_ω such that S[0,k]Kn𝑆0𝑘subscript𝐾𝑛S\setminus[0,k]\in K_{n}italic_S ∖ [ 0 , italic_k ] ∈ italic_K start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT, see [8, 19, 20]. It is known that all analytic P-ideals are Farah and that all Farah ideals are Fσδsubscript𝐹𝜎𝛿F_{\sigma\delta}italic_F start_POSTSUBSCRIPT italic_σ italic_δ end_POSTSUBSCRIPT. On the other hand, it is still unknown whether the converse holds, namely, all Fσδsubscript𝐹𝜎𝛿F_{\sigma\delta}italic_F start_POSTSUBSCRIPT italic_σ italic_δ end_POSTSUBSCRIPT ideals are Farah ideals, see [7] and [19, p. 60].

Theorem 1.3.

Let X𝑋Xitalic_X be a first countable Hausdorff space. Then:

  1. (i)

    ()Π10subscriptsuperscriptΠ01\mathscr{L}(\mathcal{I})\subseteq\Pi^{0}_{1}script_L ( caligraphic_I ) ⊆ roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT, provided that \mathcal{I}caligraphic_I is P+superscript𝑃P^{+}italic_P start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT-ideal.

  2. (ii)

    ()Π10subscriptsuperscriptΠ01\mathscr{L}(\mathcal{I})\subseteq\Pi^{0}_{1}script_L ( caligraphic_I ) ⊆ roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT if and only if \mathcal{I}caligraphic_I is a P+superscript𝑃P^{+}italic_P start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT-ideal, provided that \mathcal{I}caligraphic_I has the hereditary Baire property and X𝑋Xitalic_X is nondiscrete metrizable.

  3. (iii)

    ()Σ20subscriptsuperscriptΣ02\mathscr{L}(\mathcal{I})\subseteq\Sigma^{0}_{2}script_L ( caligraphic_I ) ⊆ roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT, provided that \mathcal{I}caligraphic_I is a Farah ideal.

  4. (iv)

    ()={{η}:ηX}conditional-set𝜂𝜂𝑋\mathscr{L}(\mathcal{I})=\{\{\eta\}:\eta\in X\}script_L ( caligraphic_I ) = { { italic_η } : italic_η ∈ italic_X } if and only if \mathcal{I}caligraphic_I is maximal, provided that |X|2.𝑋2.|X|\geq 2\textup{.}| italic_X | ≥ 2 .

If, in addition, X𝑋Xitalic_X is second countable, then:

  1. (v)

    Π10()subscriptsuperscriptΠ01\Pi^{0}_{1}\subseteq\mathscr{L}(\mathcal{I})roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ⊆ script_L ( caligraphic_I ), provided that \mathcal{I}caligraphic_I has the hereditary Baire property.

  2. (vi)

    Σ20()subscriptsuperscriptΣ02\Sigma^{0}_{2}\subseteq\mathscr{L}(\mathcal{I})roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ⊆ script_L ( caligraphic_I ), provided that \mathcal{I}caligraphic_I has the hereditary Baire property and is not a P+superscript𝑃P^{+}italic_P start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT-ideal.

Proof.

It follows by [16, Proposition 1.4, Corollary 3.9, and Theorem 4.5] and [17, Corollary 2.5 and Theorems 2.4, 2.8, and 2.10]. (We added the only if part of item (iv), which is straighforward.)

In what follows, we divide our main results into four sections. First, we show in Section 2 that ()\mathscr{L}(\mathcal{I})script_L ( caligraphic_I ) can be equal to families of arbitrarily high Borel complexity. Also, we prove that we cannot have the equality ()=Π20subscriptsuperscriptΠ02\mathscr{L}(\mathcal{I})=\Pi^{0}_{2}script_L ( caligraphic_I ) = roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT: more precisely, ()Π20subscriptsuperscriptΠ02\mathscr{L}(\mathcal{I})\subseteq\Pi^{0}_{2}script_L ( caligraphic_I ) ⊆ roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT if and only if ()Π10subscriptsuperscriptΠ01\mathscr{L}(\mathcal{I})\subseteq\Pi^{0}_{1}script_L ( caligraphic_I ) ⊆ roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT. Then, we show that, if X𝑋Xitalic_X is a first countable space and \mathcal{I}caligraphic_I is a simply coanalytic ideal (see Section 3 for details), then every Λ𝒙()subscriptΛ𝒙\Lambda_{\bm{x}}(\mathcal{I})roman_Λ start_POSTSUBSCRIPT bold_italic_x end_POSTSUBSCRIPT ( caligraphic_I ) is simply analytic. Lastly, we study structural and topological properties of “smallest”  ideals \mathcal{I}caligraphic_I such that ()\mathscr{L}(\mathcal{I})script_L ( caligraphic_I ) contains a given subsets of X𝑋Xitalic_X.

2. Large and small Borel complexities

Our first main result computes explicitly some families ()\mathscr{L}(\mathcal{I})script_L ( caligraphic_I ), proving that they may attain arbitrarily large Borel complexity (Theorem 2.5 below). This is somehow related to [16, Question 3.11], which asks about the existence of a Borel ideal \mathcal{I}caligraphic_I such that ()\mathscr{L}(\mathcal{I})script_L ( caligraphic_I ) contains sets with large Borel complexities.

For, given (possibly nonadmissible) ideals \mathcal{I}caligraphic_I and 𝒥𝒥\mathcal{J}caligraphic_J on two countably infinite sets Z𝑍Zitalic_Z and W𝑊Witalic_W, respectively, we define their Fubini product by

×𝒥:={SZ×W:{nZ:{kW:(n,k)S}𝒥}}},\mathcal{I}\times\mathcal{J}:=\{S\subseteq Z\times W\colon\{n\in Z:\{k\in W:(n% ,k)\in S\}\notin\mathcal{J}\}\in\mathcal{I}\}\},caligraphic_I × caligraphic_J := { italic_S ⊆ italic_Z × italic_W : { italic_n ∈ italic_Z : { italic_k ∈ italic_W : ( italic_n , italic_k ) ∈ italic_S } ∉ caligraphic_J } ∈ caligraphic_I } } ,

which is an ideal on the countably infinite set Z×W𝑍𝑊Z\times Witalic_Z × italic_W, see e.g. [7, Chapter 1]. Hence, recursively, Finα:=Fin×Finα1assignsuperscriptFin𝛼FinsuperscriptFin𝛼1\mathrm{Fin}^{\alpha}:=\mathrm{Fin}\times\mathrm{Fin}^{\alpha-1}roman_Fin start_POSTSUPERSCRIPT italic_α end_POSTSUPERSCRIPT := roman_Fin × roman_Fin start_POSTSUPERSCRIPT italic_α - 1 end_POSTSUPERSCRIPT for all integers α2𝛼2\alpha\geq 2italic_α ≥ 2 is an ideal on ωαsuperscript𝜔𝛼\omega^{\alpha}italic_ω start_POSTSUPERSCRIPT italic_α end_POSTSUPERSCRIPT.

Theorem 2.1.

Let ,𝒥𝒥\mathcal{I},\mathcal{J}caligraphic_I , caligraphic_J be (possibly nonadmissible) ideals on ω𝜔\omegaitalic_ω such that ×𝒥𝒥\mathcal{I}\times\mathcal{J}caligraphic_I × caligraphic_J is an admissible ideal on ω2superscript𝜔2\omega^{2}italic_ω start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT. Fix also a bijection h:ω2ω:superscript𝜔2𝜔h:\omega^{2}\to\omegaitalic_h : italic_ω start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT → italic_ω and let 𝐱=(xn)𝐱subscript𝑥𝑛\bm{x}=(x_{n})bold_italic_x = ( italic_x start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT ) be a sequence with values in a first countable space X𝑋Xitalic_X. For each nω𝑛𝜔n\in\omegaitalic_n ∈ italic_ω, define the sequence 𝐱(n)=(xk(n):kω)\bm{x}^{(n)}=(x^{(n)}_{k}:k\in\omega)bold_italic_x start_POSTSUPERSCRIPT ( italic_n ) end_POSTSUPERSCRIPT = ( italic_x start_POSTSUPERSCRIPT ( italic_n ) end_POSTSUPERSCRIPT start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT : italic_k ∈ italic_ω ) by xk(n):=xh(n,k)assignsubscriptsuperscript𝑥𝑛𝑘subscript𝑥𝑛𝑘x^{(n)}_{k}:=x_{h(n,k)}italic_x start_POSTSUPERSCRIPT ( italic_n ) end_POSTSUPERSCRIPT start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT := italic_x start_POSTSUBSCRIPT italic_h ( italic_n , italic_k ) end_POSTSUBSCRIPT for all kω𝑘𝜔k\in\omegaitalic_k ∈ italic_ω. Then

Λ𝒙(h[×𝒥])={ηX:{nω:ηΛ𝒙(n)(𝒥)}},subscriptΛ𝒙delimited-[]𝒥conditional-set𝜂𝑋conditional-set𝑛𝜔𝜂subscriptΛsuperscript𝒙𝑛𝒥\Lambda_{\bm{x}}(h[\,\mathcal{I}\times\mathcal{J}])=\left\{\eta\in X:\left\{n% \in\omega:\eta\in\Lambda_{\bm{x}^{(n)}}(\mathcal{J})\right\}\notin\mathcal{I}% \right\},roman_Λ start_POSTSUBSCRIPT bold_italic_x end_POSTSUBSCRIPT ( italic_h [ caligraphic_I × caligraphic_J ] ) = { italic_η ∈ italic_X : { italic_n ∈ italic_ω : italic_η ∈ roman_Λ start_POSTSUBSCRIPT bold_italic_x start_POSTSUPERSCRIPT ( italic_n ) end_POSTSUPERSCRIPT end_POSTSUBSCRIPT ( caligraphic_J ) } ∉ caligraphic_I } , (1)

where h[×𝒥]delimited-[]𝒥h[\,\mathcal{I}\times\mathcal{J}]italic_h [ caligraphic_I × caligraphic_J ] stands for the family {h[S]:S×𝒥}conditional-setdelimited-[]𝑆𝑆𝒥\{h[S]:S\in\mathcal{I}\times\mathcal{J}\}{ italic_h [ italic_S ] : italic_S ∈ caligraphic_I × caligraphic_J }.

Proof.

Let A𝐴Aitalic_A and B𝐵Bitalic_B be the left and right hand side of (1), respectively.

Inclusion AB𝐴𝐵A\subseteq Bitalic_A ⊆ italic_B. The inclusion is clear if A=𝐴A=\emptysetitalic_A = ∅. Otherwise fix a point ηA𝜂𝐴\eta\in Aitalic_η ∈ italic_A. Hence there exists a set Sω2𝑆superscript𝜔2S\subseteq\omega^{2}italic_S ⊆ italic_ω start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT such that S×𝒥𝑆𝒥S\notin\mathcal{I}\times\mathcal{J}italic_S ∉ caligraphic_I × caligraphic_J and the subsequence (xh(s):sS):subscript𝑥𝑠𝑠𝑆(x_{h(s)}:s\in S)( italic_x start_POSTSUBSCRIPT italic_h ( italic_s ) end_POSTSUBSCRIPT : italic_s ∈ italic_S ) is convergent to η𝜂\etaitalic_η. By the definition of Fubini product ×𝒥𝒥\mathcal{I}\times\mathcal{J}caligraphic_I × caligraphic_J,

N:={nω:Kn𝒥+}+, where Kn:={kω:(n,k)S}.formulae-sequenceassign𝑁conditional-set𝑛𝜔subscript𝐾𝑛superscript𝒥superscriptassign where subscript𝐾𝑛conditional-set𝑘𝜔𝑛𝑘𝑆N:=\{n\in\omega:K_{n}\in\mathcal{J}^{+}\}\in\mathcal{I}^{+},\,\,\text{ where }% \,\,K_{n}:=\{k\in\omega:(n,k)\in S\}.italic_N := { italic_n ∈ italic_ω : italic_K start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT ∈ caligraphic_J start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT } ∈ caligraphic_I start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT , where italic_K start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT := { italic_k ∈ italic_ω : ( italic_n , italic_k ) ∈ italic_S } .

At this point, for each nN𝑛𝑁n\in Nitalic_n ∈ italic_N, the subsequence (xk(n):kKn):subscriptsuperscript𝑥𝑛𝑘𝑘subscript𝐾𝑛(x^{(n)}_{k}:k\in K_{n})( italic_x start_POSTSUPERSCRIPT ( italic_n ) end_POSTSUPERSCRIPT start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT : italic_k ∈ italic_K start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT ) is convergent to η𝜂\etaitalic_η and Kn𝒥+subscript𝐾𝑛superscript𝒥K_{n}\in\mathcal{J}^{+}italic_K start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT ∈ caligraphic_J start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT. Since N+𝑁superscriptN\in\mathcal{I}^{+}italic_N ∈ caligraphic_I start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT, we obtain that ηB𝜂𝐵\eta\in Bitalic_η ∈ italic_B.

Inclusion BA𝐵𝐴B\subseteq Aitalic_B ⊆ italic_A. The inclusion is clear if B=𝐵B=\emptysetitalic_B = ∅. Otherwise fix a point ηB𝜂𝐵\eta\in Bitalic_η ∈ italic_B and let (Un)subscript𝑈𝑛(U_{n})( italic_U start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT ) be a decreasing local base of neighborhoods at η𝜂\etaitalic_η. Hence there exists N+𝑁superscriptN\in\mathcal{I}^{+}italic_N ∈ caligraphic_I start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT such that η𝜂\etaitalic_η is a 𝒥𝒥\mathcal{J}caligraphic_J-limit point of 𝒙(n)superscript𝒙𝑛\bm{x}^{(n)}bold_italic_x start_POSTSUPERSCRIPT ( italic_n ) end_POSTSUPERSCRIPT for each nN𝑛𝑁n\in Nitalic_n ∈ italic_N, let us say limkKnxk(n)=ηsubscript𝑘subscript𝐾𝑛subscriptsuperscript𝑥𝑛𝑘𝜂\lim_{k\in K_{n}}x^{(n)}_{k}=\etaroman_lim start_POSTSUBSCRIPT italic_k ∈ italic_K start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT end_POSTSUBSCRIPT italic_x start_POSTSUPERSCRIPT ( italic_n ) end_POSTSUPERSCRIPT start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT = italic_η for some Kn𝒥+subscript𝐾𝑛superscript𝒥K_{n}\in\mathcal{J}^{+}italic_K start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT ∈ caligraphic_J start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT. Upon removing finitely many elements, we can suppose without loss of generality that

nN,kKn,xk(n)Un.formulae-sequencefor-all𝑛𝑁formulae-sequencefor-all𝑘subscript𝐾𝑛subscriptsuperscript𝑥𝑛𝑘subscript𝑈𝑛\forall n\in N,\forall k\in K_{n},\quad x^{(n)}_{k}\in U_{n}.∀ italic_n ∈ italic_N , ∀ italic_k ∈ italic_K start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT , italic_x start_POSTSUPERSCRIPT ( italic_n ) end_POSTSUPERSCRIPT start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ∈ italic_U start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT .

Now, set S:={(n,k)ω2:nN,kKn}assign𝑆conditional-set𝑛𝑘superscript𝜔2formulae-sequence𝑛𝑁𝑘subscript𝐾𝑛S:=\{(n,k)\in\omega^{2}:n\in N,k\in K_{n}\}italic_S := { ( italic_n , italic_k ) ∈ italic_ω start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT : italic_n ∈ italic_N , italic_k ∈ italic_K start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT } and note that S×𝒥𝑆𝒥S\notin\mathcal{I}\times\mathcal{J}italic_S ∉ caligraphic_I × caligraphic_J. It follows that the subsequence (xh(s):sS):subscript𝑥𝑠𝑠𝑆(x_{h(s)}:s\in S)( italic_x start_POSTSUBSCRIPT italic_h ( italic_s ) end_POSTSUBSCRIPT : italic_s ∈ italic_S ) is convergent to η𝜂\etaitalic_η: indeed there are only finitely many elements of the subsequence outside each Unsubscript𝑈𝑛U_{n}italic_U start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT. Therefore ηA𝜂𝐴\eta\in Aitalic_η ∈ italic_A, which concludes the proof. ∎

Remark 2.2.

The above result holds for every topological space X𝑋Xitalic_X if ={}\mathcal{I}=\{\emptyset\}caligraphic_I = { ∅ }. Indeed, in the second part of the proof it is enough to let N𝑁Nitalic_N be a singleton.

It is worth noting that if \mathcal{I}caligraphic_I is an ideal on a countably infinite set Z𝑍Zitalic_Z, ϕ:Zω:italic-ϕ𝑍𝜔\phi:Z\to\omegaitalic_ϕ : italic_Z → italic_ω is a bijection, and 𝒙XZ𝒙superscript𝑋𝑍\bm{x}\in X^{Z}bold_italic_x ∈ italic_X start_POSTSUPERSCRIPT italic_Z end_POSTSUPERSCRIPT is a Z𝑍Zitalic_Z-indexed sequence with values in a topological space X𝑋Xitalic_X, then ϕ[]:={ϕ[S]:S}assignitalic-ϕdelimited-[]conditional-setitalic-ϕdelimited-[]𝑆𝑆\phi[\,\mathcal{I}]:=\{\phi[S]:S\in\mathcal{I}\}italic_ϕ [ caligraphic_I ] := { italic_ϕ [ italic_S ] : italic_S ∈ caligraphic_I } is an ideal on ω𝜔\omegaitalic_ω and Λ𝒙()=Λ𝒚(ϕ[]),subscriptΛ𝒙subscriptΛ𝒚italic-ϕdelimited-[]\Lambda_{\bm{x}}(\mathcal{I})=\Lambda_{\bm{y}}(\phi[\,\mathcal{I}]),roman_Λ start_POSTSUBSCRIPT bold_italic_x end_POSTSUBSCRIPT ( caligraphic_I ) = roman_Λ start_POSTSUBSCRIPT bold_italic_y end_POSTSUBSCRIPT ( italic_ϕ [ caligraphic_I ] ) , where 𝒚Xω𝒚superscript𝑋𝜔\bm{y}\in X^{\omega}bold_italic_y ∈ italic_X start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT is the sequence defined by yn:=xϕ1(n)assignsubscript𝑦𝑛subscript𝑥superscriptitalic-ϕ1𝑛y_{n}:=x_{\phi^{-1}(n)}italic_y start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT := italic_x start_POSTSUBSCRIPT italic_ϕ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT ( italic_n ) end_POSTSUBSCRIPT for all nω𝑛𝜔n\in\omegaitalic_n ∈ italic_ω. Hence we may use interchangeably ()\mathscr{L}(\mathcal{I})script_L ( caligraphic_I ) or (ϕ[])italic-ϕdelimited-[]\mathscr{L}(\phi[\,\mathcal{I}])script_L ( italic_ϕ [ caligraphic_I ] ). In particular, Theorem 2.1 allows to compute explicitly families of the type (×𝒥)𝒥\mathscr{L}(\mathcal{I}\times\mathcal{J})script_L ( caligraphic_I × caligraphic_J ). For, we state two consequence of Theorem 2.1:

Corollary 2.3.

Let X𝑋Xitalic_X be a topological space and \mathcal{I}caligraphic_I be an ideal on ω𝜔\omegaitalic_ω. Then

(×)={nAn:A0,A1,()}.conditional-setsubscript𝑛subscript𝐴𝑛subscript𝐴0subscript𝐴1\mathscr{L}(\emptyset\times\mathcal{I})=\left\{\bigcup\nolimits_{n}A_{n}:A_{0}% ,A_{1},\ldots\in\mathscr{L}(\mathcal{I})\right\}.script_L ( ∅ × caligraphic_I ) = { ⋃ start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT italic_A start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT : italic_A start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , italic_A start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , … ∈ script_L ( caligraphic_I ) } .
Proof.

It follows by Theorem 2.1 and Remark 2.2 that

(×)={Λ𝒙(×):𝒙Xω}={nΛ𝒙(n)():𝒙Xω}={nΛ𝒙(n)():𝒙(0),𝒙(1),Xω}={nAn:A0,A1,()},conditional-setsubscriptΛ𝒙𝒙superscript𝑋𝜔conditional-setsubscript𝑛subscriptΛsuperscript𝒙𝑛𝒙superscript𝑋𝜔conditional-setsubscript𝑛subscriptΛsuperscript𝒙𝑛superscript𝒙0superscript𝒙1superscript𝑋𝜔conditional-setsubscript𝑛subscript𝐴𝑛subscript𝐴0subscript𝐴1\begin{split}\mathscr{L}(\emptyset\times\mathcal{I})&=\left\{\Lambda_{\bm{x}}(% \emptyset\times\mathcal{I}):{\bm{x}}\in X^{\omega}\right\}\\ &=\left\{\bigcup\nolimits_{n}\Lambda_{{\bm{x}}^{(n)}}(\mathcal{I}):{\bm{x}}\in X% ^{\omega}\right\}\\ &=\left\{\bigcup\nolimits_{n}\Lambda_{{\bm{x}}^{(n)}}(\mathcal{I}):{\bm{x}}^{(% 0)},{\bm{x}}^{(1)},\ldots\in X^{\omega}\right\}\\ &=\left\{\bigcup\nolimits_{n}A_{n}:A_{0},A_{1},\ldots\in\mathscr{L}(\mathcal{I% })\right\},\end{split}start_ROW start_CELL script_L ( ∅ × caligraphic_I ) end_CELL start_CELL = { roman_Λ start_POSTSUBSCRIPT bold_italic_x end_POSTSUBSCRIPT ( ∅ × caligraphic_I ) : bold_italic_x ∈ italic_X start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT } end_CELL end_ROW start_ROW start_CELL end_CELL start_CELL = { ⋃ start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT roman_Λ start_POSTSUBSCRIPT bold_italic_x start_POSTSUPERSCRIPT ( italic_n ) end_POSTSUPERSCRIPT end_POSTSUBSCRIPT ( caligraphic_I ) : bold_italic_x ∈ italic_X start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT } end_CELL end_ROW start_ROW start_CELL end_CELL start_CELL = { ⋃ start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT roman_Λ start_POSTSUBSCRIPT bold_italic_x start_POSTSUPERSCRIPT ( italic_n ) end_POSTSUPERSCRIPT end_POSTSUBSCRIPT ( caligraphic_I ) : bold_italic_x start_POSTSUPERSCRIPT ( 0 ) end_POSTSUPERSCRIPT , bold_italic_x start_POSTSUPERSCRIPT ( 1 ) end_POSTSUPERSCRIPT , … ∈ italic_X start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT } end_CELL end_ROW start_ROW start_CELL end_CELL start_CELL = { ⋃ start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT italic_A start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT : italic_A start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , italic_A start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , … ∈ script_L ( caligraphic_I ) } , end_CELL end_ROW

completing the proof. ∎

Corollary 2.4.

Let X𝑋Xitalic_X be a first countable space and \mathcal{I}caligraphic_I be an ideal on ω𝜔\omegaitalic_ω. Then

(Fin×)={lim supnAn:A0,A1,()}.Finconditional-setsubscriptlimit-supremum𝑛subscript𝐴𝑛subscript𝐴0subscript𝐴1\mathscr{L}(\mathrm{Fin}\times\mathcal{I})=\left\{\limsup\nolimits_{n}A_{n}:A_% {0},A_{1},\ldots\in\mathscr{L}(\mathcal{I})\right\}.script_L ( roman_Fin × caligraphic_I ) = { lim sup start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT italic_A start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT : italic_A start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , italic_A start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , … ∈ script_L ( caligraphic_I ) } .
Proof.

It follows by Theorem 2.1 that

(Fin×)={Λ𝒙(Fin×):𝒙Xω}={{ηX:nω,ηΛ𝒙(n)()}:𝒙Xω}={nknΛ𝒙(n)():𝒙(0),𝒙(1),Xω}={lim supnAn:A0,A1,()},Finconditional-setsubscriptΛ𝒙Fin𝒙superscript𝑋𝜔conditional-setconditional-set𝜂𝑋formulae-sequencesuperscript𝑛𝜔𝜂subscriptΛsuperscript𝒙𝑛𝒙superscript𝑋𝜔conditional-setsubscript𝑛subscript𝑘𝑛subscriptΛsuperscript𝒙𝑛superscript𝒙0superscript𝒙1superscript𝑋𝜔conditional-setsubscriptlimit-supremum𝑛subscript𝐴𝑛subscript𝐴0subscript𝐴1\begin{split}\mathscr{L}(\mathrm{Fin}\times\mathcal{I})&=\left\{\Lambda_{\bm{x% }}(\mathrm{Fin}\times\mathcal{I}):{\bm{x}}\in X^{\omega}\right\}\\ &=\left\{\left\{\eta\in X:\exists^{\infty}n\in\omega,\eta\in\Lambda_{{\bm{x}}^% {(n)}}(\mathcal{I})\right\}:{\bm{x}}\in X^{\omega}\right\}\\ &=\left\{\bigcap\nolimits_{n}\bigcup\nolimits_{k\geq n}\Lambda_{{\bm{x}}^{(n)}% }(\mathcal{I}):{\bm{x}}^{(0)},{\bm{x}}^{(1)},\ldots\in X^{\omega}\right\}\\ &=\left\{\limsup\nolimits_{n}A_{n}:A_{0},A_{1},\ldots\in\mathscr{L}(\mathcal{I% })\right\},\end{split}start_ROW start_CELL script_L ( roman_Fin × caligraphic_I ) end_CELL start_CELL = { roman_Λ start_POSTSUBSCRIPT bold_italic_x end_POSTSUBSCRIPT ( roman_Fin × caligraphic_I ) : bold_italic_x ∈ italic_X start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT } end_CELL end_ROW start_ROW start_CELL end_CELL start_CELL = { { italic_η ∈ italic_X : ∃ start_POSTSUPERSCRIPT ∞ end_POSTSUPERSCRIPT italic_n ∈ italic_ω , italic_η ∈ roman_Λ start_POSTSUBSCRIPT bold_italic_x start_POSTSUPERSCRIPT ( italic_n ) end_POSTSUPERSCRIPT end_POSTSUBSCRIPT ( caligraphic_I ) } : bold_italic_x ∈ italic_X start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT } end_CELL end_ROW start_ROW start_CELL end_CELL start_CELL = { ⋂ start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT ⋃ start_POSTSUBSCRIPT italic_k ≥ italic_n end_POSTSUBSCRIPT roman_Λ start_POSTSUBSCRIPT bold_italic_x start_POSTSUPERSCRIPT ( italic_n ) end_POSTSUPERSCRIPT end_POSTSUBSCRIPT ( caligraphic_I ) : bold_italic_x start_POSTSUPERSCRIPT ( 0 ) end_POSTSUPERSCRIPT , bold_italic_x start_POSTSUPERSCRIPT ( 1 ) end_POSTSUPERSCRIPT , … ∈ italic_X start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT } end_CELL end_ROW start_ROW start_CELL end_CELL start_CELL = { lim sup start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT italic_A start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT : italic_A start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , italic_A start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , … ∈ script_L ( caligraphic_I ) } , end_CELL end_ROW

completing the proof. ∎

At this point, recall that FinFin\mathrm{Fin}roman_Fin is a Fσsubscript𝐹𝜎F_{\sigma}italic_F start_POSTSUBSCRIPT italic_σ end_POSTSUBSCRIPT-ideal, and that ×FinFin\emptyset\times\mathrm{Fin}∅ × roman_Fin is an analytic P𝑃Pitalic_P-ideal which is not Fσsubscript𝐹𝜎F_{\sigma}italic_F start_POSTSUBSCRIPT italic_σ end_POSTSUBSCRIPT, see [7, Example 1.2.3] and [25, Exercise 23.1]. Thus, it follows by Theorem 1.2 that

(Fin)=Π10 and (×Fin)=Σ20.formulae-sequenceFinsubscriptsuperscriptΠ01 and FinsubscriptsuperscriptΣ02\mathscr{L}(\mathrm{Fin})=\Pi^{0}_{1}\quad\text{ and }\quad\mathscr{L}(% \emptyset\times\mathrm{Fin})=\Sigma^{0}_{2}.script_L ( roman_Fin ) = roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT and script_L ( ∅ × roman_Fin ) = roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT . (2)

With the above premises, we are able to extend (2) to certain ideals with large Borel complexity families ()\mathscr{L}(\mathcal{I})script_L ( caligraphic_I ).

Theorem 2.5.

Let X𝑋Xitalic_X be a complete metrizable space. Then, for each positive integer α𝛼\alphaitalic_α, we have:

  1. (i)

    FinαsuperscriptFin𝛼\mathrm{Fin}^{\alpha}roman_Fin start_POSTSUPERSCRIPT italic_α end_POSTSUPERSCRIPT is a Σ2α0subscriptsuperscriptΣ02𝛼\Sigma^{0}_{2\alpha}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 italic_α end_POSTSUBSCRIPT-ideal and (Finα)=Π2α10superscriptFin𝛼subscriptsuperscriptΠ02𝛼1\mathscr{L}(\mathrm{Fin}^{\alpha})=\Pi^{0}_{2\alpha-1}script_L ( roman_Fin start_POSTSUPERSCRIPT italic_α end_POSTSUPERSCRIPT ) = roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 italic_α - 1 end_POSTSUBSCRIPT.

  2. (ii)

    ×FinαsuperscriptFin𝛼\emptyset\times\mathrm{Fin}^{\alpha}∅ × roman_Fin start_POSTSUPERSCRIPT italic_α end_POSTSUPERSCRIPT is a Π2α+10subscriptsuperscriptΠ02𝛼1\Pi^{0}_{2\alpha+1}roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 italic_α + 1 end_POSTSUBSCRIPT-ideal and (×Finα)=Σ2α0superscriptFin𝛼subscriptsuperscriptΣ02𝛼\mathscr{L}(\emptyset\times\mathrm{Fin}^{\alpha})=\Sigma^{0}_{2\alpha}script_L ( ∅ × roman_Fin start_POSTSUPERSCRIPT italic_α end_POSTSUPERSCRIPT ) = roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 italic_α end_POSTSUBSCRIPT.

Proof.

(i) The complexity of FinαsuperscriptFin𝛼\mathrm{Fin}^{\alpha}roman_Fin start_POSTSUPERSCRIPT italic_α end_POSTSUPERSCRIPT is obtained applying recursively [28, Proposition 1.6.16], while the computation of (Finα)superscriptFin𝛼\mathscr{L}(\mathrm{Fin}^{\alpha})script_L ( roman_Fin start_POSTSUPERSCRIPT italic_α end_POSTSUPERSCRIPT ) is obtained putting together the base case (2), Corollary 2.4, and [25, Exercise 23.5(i)].

The proof of (ii) goes similarly, replacing Corollary 2.4 with Corollary 2.3. ∎

In particular, if X𝑋Xitalic_X is a complete metrizable space, X(Fin2)=Π30subscript𝑋superscriptFin2subscriptsuperscriptΠ03\mathscr{L}_{X}(\mathrm{Fin}^{2})=\Pi^{0}_{3}script_L start_POSTSUBSCRIPT italic_X end_POSTSUBSCRIPT ( roman_Fin start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT ) = roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT. This provides a generalization of [2, Example 4.2], where it is proved constructively that there exists a real sequence 𝒙𝒙\bm{x}bold_italic_x such that Λ𝒙(Fin2)subscriptΛ𝒙superscriptFin2\Lambda_{\bm{x}}(\mathrm{Fin}^{2})roman_Λ start_POSTSUBSCRIPT bold_italic_x end_POSTSUBSCRIPT ( roman_Fin start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT ) is equal to [0,1]01[0,1]\setminus\mathbb{Q}[ 0 , 1 ] ∖ blackboard_Q (note that the latter is not a Fσsubscript𝐹𝜎F_{\sigma}italic_F start_POSTSUBSCRIPT italic_σ end_POSTSUBSCRIPT-set, hence (Fin2)(Π20Σ20)subscriptsuperscriptFin2subscriptsuperscriptΠ02subscriptsuperscriptΣ02\mathscr{L}_{\mathbb{R}}(\mathrm{Fin}^{2})\cap(\Pi^{0}_{2}\setminus\Sigma^{0}_% {2})script_L start_POSTSUBSCRIPT blackboard_R end_POSTSUBSCRIPT ( roman_Fin start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT ) ∩ ( roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ∖ roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ) is nonempty).

Our second main result deals with ideals with small topological complexity. Suppose that X𝑋Xitalic_X is a first countable space, and recall that

()Π10,subscriptsuperscriptΠ01\mathscr{L}(\mathcal{I})\subseteq\Pi^{0}_{1},script_L ( caligraphic_I ) ⊆ roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , (3)

provided \mathcal{I}caligraphic_I is a Σ20subscriptsuperscriptΣ02\Sigma^{0}_{2}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT-ideal, see [2, Theorem 2.3]; note that the Hausdorffness hypothesis is not needed here. In the next result, we are going to show that the same conclusion holds if \mathcal{I}caligraphic_I is a Σ30subscriptsuperscriptΣ03\Sigma^{0}_{3}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT-ideal.

Theorem 2.6.

Let X𝑋Xitalic_X be a first countable space, and suppose that \mathcal{I}caligraphic_I is a Σ30subscriptsuperscriptΣ03\Sigma^{0}_{3}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT-ideal. Then inclusion (3) holds.

It is worth noting that Theorem 2.6 is not a consequence of the former result as it really includes new cases: indeed, as remarked also by Solecki in [30, p. 345], there exists a Δ30subscriptsuperscriptΔ03\Delta^{0}_{3}roman_Δ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT-ideal on ω𝜔\omegaitalic_ω (hence, both Σ30subscriptsuperscriptΣ03\Sigma^{0}_{3}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT and Π30subscriptsuperscriptΠ03\Pi^{0}_{3}roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT) which is neither Σ20subscriptsuperscriptΣ02\Sigma^{0}_{2}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT nor Π20subscriptsuperscriptΠ02\Pi^{0}_{2}roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT, see [34]; cf. also [4, 5] and [28, Proposition 1.2.1].

In addition, Theorem 2.6 allows us to prove a generalization of the folklore result that every Σ20subscriptsuperscriptΣ02\Sigma^{0}_{2}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT-ideal is a P+superscript𝑃P^{+}italic_P start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT-ideal, see [23]. For a different proof of the second part, see also [9, Proposition 10.1].

Corollary 2.7.

Let \mathcal{I}caligraphic_I be a Σ30subscriptsuperscriptΣ03\Sigma^{0}_{3}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT-ideal on ω𝜔\omegaitalic_ω. Then \mathcal{I}caligraphic_I is a P+superscript𝑃P^{+}italic_P start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT-ideal, and there exists a Σ20subscriptsuperscriptΣ02\Sigma^{0}_{2}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT-ideal 𝒥𝒥\mathcal{J}caligraphic_J such that 𝒥𝒥\mathcal{I}\subseteq\mathcal{J}caligraphic_I ⊆ caligraphic_J.

Proof.

The first part follows putting together Theorem 2.6 and Theorem 1.3.(ii) (note that \mathcal{I}caligraphic_I is Borel, hence with the hereditary Baire property). The second part follows by the known fact that a Borel ideal on ω𝜔\omegaitalic_ω is contained in a Σ20subscriptsuperscriptΣ02\Sigma^{0}_{2}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT-ideal if and only if it is contained in a P+superscript𝑃P^{+}italic_P start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT-ideal, see [23]. ∎

At this point, we divide the proof of Theorem 2.6 into two intermediate steps. To this aim, we recall that properties of ideals can be often expressed by finding critical ideals with respect to some preorder, cf. e.g. the survey [19]. To this aim, let \mathcal{I}caligraphic_I and 𝒥𝒥\mathcal{J}caligraphic_J be two ideals on two countably infinite sets Z𝑍Zitalic_Z and W𝑊Witalic_W, respectively. Then we say that \mathcal{I}caligraphic_I is below 𝒥𝒥\mathcal{J}caligraphic_J in the Rudin–Blass ordering, shortened as

RB𝒥,subscriptRB𝒥\mathcal{I}\leq_{\mathrm{RB}}\mathcal{J},caligraphic_I ≤ start_POSTSUBSCRIPT roman_RB end_POSTSUBSCRIPT caligraphic_J ,

if there is a finite-to-one map ϕ:WZ:italic-ϕ𝑊𝑍\phi:W\to Zitalic_ϕ : italic_W → italic_Z such that S𝑆S\in\mathcal{I}italic_S ∈ caligraphic_I if and only if ϕ1[S]𝒥superscriptitalic-ϕ1delimited-[]𝑆𝒥\phi^{-1}[S]\in\mathcal{J}italic_ϕ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_S ] ∈ caligraphic_J for all subsets SZ𝑆𝑍S\subseteq Zitalic_S ⊆ italic_Z. The restriction of these orderings to maximal ideals \mathcal{I}caligraphic_I, and the Borel cardinality of the quotients 𝒫(Z)/𝒫𝑍\mathcal{P}(Z)/\mathcal{I}caligraphic_P ( italic_Z ) / caligraphic_I have been extensively studied, see e.g. [6, 18, 26] and references therein.

Lemma 2.8.

Let X𝑋Xitalic_X be a first countable Hausdorff space and \mathcal{I}caligraphic_I be an ideal on ω𝜔\omegaitalic_ω with the hereditary Baire property. Suppose that inclusion (3) does not hold. Then ×FinRBsubscriptRBFin\emptyset\times\mathrm{Fin}\leq_{\mathrm{RB}}\mathcal{I}∅ × roman_Fin ≤ start_POSTSUBSCRIPT roman_RB end_POSTSUBSCRIPT caligraphic_I.

Proof.

Since inclusion (3) fails and X𝑋Xitalic_X is first countable, there exists a sequence 𝒙𝒙\bm{x}bold_italic_x such that S:=Λ𝒙()assign𝑆subscriptΛ𝒙S:=\Lambda_{\bm{x}}(\mathcal{I})italic_S := roman_Λ start_POSTSUBSCRIPT bold_italic_x end_POSTSUBSCRIPT ( caligraphic_I ) is not closed, hence not sequentially closed. Therefore there exists a sequence 𝒚𝒚\bm{y}bold_italic_y taking values in S𝑆Sitalic_S which is convergent to some limit ηXS𝜂𝑋𝑆\eta\in X\setminus Sitalic_η ∈ italic_X ∖ italic_S. Since X𝑋Xitalic_X is Hausdorff, we may suppose without loss of generality that ynymsubscript𝑦𝑛subscript𝑦𝑚y_{n}\neq y_{m}italic_y start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT ≠ italic_y start_POSTSUBSCRIPT italic_m end_POSTSUBSCRIPT for all distinct n,mω𝑛𝑚𝜔n,m\in\omegaitalic_n , italic_m ∈ italic_ω. Now, for each nω𝑛𝜔n\in\omegaitalic_n ∈ italic_ω, there exists An+subscript𝐴𝑛superscriptA_{n}\in\mathcal{I}^{+}italic_A start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT ∈ caligraphic_I start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT such that limkAnxk=ynsubscript𝑘subscript𝐴𝑛subscript𝑥𝑘subscript𝑦𝑛\lim_{k\in A_{n}}x_{k}=y_{n}roman_lim start_POSTSUBSCRIPT italic_k ∈ italic_A start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT end_POSTSUBSCRIPT italic_x start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT = italic_y start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT. Define B0:=A0(ωnAn)assignsubscript𝐵0subscript𝐴0𝜔subscript𝑛subscript𝐴𝑛B_{0}:=A_{0}\cup(\omega\setminus\bigcup_{n}A_{n})italic_B start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT := italic_A start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT ∪ ( italic_ω ∖ ⋃ start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT italic_A start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT ) and, recursively, Bn+1:=An+1knBkassignsubscript𝐵𝑛1subscript𝐴𝑛1subscript𝑘𝑛subscript𝐵𝑘B_{n+1}:=A_{n+1}\setminus\bigcup_{k\leq n}B_{k}italic_B start_POSTSUBSCRIPT italic_n + 1 end_POSTSUBSCRIPT := italic_A start_POSTSUBSCRIPT italic_n + 1 end_POSTSUBSCRIPT ∖ ⋃ start_POSTSUBSCRIPT italic_k ≤ italic_n end_POSTSUBSCRIPT italic_B start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT for all nω𝑛𝜔n\in\omegaitalic_n ∈ italic_ω. Hence {Bn:nω}conditional-setsubscript𝐵𝑛𝑛𝜔\{B_{n}:n\in\omega\}{ italic_B start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT : italic_n ∈ italic_ω } is a partition of ω𝜔\omegaitalic_ω into \mathcal{I}caligraphic_I-positive sets, for each nω𝑛𝜔n\in\omegaitalic_n ∈ italic_ω, the restriction Bnsubscript𝐵𝑛\mathcal{I}\upharpoonright B_{n}caligraphic_I ↾ italic_B start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT is an ideal on Bnsubscript𝐵𝑛B_{n}italic_B start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT with the Baire property.

It follows by Talagrand’s characterization of meager ideals that FinRBBnsubscriptRBFinsubscript𝐵𝑛\mathrm{Fin}\leq_{\mathrm{RB}}\mathcal{I}\upharpoonright B_{n}roman_Fin ≤ start_POSTSUBSCRIPT roman_RB end_POSTSUBSCRIPT caligraphic_I ↾ italic_B start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT for each nω𝑛𝜔n\in\omegaitalic_n ∈ italic_ω. More explicitly, for each nω𝑛𝜔n\in\omegaitalic_n ∈ italic_ω there exists a finite-to-one map ϕn:Bnω:subscriptitalic-ϕ𝑛subscript𝐵𝑛𝜔\phi_{n}:B_{n}\to\omegaitalic_ϕ start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT : italic_B start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT → italic_ω such that

Wω,ϕn1[W] if and only if WFin,formulae-sequencefor-all𝑊𝜔superscriptsubscriptitalic-ϕ𝑛1delimited-[]𝑊 if and only if 𝑊Fin\forall W\subseteq\omega,\quad\phi_{n}^{-1}[W]\in\mathcal{I}\,\,\,\text{ if % and only if }\,\,\,W\in\mathrm{Fin},∀ italic_W ⊆ italic_ω , italic_ϕ start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_W ] ∈ caligraphic_I if and only if italic_W ∈ roman_Fin , (4)

see [33, Theorem 2.1]; cf. also [1] for further characterizations of meager ideals based of \mathcal{I}caligraphic_I-limit points of sequences. Define the map ϕ:ωω2:italic-ϕ𝜔superscript𝜔2\phi:\omega\to\omega^{2}italic_ϕ : italic_ω → italic_ω start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT by

nω,kBn,ϕ(k)=(n,ϕn(k)).formulae-sequencefor-all𝑛𝜔formulae-sequencefor-all𝑘subscript𝐵𝑛italic-ϕ𝑘𝑛subscriptitalic-ϕ𝑛𝑘\forall n\in\omega,\forall k\in B_{n},\quad\phi(k)=(n,\phi_{n}(k)).∀ italic_n ∈ italic_ω , ∀ italic_k ∈ italic_B start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT , italic_ϕ ( italic_k ) = ( italic_n , italic_ϕ start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT ( italic_k ) ) .

We claim that ϕitalic-ϕ\phiitalic_ϕ is a witnessing function for ×FinRBsubscriptRBFin\emptyset\times\mathrm{Fin}\leq_{\mathrm{RB}}\mathcal{I}∅ × roman_Fin ≤ start_POSTSUBSCRIPT roman_RB end_POSTSUBSCRIPT caligraphic_I. For, suppose that Wω2𝑊superscript𝜔2W\subseteq\omega^{2}italic_W ⊆ italic_ω start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT belongs to ×FinFin\emptyset\times\mathrm{Fin}∅ × roman_Fin. Then

ϕ1[W]=nωϕn1[{kω:(n,k)W}]superscriptitalic-ϕ1delimited-[]𝑊subscript𝑛𝜔superscriptsubscriptitalic-ϕ𝑛1delimited-[]conditional-set𝑘𝜔𝑛𝑘𝑊\phi^{-1}[W]=\bigcup_{n\in\omega}\phi_{n}^{-1}[\{k\in\omega:(n,k)\in W\}]italic_ϕ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_W ] = ⋃ start_POSTSUBSCRIPT italic_n ∈ italic_ω end_POSTSUBSCRIPT italic_ϕ start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ { italic_k ∈ italic_ω : ( italic_n , italic_k ) ∈ italic_W } ] (5)

Since each ϕnsubscriptitalic-ϕ𝑛\phi_{n}italic_ϕ start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT is finite-to-one, ϕ1[W]superscriptitalic-ϕ1delimited-[]𝑊\phi^{-1}[W]italic_ϕ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_W ] has finite intersection with each Bnsubscript𝐵𝑛B_{n}italic_B start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT. Then either ϕ1[W]superscriptitalic-ϕ1delimited-[]𝑊\phi^{-1}[W]italic_ϕ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_W ] is finite or the subsequence (xn:nϕ1[W]):subscript𝑥𝑛𝑛superscriptitalic-ϕ1delimited-[]𝑊(x_{n}:n\in\phi^{-1}[W])( italic_x start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT : italic_n ∈ italic_ϕ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_W ] ) is convergent to η𝜂\etaitalic_η, while η𝜂\etaitalic_η is not an \mathcal{I}caligraphic_I-limit point of 𝒙𝒙\bm{x}bold_italic_x. Hence, in both cases, ϕ1[W]superscriptitalic-ϕ1delimited-[]𝑊\phi^{-1}[W]\in\mathcal{I}italic_ϕ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_W ] ∈ caligraphic_I. Conversely, suppose that W𝑊Witalic_W does not belong to ×FinFin\emptyset\times\mathrm{Fin}∅ × roman_Fin, so that there exists n0ωsubscript𝑛0𝜔n_{0}\in\omegaitalic_n start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT ∈ italic_ω such that {kω:(n0,k)W}Finconditional-set𝑘𝜔subscript𝑛0𝑘𝑊Fin\{k\in\omega:(n_{0},k)\in W\}\notin\mathrm{Fin}{ italic_k ∈ italic_ω : ( italic_n start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , italic_k ) ∈ italic_W } ∉ roman_Fin. It follows by (4) and (5) that ϕ1[W]superscriptitalic-ϕ1delimited-[]𝑊\phi^{-1}[W]italic_ϕ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_W ] contains

ϕn01[{kω:(n0,k)W}]+.superscriptsubscriptitalic-ϕsubscript𝑛01delimited-[]conditional-set𝑘𝜔subscript𝑛0𝑘𝑊superscript\phi_{n_{0}}^{-1}[\{k\in\omega:(n_{0},k)\in W\}]\in\mathcal{I}^{+}.italic_ϕ start_POSTSUBSCRIPT italic_n start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT end_POSTSUBSCRIPT start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ { italic_k ∈ italic_ω : ( italic_n start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , italic_k ) ∈ italic_W } ] ∈ caligraphic_I start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT .

Therefore ϕ1[W]superscriptitalic-ϕ1delimited-[]𝑊\phi^{-1}[W]\in\mathcal{I}italic_ϕ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_W ] ∈ caligraphic_I if and only if W×Fin𝑊FinW\in\emptyset\times\mathrm{Fin}italic_W ∈ ∅ × roman_Fin for each Wω2𝑊superscript𝜔2W\subseteq\omega^{2}italic_W ⊆ italic_ω start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT. ∎

Corollary 2.9.

Under the same hypotheses of Lemma 2.8, Σ20()subscriptsuperscriptΣ02\Sigma^{0}_{2}\subseteq\mathscr{L}(\mathcal{I})roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ⊆ script_L ( caligraphic_I ).

Proof.

Thanks to Lemma 2.8 and [16, Proposition 3.8], we have (×Fin)()Fin\mathscr{L}(\emptyset\times\mathrm{Fin})\subseteq\mathscr{L}(\mathcal{I})script_L ( ∅ × roman_Fin ) ⊆ script_L ( caligraphic_I ). The claim follows by Equation (2). ∎

For the next intermediate result, given topological spaces X,Y𝑋𝑌X,Yitalic_X , italic_Y and subsets AX𝐴𝑋A\subseteq Xitalic_A ⊆ italic_X and BY𝐵𝑌B\subseteq Yitalic_B ⊆ italic_Y, we say that A𝐴Aitalic_A is Wadge reducible to B𝐵Bitalic_B, shortened as

AWB,subscriptW𝐴𝐵A\leq_{\mathrm{W}}B,italic_A ≤ start_POSTSUBSCRIPT roman_W end_POSTSUBSCRIPT italic_B ,

if there exists a continuous map Φ:XY:Φ𝑋𝑌\Phi:X\to Yroman_Φ : italic_X → italic_Y such that Φ1[B]=AsuperscriptΦ1delimited-[]𝐵𝐴\Phi^{-1}[B]=Aroman_Φ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_B ] = italic_A (or, equivalently, xA𝑥𝐴x\in Aitalic_x ∈ italic_A if and only if Φ(x)BΦ𝑥𝐵\Phi(x)\in Broman_Φ ( italic_x ) ∈ italic_B for all xX𝑥𝑋x\in Xitalic_x ∈ italic_X), see e.g. [25, Definition 21.13]. If, in addition, X𝑋Xitalic_X and Y𝑌Yitalic_Y are Polish spaces with X𝑋Xitalic_X zero-dimensional, then B𝐵Bitalic_B is said to be Π30subscriptsuperscriptΠ03\Pi^{0}_{3}roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT-hard if AWBsubscriptW𝐴𝐵A\leq_{\mathrm{W}}Bitalic_A ≤ start_POSTSUBSCRIPT roman_W end_POSTSUBSCRIPT italic_B for some AΠ30(X)𝐴subscriptsuperscriptΠ03𝑋A\in\Pi^{0}_{3}(X)italic_A ∈ roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT ( italic_X ). Lastly, if B𝐵Bitalic_B is a Π30(Y)subscriptsuperscriptΠ03𝑌\Pi^{0}_{3}(Y)roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT ( italic_Y ) set which is also Π30subscriptsuperscriptΠ03\Pi^{0}_{3}roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT-hard, then it is called Π30subscriptsuperscriptΠ03\Pi^{0}_{3}roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT-complete. (Analogous definitions can be given for other classes of sets in Polish spaces, see [25, Definition 22.9].)

Lemma 2.10.

Let \mathcal{I}caligraphic_I be an ideal on ω𝜔\omegaitalic_ω such that ×FinRBsubscriptRBFin\emptyset\times\mathrm{Fin}\leq_{\mathrm{RB}}\mathcal{I}∅ × roman_Fin ≤ start_POSTSUBSCRIPT roman_RB end_POSTSUBSCRIPT caligraphic_I. Then \mathcal{I}caligraphic_I is Π30subscriptsuperscriptΠ03\Pi^{0}_{3}roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT-hard.

Proof.

First, recall that ×FinFin\emptyset\times\mathrm{Fin}∅ × roman_Fin is a Π30subscriptsuperscriptΠ03\Pi^{0}_{3}roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT-complete subset of 𝒫(ω2)𝒫superscript𝜔2\mathcal{P}(\omega^{2})caligraphic_P ( italic_ω start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT ), see e.g. [25, Exercise 23.1]. Hence, to complete the proof, it is sufficient to show that

×FinW.subscriptWFin\emptyset\times\mathrm{Fin}\leq_{\mathrm{W}}\mathcal{I}.∅ × roman_Fin ≤ start_POSTSUBSCRIPT roman_W end_POSTSUBSCRIPT caligraphic_I .

By hypothesis, there exists a finite-to-one function ϕ:ωω2:italic-ϕ𝜔superscript𝜔2\phi:\omega\to\omega^{2}italic_ϕ : italic_ω → italic_ω start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT such that S×Fin𝑆FinS\in\emptyset\times\mathrm{Fin}italic_S ∈ ∅ × roman_Fin if and only if ϕ1[S]superscriptitalic-ϕ1delimited-[]𝑆\phi^{-1}[S]\in\mathcal{I}italic_ϕ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_S ] ∈ caligraphic_I. Now, define the map Φ:𝒫(ω2)𝒫(ω):Φ𝒫superscript𝜔2𝒫𝜔\Phi:\mathcal{P}(\omega^{2})\to\mathcal{P}(\omega)roman_Φ : caligraphic_P ( italic_ω start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT ) → caligraphic_P ( italic_ω ) by Φ(S):=ϕ1[S]assignΦ𝑆superscriptitalic-ϕ1delimited-[]𝑆\Phi(S):=\phi^{-1}[S]roman_Φ ( italic_S ) := italic_ϕ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_S ], so that S×Fin𝑆FinS\in\emptyset\times\mathrm{Fin}italic_S ∈ ∅ × roman_Fin if and only if Φ(S)Φ𝑆\Phi(S)\in\mathcal{I}roman_Φ ( italic_S ) ∈ caligraphic_I. Hence, we only need to show that ΦΦ\Phiroman_Φ is continuous. For, fix nω𝑛𝜔n\in\omegaitalic_n ∈ italic_ω and define S:=ϕ[{0,,n}]assign𝑆italic-ϕdelimited-[]0𝑛S:=\phi[\{0,\ldots,n\}]italic_S := italic_ϕ [ { 0 , … , italic_n } ]. It follows that, for all A,Bω2𝐴𝐵superscript𝜔2A,B\subseteq\omega^{2}italic_A , italic_B ⊆ italic_ω start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT with AS=BS𝐴𝑆𝐵𝑆A\cap S=B\cap Sitalic_A ∩ italic_S = italic_B ∩ italic_S,

Φ(A)[0,n]={kω:ϕ(k)A and kn}={kω:ϕ(k)AS}=Φ(B)[0,n],Φ𝐴0𝑛conditional-set𝑘𝜔italic-ϕ𝑘𝐴 and 𝑘𝑛conditional-set𝑘𝜔italic-ϕ𝑘𝐴𝑆Φ𝐵0𝑛\begin{split}\Phi(A)\cap[0,n]&=\{k\in\omega:\phi(k)\in A\text{ and }k\leq n\}% \\ &=\{k\in\omega:\phi(k)\in A\cap S\}=\Phi(B)\cap[0,n],\end{split}start_ROW start_CELL roman_Φ ( italic_A ) ∩ [ 0 , italic_n ] end_CELL start_CELL = { italic_k ∈ italic_ω : italic_ϕ ( italic_k ) ∈ italic_A and italic_k ≤ italic_n } end_CELL end_ROW start_ROW start_CELL end_CELL start_CELL = { italic_k ∈ italic_ω : italic_ϕ ( italic_k ) ∈ italic_A ∩ italic_S } = roman_Φ ( italic_B ) ∩ [ 0 , italic_n ] , end_CELL end_ROW

which proves the continuity of ΦΦ\Phiroman_Φ. ∎

Proof of Theorem 2.6.

Suppose that X𝑋Xitalic_X is a first countable space and \mathcal{I}caligraphic_I is an ideal on ω𝜔\omegaitalic_ω such that inclusion (3) fails. We claim that \mathcal{I}caligraphic_I is not a Σ30subscriptsuperscriptΣ03\Sigma^{0}_{3}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT-ideal.

If \mathcal{I}caligraphic_I is not a Borel ideal, then the claim is trivial. Hence, let us suppose hereafter that \mathcal{I}caligraphic_I is Borel. In particular, \mathcal{I}caligraphic_I has the hereditary Baire property. At this point, it follows by Lemma 2.8 and Lemma 2.10 that \mathcal{I}caligraphic_I is a Π30subscriptsuperscriptΠ03\Pi^{0}_{3}roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT-hard. To sum up, \mathcal{I}caligraphic_I is a Borel subset of a zero-dimensional Polish space and it is Π30subscriptsuperscriptΠ03\Pi^{0}_{3}roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT-hard. We conclude by [25, Therem 22.10] that the \mathcal{I}caligraphic_I is not a Σ30subscriptsuperscriptΣ03\Sigma^{0}_{3}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT-ideal. ∎

3. Simply analyticity

In [2, Proposition 4.1], the first and last-named authors proved the following:

Proposition 3.1.

Let 𝐱𝐱\bm{x}bold_italic_x be a sequence taking values in a Hausdorff regular first countable space X𝑋Xitalic_X. Let also \mathcal{I}caligraphic_I be a coanalytic ideal on ω𝜔\omegaitalic_ω. Then Λ𝐱()subscriptΛ𝐱\Lambda_{\bm{x}}(\mathcal{I})roman_Λ start_POSTSUBSCRIPT bold_italic_x end_POSTSUBSCRIPT ( caligraphic_I ) is analytic.

However, the classical definition of analytic sets AX𝐴𝑋A\subseteq Xitalic_A ⊆ italic_X as projections of Borel subsets of X×X𝑋𝑋X\times Xitalic_X × italic_X is usually considered in Polish spaces X𝑋Xitalic_X, cf. e.g. [31, Chapter 4] or [25, Chapter 14]. In addition, the above result has been re-proved in [17, Theorem 4.1] for zero-dimensional Polish spaces X𝑋Xitalic_X, so that

()Σ11.subscriptsuperscriptΣ11\mathscr{L}(\mathcal{I})\subseteq\Sigma^{1}_{1}.script_L ( caligraphic_I ) ⊆ roman_Σ start_POSTSUPERSCRIPT 1 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT .

whenever \mathcal{I}caligraphic_I is a Π11subscriptsuperscriptΠ11\Pi^{1}_{1}roman_Π start_POSTSUPERSCRIPT 1 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT-ideal (notice that the above notation is meaningful).

In this Section, our aim is to reformulate and clarify the statement and the proof of Proposition 3.1, extending in turn both the latter and the special case treated in [17, Theorem 4.1].

Note that there are several papers in which the notion of analytic set is adapted to more general topological spaces, see e.g. [15, 29]. The theory of the latter extension, called K𝐾Kitalic_K-analytic sets, is important and nontrivial, and it does not proceed verbatim as in the classical one of analytic sets.

For our purposes, we generalize in a straight way one of possible definitions of analytic sets to the case of arbitrary topological spaces. Hereafter, πXsubscript𝜋𝑋\pi_{X}italic_π start_POSTSUBSCRIPT italic_X end_POSTSUBSCRIPT stands for the usual projection on X𝑋Xitalic_X.

Definition 3.2.

Let X𝑋Xitalic_X be a topological space. A subset AX𝐴𝑋A\subseteq Xitalic_A ⊆ italic_X is said to be simply analytic, shortened as s-analytic, if there exists an uncountable Polish space Y𝑌Yitalic_Y and a Borel subset BX×Y𝐵𝑋𝑌B\subseteq X\times Yitalic_B ⊆ italic_X × italic_Y such that A=πX[B]𝐴subscript𝜋𝑋delimited-[]𝐵A=\pi_{X}[B]italic_A = italic_π start_POSTSUBSCRIPT italic_X end_POSTSUBSCRIPT [ italic_B ].

Note that Definition 3.2 is coherent with the classical notion of analytic set. Indeed, it is well known that a subset A𝐴Aitalic_A of a Polish space X𝑋Xitalic_X is analytic if and only if there exists an uncountable Polish space Y𝑌Yitalic_Y and a Borel BX×Y𝐵𝑋𝑌B\subseteq X\times Yitalic_B ⊆ italic_X × italic_Y such that A=πX[B]𝐴subscript𝜋𝑋delimited-[]𝐵A=\pi_{X}[B]italic_A = italic_π start_POSTSUBSCRIPT italic_X end_POSTSUBSCRIPT [ italic_B ], cf. [31, Proposition 4.1.1]. At this point, let Y1,Y2subscript𝑌1subscript𝑌2Y_{1},Y_{2}italic_Y start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , italic_Y start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT be two uncountable Polish spaces. Then there exists a Borel isomorphism h:Y2Y1:subscript𝑌2subscript𝑌1h:Y_{2}\to Y_{1}italic_h : italic_Y start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT → italic_Y start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT, as it follows by [31, Theorem 3.3.13]. Let A:=πX[B1]assign𝐴subscript𝜋𝑋delimited-[]subscript𝐵1A:=\pi_{X}[B_{1}]italic_A := italic_π start_POSTSUBSCRIPT italic_X end_POSTSUBSCRIPT [ italic_B start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ] be a s-analytic subset for a Borel set B1X×Y1subscript𝐵1𝑋subscript𝑌1B_{1}\subseteq X\times Y_{1}italic_B start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ⊆ italic_X × italic_Y start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT. Then B2:={(x,y)X×Y2:(x,h(y))B1}assignsubscript𝐵2conditional-set𝑥𝑦𝑋subscript𝑌2𝑥𝑦subscript𝐵1B_{2}:=\{(x,y)\in X\times Y_{2}:(x,h(y))\in B_{1}\}italic_B start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT := { ( italic_x , italic_y ) ∈ italic_X × italic_Y start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT : ( italic_x , italic_h ( italic_y ) ) ∈ italic_B start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT } is a Borel subset of X×Y2𝑋subscript𝑌2X\times Y_{2}italic_X × italic_Y start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT by Lemma 3.4 below and, clearly, πX[B2]=Asubscript𝜋𝑋delimited-[]subscript𝐵2𝐴\pi_{X}[B_{2}]=Aitalic_π start_POSTSUBSCRIPT italic_X end_POSTSUBSCRIPT [ italic_B start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ] = italic_A. Therefore, in Definition 3.2 one may assume without loss of generality that Y=ωω𝑌superscript𝜔𝜔Y=\omega^{\omega}italic_Y = italic_ω start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT.

Lastly, a subset AX𝐴𝑋A\subseteq Xitalic_A ⊆ italic_X is said to be s-coanalytic if Ac:=XAassignsuperscript𝐴𝑐𝑋𝐴A^{c}:=X\setminus Aitalic_A start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT := italic_X ∖ italic_A is s-analytic. Observe that every Borel set in X𝑋Xitalic_X is both s-analytic and s-coanalytic.

These observations enable us to consider Proposition 3.1 valid as it was stated in the original version in [2]. They allow, in addition, to remove in its statement the hypothesis of regularity (and also the property of being Hausdorff; however, the latter one has been used in [2] only because a first countable space is Hausdorff if and only if every sequence has at most one limit):

Theorem 3.3.

Let 𝐱𝐱{\bm{x}}bold_italic_x be a sequence taking values in a first countable space X𝑋Xitalic_X. Let also \mathcal{I}caligraphic_I be a s-coanalytic ideal on ω𝜔\omegaitalic_ω. Then Λ𝐱()subscriptΛ𝐱\Lambda_{\bm{x}}(\mathcal{I})roman_Λ start_POSTSUBSCRIPT bold_italic_x end_POSTSUBSCRIPT ( caligraphic_I ) is s-analytic.

For the proof of Theorem 3.3, we will need some intermediate lemmas.

Lemma 3.4.

Let f:XZ:𝑓𝑋𝑍f:X\to Zitalic_f : italic_X → italic_Z and g:YW:𝑔𝑌𝑊g:Y\to Witalic_g : italic_Y → italic_W be Borel functions, where X,Y,Z,W𝑋𝑌𝑍𝑊X,Y,Z,Witalic_X , italic_Y , italic_Z , italic_W are topological spaces. Then the map h:X×YZ×W:𝑋𝑌𝑍𝑊h:X\times Y\to Z\times Witalic_h : italic_X × italic_Y → italic_Z × italic_W defined by

(x,y)X×Y,h(x,y):=(f(x),g(y))formulae-sequencefor-all𝑥𝑦𝑋𝑌assign𝑥𝑦𝑓𝑥𝑔𝑦\forall(x,y)\in X\times Y,\quad h(x,y):=(f(x),g(y))∀ ( italic_x , italic_y ) ∈ italic_X × italic_Y , italic_h ( italic_x , italic_y ) := ( italic_f ( italic_x ) , italic_g ( italic_y ) )

is Borel.

Proof.

Let UZ𝑈𝑍U\subseteq Zitalic_U ⊆ italic_Z and VW𝑉𝑊V\subseteq Witalic_V ⊆ italic_W be arbitrary open sets. It is enough to show that h1[U×V]superscript1delimited-[]𝑈𝑉h^{-1}[U\times V]italic_h start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_U × italic_V ] is a Borel subset of X×Y𝑋𝑌X\times Yitalic_X × italic_Y. The latter set is equal to f1[U]×g1[V]superscript𝑓1delimited-[]𝑈superscript𝑔1delimited-[]𝑉f^{-1}[U]\times g^{-1}[V]italic_f start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_U ] × italic_g start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_V ], which belongs to (X)(Y)tensor-product𝑋𝑌\mathscr{B}(X)\otimes\mathscr{B}(Y)script_B ( italic_X ) ⊗ script_B ( italic_Y ). The claim follows by the inclusion (X)(Y)(X×Y)tensor-product𝑋𝑌𝑋𝑌\mathscr{B}(X)\otimes\mathscr{B}(Y)\subseteq\mathscr{B}(X\times Y)script_B ( italic_X ) ⊗ script_B ( italic_Y ) ⊆ script_B ( italic_X × italic_Y ), see e.g. [11, Proposition 251E]. (For the converse inclusion of the latter, which does not hold for every X𝑋Xitalic_X and Y𝑌Yitalic_Y, see [12].) ∎

Lemma 3.5.

Let f:XY:𝑓𝑋𝑌f\colon X\to Yitalic_f : italic_X → italic_Y be a Borel function, where X𝑋Xitalic_X is a topological space and Y𝑌Yitalic_Y is an uncountable Polish space. If AY𝐴𝑌A\subseteq Yitalic_A ⊆ italic_Y is an analytic set, then f1[A]superscript𝑓1delimited-[]𝐴f^{-1}[A]italic_f start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_A ] is a s-analytic subset of X𝑋Xitalic_X.

Proof.

Since AY𝐴𝑌A\subseteq Yitalic_A ⊆ italic_Y is analytic, there exists a Borel subset BY×Y𝐵𝑌𝑌B\subseteq Y\times Yitalic_B ⊆ italic_Y × italic_Y such that A𝐴Aitalic_A is the projection on the first coordinate of B𝐵Bitalic_B. At this point, define C:={(x,y)X×Y:(f(x),y)B}assign𝐶conditional-set𝑥𝑦𝑋𝑌𝑓𝑥𝑦𝐵C:=\left\{(x,y)\in X\times Y:(f(x),y)\in B\right\}italic_C := { ( italic_x , italic_y ) ∈ italic_X × italic_Y : ( italic_f ( italic_x ) , italic_y ) ∈ italic_B }. Then C𝐶Citalic_C is Borel set by Lemma 3.4 and πX[C]=f1[A]subscript𝜋𝑋delimited-[]𝐶superscript𝑓1delimited-[]𝐴\pi_{X}[C]=f^{-1}[A]italic_π start_POSTSUBSCRIPT italic_X end_POSTSUBSCRIPT [ italic_C ] = italic_f start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_A ]. Therefore f1[A]superscript𝑓1delimited-[]𝐴f^{-1}[A]italic_f start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_A ] is a s-analytic set, which concludes the proof. ∎

Lemma 3.6.

Let AX×Y𝐴𝑋𝑌A\subseteq X\times Yitalic_A ⊆ italic_X × italic_Y be a s-analytic set, where X𝑋Xitalic_X is a topological space and Y𝑌Yitalic_Y is an uncountable Polish space. Then C:=πX[A]assign𝐶subscript𝜋𝑋delimited-[]𝐴C:=\pi_{X}[A]italic_C := italic_π start_POSTSUBSCRIPT italic_X end_POSTSUBSCRIPT [ italic_A ] is a s-analytic set.

Proof.

Let Z𝑍Zitalic_Z be an uncountable Polish space and BX×Y×Z𝐵𝑋𝑌𝑍B\subseteq X\times Y\times Zitalic_B ⊆ italic_X × italic_Y × italic_Z be a Borel set such that A=πX×Y[B]𝐴subscript𝜋𝑋𝑌delimited-[]𝐵A=\pi_{X\times Y}[B]italic_A = italic_π start_POSTSUBSCRIPT italic_X × italic_Y end_POSTSUBSCRIPT [ italic_B ]. Now, observe that C=πX[B]𝐶subscript𝜋𝑋delimited-[]𝐵C=\pi_{X}[B]italic_C = italic_π start_POSTSUBSCRIPT italic_X end_POSTSUBSCRIPT [ italic_B ] and obviously Y×Z𝑌𝑍Y\times Zitalic_Y × italic_Z is an uncountable Polish space. ∎

Lemma 3.7.

Let (An)subscript𝐴𝑛(A_{n})( italic_A start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT ) be a sequence of s-analytic subsets of a topological space X𝑋Xitalic_X. Then both nAnsubscript𝑛subscript𝐴𝑛\bigcup\nolimits_{n}A_{n}⋃ start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT italic_A start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT and nAnsubscript𝑛subscript𝐴𝑛\bigcap\nolimits_{n}A_{n}⋂ start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT italic_A start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT are s-analytic.

Proof.

The proof proceeds verbatim as in [31, Proposition 4.1.2]. ∎

We are ready for the proof of Theorem 3.3.

Proof of Theorem 3.3.

Note that +superscript\mathcal{I}^{+}caligraphic_I start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT is s-analytic and let 𝒩𝒩\mathcal{N}caligraphic_N be the set of strictly increasing sequences (nk)subscript𝑛𝑘(n_{k})( italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ) of nonnegative integers. Since 𝒩𝒩\mathcal{N}caligraphic_N can be regarded as a closed subset of ωωsuperscript𝜔𝜔\omega^{\omega}italic_ω start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT, it is a Polish space by Alexandrov’s theorem, see e.g. [31, Theorem 2.2.1]. For each ηX𝜂𝑋\eta\in Xitalic_η ∈ italic_X, fix a decreasing local base (Uη,m:mω):subscript𝑈𝜂𝑚𝑚𝜔\left(U_{\eta,m}:m\in\omega\right)( italic_U start_POSTSUBSCRIPT italic_η , italic_m end_POSTSUBSCRIPT : italic_m ∈ italic_ω ) of open neighbourhoods of η𝜂\etaitalic_η. Then η𝜂\etaitalic_η is an \mathcal{I}caligraphic_I-limit point of 𝒙𝒙\bm{x}bold_italic_x if and only if there exists a sequence (nk)𝒩subscript𝑛𝑘𝒩(n_{k})\in\mathcal{N}( italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ) ∈ caligraphic_N such that {nk:kω}+conditional-setsubscript𝑛𝑘𝑘𝜔superscript\{n_{k}:k\in\omega\}\in\mathcal{I}^{+}{ italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT : italic_k ∈ italic_ω } ∈ caligraphic_I start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT and {kω:xnkUη,m}Finconditional-set𝑘𝜔subscript𝑥subscript𝑛𝑘subscript𝑈𝜂𝑚Fin\{k\in\omega:x_{n_{k}}\notin U_{\eta,m}\}\in\mathrm{Fin}{ italic_k ∈ italic_ω : italic_x start_POSTSUBSCRIPT italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT end_POSTSUBSCRIPT ∉ italic_U start_POSTSUBSCRIPT italic_η , italic_m end_POSTSUBSCRIPT } ∈ roman_Fin for all mω𝑚𝜔m\in\omegaitalic_m ∈ italic_ω.

At this point, define the continuous function

ψ:𝒩{0,1}ω:(nk)χ{nk:kω},:𝜓𝒩superscript01𝜔:maps-tosubscript𝑛𝑘subscript𝜒conditional-setsubscript𝑛𝑘𝑘𝜔\psi\colon\mathcal{N}\to\{0,1\}^{\omega}:(n_{k})\mapsto\chi_{\{n_{k}\colon k% \in\omega\}},italic_ψ : caligraphic_N → { 0 , 1 } start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT : ( italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ) ↦ italic_χ start_POSTSUBSCRIPT { italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT : italic_k ∈ italic_ω } end_POSTSUBSCRIPT ,

where χSsubscript𝜒𝑆\chi_{S}italic_χ start_POSTSUBSCRIPT italic_S end_POSTSUBSCRIPT stands for the characteristic function of a set Sω𝑆𝜔S\subseteq\omegaitalic_S ⊆ italic_ω. In addition, for each mω𝑚𝜔m\in\omegaitalic_m ∈ italic_ω, define the function ζm:𝒩×X{0,1}ω:subscript𝜁𝑚𝒩𝑋superscript01𝜔\zeta_{m}\colon\mathcal{N}\times X\to\{0,1\}^{\omega}italic_ζ start_POSTSUBSCRIPT italic_m end_POSTSUBSCRIPT : caligraphic_N × italic_X → { 0 , 1 } start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT by

(nk)𝒩,ηX,ζm((nk),η):=χ{tω:xntUη,m}formulae-sequencefor-allsubscript𝑛𝑘𝒩formulae-sequencefor-all𝜂𝑋assignsubscript𝜁𝑚subscript𝑛𝑘𝜂subscript𝜒conditional-set𝑡𝜔subscript𝑥subscript𝑛𝑡subscript𝑈𝜂𝑚\forall(n_{k})\in\mathcal{N},\forall\eta\in X,\quad\zeta_{m}\left((n_{k}),\eta% \right):=\chi_{\{\,t\in\omega:\,x_{n_{t}}\notin U_{\eta,m}\}}∀ ( italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ) ∈ caligraphic_N , ∀ italic_η ∈ italic_X , italic_ζ start_POSTSUBSCRIPT italic_m end_POSTSUBSCRIPT ( ( italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ) , italic_η ) := italic_χ start_POSTSUBSCRIPT { italic_t ∈ italic_ω : italic_x start_POSTSUBSCRIPT italic_n start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT end_POSTSUBSCRIPT ∉ italic_U start_POSTSUBSCRIPT italic_η , italic_m end_POSTSUBSCRIPT } end_POSTSUBSCRIPT

Identifying each Sω𝑆𝜔S\subseteq\omegaitalic_S ⊆ italic_ω with its charateristic function χSsubscript𝜒𝑆\chi_{S}italic_χ start_POSTSUBSCRIPT italic_S end_POSTSUBSCRIPT, it follows that

Λ𝒙()=πX[(ψ1[+]×X)mωζm1[Fin]].subscriptΛ𝒙subscript𝜋𝑋delimited-[]superscript𝜓1delimited-[]superscript𝑋subscript𝑚𝜔subscriptsuperscript𝜁1𝑚delimited-[]Fin\Lambda_{\bm{x}}(\mathcal{I})=\pi_{X}\left[\left(\psi^{-1}[\,\mathcal{I}^{+}]% \times X\right)\cap\bigcap\nolimits_{m\in\omega}\zeta^{-1}_{m}[\mathrm{Fin}]% \right].roman_Λ start_POSTSUBSCRIPT bold_italic_x end_POSTSUBSCRIPT ( caligraphic_I ) = italic_π start_POSTSUBSCRIPT italic_X end_POSTSUBSCRIPT [ ( italic_ψ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ caligraphic_I start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT ] × italic_X ) ∩ ⋂ start_POSTSUBSCRIPT italic_m ∈ italic_ω end_POSTSUBSCRIPT italic_ζ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT italic_m end_POSTSUBSCRIPT [ roman_Fin ] ] . (6)
Claim 1.

ψ1[+]×Xsuperscript𝜓1delimited-[]superscript𝑋\psi^{-1}[\,\mathcal{I}^{+}]\times Xitalic_ψ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ caligraphic_I start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT ] × italic_X is a s-analytic subset of 𝒩×X𝒩𝑋\mathcal{N}\times Xcaligraphic_N × italic_X.

Proof.

Thanks to Lemma 3.5 and the fact that ψ𝜓\psiitalic_ψ is continuous, ψ1[+]superscript𝜓1delimited-[]superscript\psi^{-1}[\,\mathcal{I}^{+}]italic_ψ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ caligraphic_I start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT ] is an (ordinary) analytic subset of 𝒩𝒩\mathcal{N}caligraphic_N. Hence there exists an uncountable Polish space Y𝑌Yitalic_Y and a Borel set B𝒩×Y𝐵𝒩𝑌B\subseteq\mathcal{N}\times Yitalic_B ⊆ caligraphic_N × italic_Y such that ψ1[+]=π𝒩[B]superscript𝜓1delimited-[]superscriptsubscript𝜋𝒩delimited-[]𝐵\psi^{-1}[\,\mathcal{I}^{+}]=\pi_{\mathcal{N}}[B]italic_ψ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ caligraphic_I start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT ] = italic_π start_POSTSUBSCRIPT caligraphic_N end_POSTSUBSCRIPT [ italic_B ]. The claim follows from the fact that

B×X(𝒩×Y)×(X)(𝒩×Y×X),𝐵𝑋𝒩𝑌𝑋𝒩𝑌𝑋B\times X\in\mathscr{B}(\mathcal{N}\times Y)\times\mathscr{B}(X)\subseteq% \mathscr{B}(\mathcal{N}\times Y\times X),italic_B × italic_X ∈ script_B ( caligraphic_N × italic_Y ) × script_B ( italic_X ) ⊆ script_B ( caligraphic_N × italic_Y × italic_X ) ,

see e.g. [11, Proposition 251E], and that π𝒩×X[B×X]=π𝒩[B]×Xsubscript𝜋𝒩𝑋delimited-[]𝐵𝑋subscript𝜋𝒩delimited-[]𝐵𝑋\pi_{\mathcal{N}\times X}[B\times X]=\pi_{\mathcal{N}}[B]\times Xitalic_π start_POSTSUBSCRIPT caligraphic_N × italic_X end_POSTSUBSCRIPT [ italic_B × italic_X ] = italic_π start_POSTSUBSCRIPT caligraphic_N end_POSTSUBSCRIPT [ italic_B ] × italic_X. ∎

Claim 2.

For each ηX𝜂𝑋\eta\in Xitalic_η ∈ italic_X and mω𝑚𝜔m\in\omegaitalic_m ∈ italic_ω, the section ζm(,η)subscript𝜁𝑚𝜂\zeta_{m}(\,\cdot,\eta)italic_ζ start_POSTSUBSCRIPT italic_m end_POSTSUBSCRIPT ( ⋅ , italic_η ) is continuous.

Proof.

Fix ηX𝜂𝑋\eta\in Xitalic_η ∈ italic_X and mω𝑚𝜔m\in\omegaitalic_m ∈ italic_ω. It is enough to show that the the section ζm(,η)subscript𝜁𝑚𝜂\zeta_{m}(\,\cdot,\eta)italic_ζ start_POSTSUBSCRIPT italic_m end_POSTSUBSCRIPT ( ⋅ , italic_η ) is sequentially continuous. For, pick a sequence (nk(p):pω):superscriptsubscript𝑛𝑘𝑝𝑝𝜔(n_{k}^{(p)}:p\in\omega)( italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT start_POSTSUPERSCRIPT ( italic_p ) end_POSTSUPERSCRIPT : italic_p ∈ italic_ω ) of elements of 𝒩𝒩\mathcal{N}caligraphic_N which is convergent to some (nk)𝒩subscript𝑛𝑘𝒩(n_{k})\in\mathcal{N}( italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ) ∈ caligraphic_N. Then, for each tω𝑡𝜔t\in\omegaitalic_t ∈ italic_ω, there exists ptωsubscript𝑝𝑡𝜔p_{t}\in\omegaitalic_p start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ∈ italic_ω such that nt(p)=ntsuperscriptsubscript𝑛𝑡𝑝subscript𝑛𝑡n_{t}^{(p)}=n_{t}italic_n start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT start_POSTSUPERSCRIPT ( italic_p ) end_POSTSUPERSCRIPT = italic_n start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT for all ppt𝑝subscript𝑝𝑡p\geq p_{t}italic_p ≥ italic_p start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT. Hence ζm((nk(p)),η)(t)=ζm((nk),η)(t)subscript𝜁𝑚superscriptsubscript𝑛𝑘𝑝𝜂𝑡subscript𝜁𝑚subscript𝑛𝑘𝜂𝑡\zeta_{m}((n_{k}^{(p)}),\eta)(t)=\zeta_{m}((n_{k}),\eta)(t)italic_ζ start_POSTSUBSCRIPT italic_m end_POSTSUBSCRIPT ( ( italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT start_POSTSUPERSCRIPT ( italic_p ) end_POSTSUPERSCRIPT ) , italic_η ) ( italic_t ) = italic_ζ start_POSTSUBSCRIPT italic_m end_POSTSUBSCRIPT ( ( italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ) , italic_η ) ( italic_t ) for all ppt𝑝subscript𝑝𝑡p\geq p_{t}italic_p ≥ italic_p start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT. This proves that limpζm((nk(p)),η)=ζm((nk),η)subscript𝑝subscript𝜁𝑚superscriptsubscript𝑛𝑘𝑝𝜂subscript𝜁𝑚subscript𝑛𝑘𝜂\lim_{p}\zeta_{m}((n_{k}^{(p)}),\eta)=\zeta_{m}((n_{k}),\eta)roman_lim start_POSTSUBSCRIPT italic_p end_POSTSUBSCRIPT italic_ζ start_POSTSUBSCRIPT italic_m end_POSTSUBSCRIPT ( ( italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT start_POSTSUPERSCRIPT ( italic_p ) end_POSTSUPERSCRIPT ) , italic_η ) = italic_ζ start_POSTSUBSCRIPT italic_m end_POSTSUBSCRIPT ( ( italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ) , italic_η ). ∎

Claim 3.

For each (nk)𝒩subscript𝑛𝑘𝒩(n_{k})\in\mathcal{N}( italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ) ∈ caligraphic_N and mω𝑚𝜔m\in\omegaitalic_m ∈ italic_ω, the section ζm((nk),)subscript𝜁𝑚subscript𝑛𝑘\zeta_{m}((n_{k}),\cdot\,)italic_ζ start_POSTSUBSCRIPT italic_m end_POSTSUBSCRIPT ( ( italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ) , ⋅ ) is Borel measurable.

Proof.

Fix (nk)𝒩subscript𝑛𝑘𝒩(n_{k})\in\mathcal{N}( italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ) ∈ caligraphic_N and mω𝑚𝜔m\in\omegaitalic_m ∈ italic_ω and define for convenience ξ:X{0,1}ω:𝜉𝑋superscript01𝜔\xi:X\to\{0,1\}^{\omega}italic_ξ : italic_X → { 0 , 1 } start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT by ξ(η):=ζm((nk),η)assign𝜉𝜂subscript𝜁𝑚subscript𝑛𝑘𝜂\xi(\eta):=\zeta_{m}((n_{k}),\eta)italic_ξ ( italic_η ) := italic_ζ start_POSTSUBSCRIPT italic_m end_POSTSUBSCRIPT ( ( italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ) , italic_η ) for all ηX𝜂𝑋\eta\in Xitalic_η ∈ italic_X. It is enough to show that preimage ξ1[U]superscript𝜉1delimited-[]𝑈\xi^{-1}[U]italic_ξ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_U ] is Borel in X𝑋Xitalic_X for every basic open set U𝒩𝑈𝒩U\subseteq\mathcal{N}italic_U ⊆ caligraphic_N. This is clear if U=𝑈U=\emptysetitalic_U = ∅. Otherwise there exist pω𝑝𝜔p\in\omegaitalic_p ∈ italic_ω and r0,r1,,rp{0,1}subscript𝑟0subscript𝑟1subscript𝑟𝑝01r_{0},r_{1},\ldots,r_{p}\in\{0,1\}italic_r start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , italic_r start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , … , italic_r start_POSTSUBSCRIPT italic_p end_POSTSUBSCRIPT ∈ { 0 , 1 } such that U={a{0,1}ω:a(t)=rt for all t=0,1,,p}𝑈conditional-set𝑎superscript01𝜔formulae-sequence𝑎𝑡subscript𝑟𝑡 for all 𝑡01𝑝U=\{a\in\{0,1\}^{\omega}:a(t)=r_{t}\text{ for all }t=0,1,\ldots,p\}italic_U = { italic_a ∈ { 0 , 1 } start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT : italic_a ( italic_t ) = italic_r start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT for all italic_t = 0 , 1 , … , italic_p }. Set A:={t{0,1,,p}:rt=0}assign𝐴conditional-set𝑡01𝑝subscript𝑟𝑡0A:=\{t\in\{0,1,\ldots,p\}:r_{t}=0\}italic_A := { italic_t ∈ { 0 , 1 , … , italic_p } : italic_r start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT = 0 } and B:={t{0,1,,p}:rt=1}assign𝐵conditional-set𝑡01𝑝subscript𝑟𝑡1B:=\{t\in\{0,1,\ldots,p\}:r_{t}=1\}italic_B := { italic_t ∈ { 0 , 1 , … , italic_p } : italic_r start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT = 1 }. It follows that

ξ1[U]={ηX:ζm((nk),η)(t)=rt for all t=0,1,,p}=(tA{ηX:ζm((nk),η)(t)=0)(tB{ηX:ζm((nk),η)(t)=1)=(tA{ηX:xntUη,m)(tB{ηX:xntUη,m)=(tAηX:xntUη,mUη,m)(tBηX:xntUη,mUη,mc).\begin{split}\xi^{-1}&[U]=\{\eta\in X:\zeta_{m}((n_{k}),\eta)(t)=r_{t}\text{ % for all }t=0,1,\ldots,p\}\\ &=\left(\bigcap_{t\in A}\{\eta\in X:\zeta_{m}((n_{k}),\eta)(t)=0\right)\cap% \left(\bigcap_{t\in B}\{\eta\in X:\zeta_{m}((n_{k}),\eta)(t)=1\right)\\ &=\left(\bigcap_{t\in A}\{\eta\in X:x_{n_{t}}\in U_{\eta,m}\right)\cap\left(% \bigcap_{t\in B}\{\eta\in X:x_{n_{t}}\notin U_{\eta,m}\right)\\ &=\left(\bigcap_{t\in A}\,\bigcup_{\eta\in X:\,x_{n_{t}}\in U_{\eta,m}}U_{\eta% ,m}\right)\cap\left(\bigcap_{t\in B}\,\bigcap_{\eta\in X:\,x_{n_{t}}\notin U_{% \eta,m}}U_{\eta,m}^{c}\right).\end{split}start_ROW start_CELL italic_ξ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT end_CELL start_CELL [ italic_U ] = { italic_η ∈ italic_X : italic_ζ start_POSTSUBSCRIPT italic_m end_POSTSUBSCRIPT ( ( italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ) , italic_η ) ( italic_t ) = italic_r start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT for all italic_t = 0 , 1 , … , italic_p } end_CELL end_ROW start_ROW start_CELL end_CELL start_CELL = ( ⋂ start_POSTSUBSCRIPT italic_t ∈ italic_A end_POSTSUBSCRIPT { italic_η ∈ italic_X : italic_ζ start_POSTSUBSCRIPT italic_m end_POSTSUBSCRIPT ( ( italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ) , italic_η ) ( italic_t ) = 0 ) ∩ ( ⋂ start_POSTSUBSCRIPT italic_t ∈ italic_B end_POSTSUBSCRIPT { italic_η ∈ italic_X : italic_ζ start_POSTSUBSCRIPT italic_m end_POSTSUBSCRIPT ( ( italic_n start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ) , italic_η ) ( italic_t ) = 1 ) end_CELL end_ROW start_ROW start_CELL end_CELL start_CELL = ( ⋂ start_POSTSUBSCRIPT italic_t ∈ italic_A end_POSTSUBSCRIPT { italic_η ∈ italic_X : italic_x start_POSTSUBSCRIPT italic_n start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT end_POSTSUBSCRIPT ∈ italic_U start_POSTSUBSCRIPT italic_η , italic_m end_POSTSUBSCRIPT ) ∩ ( ⋂ start_POSTSUBSCRIPT italic_t ∈ italic_B end_POSTSUBSCRIPT { italic_η ∈ italic_X : italic_x start_POSTSUBSCRIPT italic_n start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT end_POSTSUBSCRIPT ∉ italic_U start_POSTSUBSCRIPT italic_η , italic_m end_POSTSUBSCRIPT ) end_CELL end_ROW start_ROW start_CELL end_CELL start_CELL = ( ⋂ start_POSTSUBSCRIPT italic_t ∈ italic_A end_POSTSUBSCRIPT ⋃ start_POSTSUBSCRIPT italic_η ∈ italic_X : italic_x start_POSTSUBSCRIPT italic_n start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT end_POSTSUBSCRIPT ∈ italic_U start_POSTSUBSCRIPT italic_η , italic_m end_POSTSUBSCRIPT end_POSTSUBSCRIPT italic_U start_POSTSUBSCRIPT italic_η , italic_m end_POSTSUBSCRIPT ) ∩ ( ⋂ start_POSTSUBSCRIPT italic_t ∈ italic_B end_POSTSUBSCRIPT ⋂ start_POSTSUBSCRIPT italic_η ∈ italic_X : italic_x start_POSTSUBSCRIPT italic_n start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT end_POSTSUBSCRIPT ∉ italic_U start_POSTSUBSCRIPT italic_η , italic_m end_POSTSUBSCRIPT end_POSTSUBSCRIPT italic_U start_POSTSUBSCRIPT italic_η , italic_m end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT ) . end_CELL end_ROW

Since each Uη,msubscript𝑈𝜂𝑚U_{\eta,m}italic_U start_POSTSUBSCRIPT italic_η , italic_m end_POSTSUBSCRIPT is open, we conclude that ξ1[U]superscript𝜉1delimited-[]𝑈\xi^{-1}[U]italic_ξ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_U ] is a Borel set. ∎

Claim 4.

For each mω𝑚𝜔m\in\omegaitalic_m ∈ italic_ω, the set ζm1[Fin]superscriptsubscript𝜁𝑚1delimited-[]Fin\zeta_{m}^{-1}[\mathrm{Fin}]italic_ζ start_POSTSUBSCRIPT italic_m end_POSTSUBSCRIPT start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ roman_Fin ] is Borel.

Proof.

Thanks to Claim 2, Claim 3, and [31, Theorem 3.1.30], each map ζmsubscript𝜁𝑚\zeta_{m}italic_ζ start_POSTSUBSCRIPT italic_m end_POSTSUBSCRIPT is Borel measurable. The conclusion follows since FinFin\mathrm{Fin}roman_Fin is a Fσsubscript𝐹𝜎F_{\sigma}italic_F start_POSTSUBSCRIPT italic_σ end_POSTSUBSCRIPT-set. ∎

To conclude the proof of Proposition 3.3, we have by Claim 1 and Claim 4 that both ψ1[+]×Xsuperscript𝜓1delimited-[]superscript𝑋\psi^{-1}[\,\mathcal{I}^{+}]\times Xitalic_ψ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ caligraphic_I start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT ] × italic_X and σm1[Fin]superscriptsubscript𝜎𝑚1delimited-[]Fin\sigma_{m}^{-1}[\mathrm{Fin}]italic_σ start_POSTSUBSCRIPT italic_m end_POSTSUBSCRIPT start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ roman_Fin ] are s-analytic subsets of 𝒩×X𝒩𝑋\mathcal{N}\times Xcaligraphic_N × italic_X, hence also their intersection by Lemma 3.7. By the identity (6), Λ𝒙()subscriptΛ𝒙\Lambda_{\bm{x}}(\mathcal{I})roman_Λ start_POSTSUBSCRIPT bold_italic_x end_POSTSUBSCRIPT ( caligraphic_I ) is the projection on X𝑋Xitalic_X of a s-analytic subset of 𝒩×X𝒩𝑋\mathcal{N}\times Xcaligraphic_N × italic_X, which is s-analytic by Lemma 3.6. ∎

4. Minimal ideals Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT and their complexities

In this last Section, following the line of research initiated in [16, 17], we recall the definition of certain ideals and we study their structural and topological complexity. For, given a sequence 𝒙𝒙\bm{x}bold_italic_x taking values in a first countable Hausdorff space X𝑋Xitalic_X and a subset WX𝑊𝑋W\subseteq Xitalic_W ⊆ italic_X, define the ideal

W:={Aω:L𝒙AW=},assignsubscript𝑊conditional-set𝐴𝜔subscriptL𝒙𝐴𝑊\mathcal{I}_{W}:=\{A\subseteq\omega:\mathrm{L}_{\bm{x}\upharpoonright A}\cap W% =\emptyset\},caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT := { italic_A ⊆ italic_ω : roman_L start_POSTSUBSCRIPT bold_italic_x ↾ italic_A end_POSTSUBSCRIPT ∩ italic_W = ∅ } ,

where L𝒙A:=Λ𝒙A(Fin)assignsubscriptL𝒙𝐴subscriptΛ𝒙𝐴Fin\mathrm{L}_{\bm{x}\upharpoonright A}:=\Lambda_{\bm{x}\upharpoonright A}(% \mathrm{Fin})roman_L start_POSTSUBSCRIPT bold_italic_x ↾ italic_A end_POSTSUBSCRIPT := roman_Λ start_POSTSUBSCRIPT bold_italic_x ↾ italic_A end_POSTSUBSCRIPT ( roman_Fin ), see [16, Section 2]. More explicitly, Aω𝐴𝜔A\subseteq\omegaitalic_A ⊆ italic_ω belongs to Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT if and only if A𝐴Aitalic_A is finite or, in the opposite, A𝐴Aitalic_A is infinite and there are no infinite subsets BA𝐵𝐴B\subseteq Aitalic_B ⊆ italic_A such that (xn:nB):subscript𝑥𝑛𝑛𝐵(x_{n}:n\in B)( italic_x start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT : italic_n ∈ italic_B ) is convergent to some element of W𝑊Witalic_W. Note that Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT may be not admissible: for instance, if W=𝑊W=\emptysetitalic_W = ∅ then W=𝒫(ω)subscript𝑊𝒫𝜔\mathcal{I}_{W}=\mathcal{P}(\omega)caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT = caligraphic_P ( italic_ω ).

The main reason of its introduction is the following:

Theorem 4.1.

Let 𝐱𝐱\bm{x}bold_italic_x be a sequence taking values in a first countable Hausdorff space X𝑋Xitalic_X and fix a subset WX𝑊𝑋W\subseteq Xitalic_W ⊆ italic_X. Then

Λ𝒙(W)=WL𝒙.subscriptΛ𝒙subscript𝑊𝑊subscriptL𝒙\Lambda_{\bm{x}}(\mathcal{I}_{W})=W\cap\mathrm{L}_{\bm{x}}.roman_Λ start_POSTSUBSCRIPT bold_italic_x end_POSTSUBSCRIPT ( caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT ) = italic_W ∩ roman_L start_POSTSUBSCRIPT bold_italic_x end_POSTSUBSCRIPT . (7)

In particular, if X𝑋Xitalic_X is separable, then W(W)𝑊subscript𝑊W\in\mathscr{L}(\mathcal{I}_{W})italic_W ∈ script_L ( caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT ).

Proof.

The first part follows by [16, Theorem 2.2(i)]. This implies, if WL𝒙𝑊subscriptL𝒙W\subseteq\mathrm{L}_{\bm{x}}italic_W ⊆ roman_L start_POSTSUBSCRIPT bold_italic_x end_POSTSUBSCRIPT, then Λ𝒙(W)=WsubscriptΛ𝒙subscript𝑊𝑊\Lambda_{\bm{x}}(\mathcal{I}_{W})=Wroman_Λ start_POSTSUBSCRIPT bold_italic_x end_POSTSUBSCRIPT ( caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT ) = italic_W, so that W(W)𝑊subscript𝑊W\in\mathscr{L}(\mathcal{I}_{W})italic_W ∈ script_L ( caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT ). Hence, the second part is obtained by choosing a sequence 𝒙𝒙\bm{x}bold_italic_x with dense image. ∎

Note that that Theorem 4.1 proves that, even if X=𝑋X=\mathbb{R}italic_X = blackboard_R, the family ()\mathscr{L}(\mathcal{I})script_L ( caligraphic_I ) may contain sets which are not Borel, which answered an open question in [2].

It is also worth to remark that ideals Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT are not the only ones for which (7) holds, see [16, Theorem 2.7]; cf. also [17, Theorems 3.4 and 3.5] for further refinements of Theorem 4.1 in regular and Polish spaces. However, we show that they are the smallest ideals with such property. For, we say that \mathcal{I}caligraphic_I contains an isomorphic copy of 𝒥𝒥\mathcal{J}caligraphic_J if there exists a bijection ϕitalic-ϕ\phiitalic_ϕ on ω𝜔\omegaitalic_ω such that ϕ[𝒥]italic-ϕdelimited-[]𝒥\phi[\,\mathcal{J}]\subseteq\mathcal{I}italic_ϕ [ caligraphic_J ] ⊆ caligraphic_I, where ϕ[𝒥]italic-ϕdelimited-[]𝒥\phi[\,\mathcal{J}]italic_ϕ [ caligraphic_J ] stands for the family {ϕ[S]:S𝒥}conditional-setitalic-ϕdelimited-[]𝑆𝑆𝒥\{\phi[S]:S\in\mathcal{J}\}{ italic_ϕ [ italic_S ] : italic_S ∈ caligraphic_J }.

Proposition 4.2.

Let 𝐱𝐱\bm{x}bold_italic_x be a sequence taking values in a compact metric space X𝑋Xitalic_X and fix a dense subset WX𝑊𝑋W\subseteq Xitalic_W ⊆ italic_X. Let also \mathcal{I}caligraphic_I be an ideal on ω𝜔\omegaitalic_ω such that W()𝑊W\in\mathscr{L}(\mathcal{I})italic_W ∈ script_L ( caligraphic_I ). Suppose also that 𝐱𝐱\bm{x}bold_italic_x has dense image, and that

A+,𝒚Xω,Λ𝒚A(A).formulae-sequencefor-all𝐴superscriptformulae-sequencefor-all𝒚superscript𝑋𝜔subscriptΛ𝒚𝐴𝐴\forall A\in\mathcal{I}^{+},\forall\bm{y}\in X^{\omega},\quad\Lambda_{\bm{y}% \upharpoonright A}(\,\mathcal{I}\upharpoonright A)\neq\emptyset.∀ italic_A ∈ caligraphic_I start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT , ∀ bold_italic_y ∈ italic_X start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT , roman_Λ start_POSTSUBSCRIPT bold_italic_y ↾ italic_A end_POSTSUBSCRIPT ( caligraphic_I ↾ italic_A ) ≠ ∅ . (8)

Then \mathcal{I}caligraphic_I contains an isomorphic copy of Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT.

Proof.

By hypothesis, there exists a sequence 𝒚𝒚\bm{y}bold_italic_y taking values in X𝑋Xitalic_X such that Λ𝒚()=WsubscriptΛ𝒚𝑊\Lambda_{\bm{y}}(\mathcal{I})=Wroman_Λ start_POSTSUBSCRIPT bold_italic_y end_POSTSUBSCRIPT ( caligraphic_I ) = italic_W. Let d𝑑ditalic_d be a compatible metric on X𝑋Xitalic_X and note that the denseness of W𝑊Witalic_W implies that L𝒚=XsubscriptL𝒚𝑋\mathrm{L}_{\bm{y}}=Xroman_L start_POSTSUBSCRIPT bold_italic_y end_POSTSUBSCRIPT = italic_X. Since X𝑋Xitalic_X is compact, there exists a strictly increasing sequence of integers (ιn:nω):subscript𝜄𝑛𝑛𝜔(\iota_{n}:n\in\omega)( italic_ι start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT : italic_n ∈ italic_ω ) such that ι0:=0assignsubscript𝜄00\iota_{0}:=0italic_ι start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT := 0 and, for each nω𝑛𝜔n\in\omegaitalic_n ∈ italic_ω, the family of open balls {B(xk,2n):k[ιn,ιn+1)}conditional-set𝐵subscript𝑥𝑘superscript2𝑛𝑘subscript𝜄𝑛subscript𝜄𝑛1\{B(x_{k},2^{-n}):k\in[\iota_{n},\iota_{n+1})\}{ italic_B ( italic_x start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT , 2 start_POSTSUPERSCRIPT - italic_n end_POSTSUPERSCRIPT ) : italic_k ∈ [ italic_ι start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT , italic_ι start_POSTSUBSCRIPT italic_n + 1 end_POSTSUBSCRIPT ) } is an open cover of X𝑋Xitalic_X.

At this point, define recursively the map ϕ:ωω:italic-ϕ𝜔𝜔\phi:\omega\to\omegaitalic_ϕ : italic_ω → italic_ω as it follows:

  1. (i)

    ϕ(0)italic-ϕ0\phi(0)italic_ϕ ( 0 ) is the smallest integer kω𝑘𝜔k\in\omegaitalic_k ∈ italic_ω such that d(yϕ(k),x0)<1𝑑subscript𝑦italic-ϕ𝑘subscript𝑥01d(y_{\phi(k)},x_{0})<1italic_d ( italic_y start_POSTSUBSCRIPT italic_ϕ ( italic_k ) end_POSTSUBSCRIPT , italic_x start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT ) < 1;

  2. (ii)

    for each integer m(1,ι1)𝑚1subscript𝜄1m\in(1,\iota_{1})italic_m ∈ ( 1 , italic_ι start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ), ϕ(m)italic-ϕ𝑚\phi(m)italic_ϕ ( italic_m ) is the smallest integer kω𝑘𝜔k\in\omegaitalic_k ∈ italic_ω such that d(yϕ(k),xm)<1𝑑subscript𝑦italic-ϕ𝑘subscript𝑥𝑚1d(y_{\phi(k)},x_{m})<1italic_d ( italic_y start_POSTSUBSCRIPT italic_ϕ ( italic_k ) end_POSTSUBSCRIPT , italic_x start_POSTSUBSCRIPT italic_m end_POSTSUBSCRIPT ) < 1 and kϕ[{0,1,,m1}]𝑘italic-ϕdelimited-[]01𝑚1k\notin\phi[\{0,1,\ldots,m-1\}]italic_k ∉ italic_ϕ [ { 0 , 1 , … , italic_m - 1 } ];

  3. (iii)

    for each n,mω𝑛𝑚𝜔n,m\in\omegaitalic_n , italic_m ∈ italic_ω with n1𝑛1n\geq 1italic_n ≥ 1 and m[ιn,ιn+1)𝑚subscript𝜄𝑛subscript𝜄𝑛1m\in[\iota_{n},\iota_{n+1})italic_m ∈ [ italic_ι start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT , italic_ι start_POSTSUBSCRIPT italic_n + 1 end_POSTSUBSCRIPT ), ϕ(m)italic-ϕ𝑚\phi(m)italic_ϕ ( italic_m ) is the smallest integer kω𝑘𝜔k\in\omegaitalic_k ∈ italic_ω such that d(yϕ(k),xm)<2n𝑑subscript𝑦italic-ϕ𝑘subscript𝑥𝑚superscript2𝑛d(y_{\phi(k)},x_{m})<2^{-n}italic_d ( italic_y start_POSTSUBSCRIPT italic_ϕ ( italic_k ) end_POSTSUBSCRIPT , italic_x start_POSTSUBSCRIPT italic_m end_POSTSUBSCRIPT ) < 2 start_POSTSUPERSCRIPT - italic_n end_POSTSUPERSCRIPT and kϕ[{0,1,,m1}]𝑘italic-ϕdelimited-[]01𝑚1k\notin\phi[\{0,1,\ldots,m-1\}]italic_k ∉ italic_ϕ [ { 0 , 1 , … , italic_m - 1 } ].

It follows by construction that ϕitalic-ϕ\phiitalic_ϕ is a bijection on ω𝜔\omegaitalic_ω which satisfies

limnd(yϕ(n),xn)=0.subscript𝑛𝑑subscript𝑦italic-ϕ𝑛subscript𝑥𝑛0\lim_{n\to\infty}d(y_{\phi(n)},x_{n})=0.roman_lim start_POSTSUBSCRIPT italic_n → ∞ end_POSTSUBSCRIPT italic_d ( italic_y start_POSTSUBSCRIPT italic_ϕ ( italic_n ) end_POSTSUBSCRIPT , italic_x start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT ) = 0 . (9)

Fix A+𝐴superscriptA\in\mathcal{I}^{+}italic_A ∈ caligraphic_I start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT and note that the set of A𝐴\mathcal{I}\upharpoonright Acaligraphic_I ↾ italic_A-limit points of 𝒚A𝒚𝐴\bm{y}\upharpoonright Abold_italic_y ↾ italic_A is contained both in L𝒚AsubscriptL𝒚𝐴\mathrm{L}_{\bm{y}\upharpoonright A}roman_L start_POSTSUBSCRIPT bold_italic_y ↾ italic_A end_POSTSUBSCRIPT and in Λ𝒚()subscriptΛ𝒚\Lambda_{\bm{y}}(\mathcal{I})roman_Λ start_POSTSUBSCRIPT bold_italic_y end_POSTSUBSCRIPT ( caligraphic_I ). Since the latter is equal to W𝑊Witalic_W, we obtain by (8) and (9) that

Λ𝒚A(A)L𝒚AW=L𝒙ϕ1[A]W.subscriptΛ𝒚𝐴𝐴subscriptL𝒚𝐴𝑊subscriptL𝒙superscriptitalic-ϕ1delimited-[]𝐴𝑊\emptyset\neq\Lambda_{\bm{y}\upharpoonright A}(\,\mathcal{I}\upharpoonright A)% \subseteq\mathrm{L}_{\bm{y}\upharpoonright A}\cap W=\mathrm{L}_{\bm{x}% \upharpoonright\phi^{-1}[A]}\cap W.∅ ≠ roman_Λ start_POSTSUBSCRIPT bold_italic_y ↾ italic_A end_POSTSUBSCRIPT ( caligraphic_I ↾ italic_A ) ⊆ roman_L start_POSTSUBSCRIPT bold_italic_y ↾ italic_A end_POSTSUBSCRIPT ∩ italic_W = roman_L start_POSTSUBSCRIPT bold_italic_x ↾ italic_ϕ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_A ] end_POSTSUBSCRIPT ∩ italic_W .

By the definition of Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT, we get ϕ1[A]Wsuperscriptitalic-ϕ1delimited-[]𝐴subscript𝑊\phi^{-1}[A]\notin\mathcal{I}_{W}italic_ϕ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_A ] ∉ caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT. Therefore ϕ[W]italic-ϕdelimited-[]subscript𝑊\phi[\,\mathcal{I}_{W}]\subseteq\mathcal{I}italic_ϕ [ caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT ] ⊆ caligraphic_I. ∎

As remarked in the Introduction, the technical condition (8) has been already studied in the literature. For instance, if X𝑋Xitalic_X is compact, all Fσsubscript𝐹𝜎F_{\sigma}italic_F start_POSTSUBSCRIPT italic_σ end_POSTSUBSCRIPT-ideals \mathcal{I}caligraphic_I satisfy (8), taking into account [2, Theorem 2.3] and [27, Lemma 3.1(vi)].

The following result, due to He et al. [16], deals with the topological complexity of ideals Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT. Hereafter, we assume for simplicity that X𝑋Xitalic_X is the Cantor space and 𝐱𝐱\bm{x}bold_italic_x is an enumeration of the rationals (i.e., finitely supported sequences) of X𝑋Xitalic_X, with x0:=(0,0,)assignsubscript𝑥000x_{0}:=(0,0,\ldots)italic_x start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT := ( 0 , 0 , … ).

Theorem 4.3.

Let W𝑊Witalic_W be a subset of the Cantor space X={0,1}ω𝑋superscript01𝜔X=\{0,1\}^{\omega}italic_X = { 0 , 1 } start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT. Then:

  1. (i)

    if W𝑊Witalic_W is closed then Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT is a Σ20subscriptsuperscriptΣ02\Sigma^{0}_{2}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT-ideal;

  2. (ii)

    if W𝑊Witalic_W is Σ20subscriptsuperscriptΣ02\Sigma^{0}_{2}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT then Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT is a Π30subscriptsuperscriptΠ03\Pi^{0}_{3}roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT-ideal;

  3. (iii)

    if W𝑊Witalic_W is open then Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT is an analytic P𝑃Pitalic_P-ideal.

Proof.

It follows by [16, Theorem 2.2(ii)] by choosing a sequence 𝒙𝒙\bm{x}bold_italic_x with dense image (the case W=𝑊W=\emptysetitalic_W = ∅ holds as well). ∎

Taking into account also the results obtained in Section 2, we prove some characterizations for the [non]closedness of the subset W𝑊Witalic_W:

Theorem 4.4.

Let W𝑊Witalic_W be a nonempty subset of the Cantor space X={0,1}ω𝑋superscript01𝜔X=\{0,1\}^{\omega}italic_X = { 0 , 1 } start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT such that Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT has the hereditary Baire property. Then the following are equivalent:

  1. (i)

    W𝑊Witalic_W is not closed;

  2. (ii)

    ×FinRBWsubscriptRBFinsubscript𝑊\emptyset\times\mathrm{Fin}\leq_{\mathrm{RB}}\mathcal{I}_{W}∅ × roman_Fin ≤ start_POSTSUBSCRIPT roman_RB end_POSTSUBSCRIPT caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT;

  3. (iii)

    Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT is Π30subscriptsuperscriptΠ03\Pi^{0}_{3}roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT-hard;

  4. (iv)

    Σ20(W)subscriptsuperscriptΣ02subscript𝑊\Sigma^{0}_{2}\subseteq\mathscr{L}(\mathcal{I}_{W})roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ⊆ script_L ( caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT ).

Proof.

(i) \implies (ii). Thanks to Theorem 4.1, W(W)𝑊subscript𝑊W\in\mathscr{L}(\mathcal{I}_{W})italic_W ∈ script_L ( caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT ), hence the inclusion (W)Π10subscript𝑊subscriptsuperscriptΠ01\mathscr{L}(\mathcal{I}_{W})\subseteq\Pi^{0}_{1}script_L ( caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT ) ⊆ roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT fails. The claim follows by Lemma 2.8.

(ii) \implies (iii). It follows by Lemma 2.10.

(ii) \implies (iv). It follows by Corollary 2.9.

(iii) \implies (i). If W𝑊Witalic_W is closed, then Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT would be a Σ20subscriptsuperscriptΣ02\Sigma^{0}_{2}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT-ideal by Theorem 4.3.

(iv) \implies (i). As in the previous implication, if W𝑊Witalic_W is closed, then Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT would be Σ20subscriptsuperscriptΣ02\Sigma^{0}_{2}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT. Hence (W)=Π10subscript𝑊subscriptsuperscriptΠ01\mathscr{L}(\mathcal{I}_{W})=\Pi^{0}_{1}script_L ( caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT ) = roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT by Theorem 1.2. However, since X𝑋Xitalic_X is separable, there exists a countable dense subset, which is Σ20subscriptsuperscriptΣ02\Sigma^{0}_{2}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT and not closed. ∎

On the same lines, we are able to characterize the openness of the set W𝑊Witalic_W and the P𝑃Pitalic_P-property of the ideal Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT, improving Theorem 4.3(iii):

Theorem 4.5.

Let W𝑊Witalic_W be a subset of the Cantor space X={0,1}ω𝑋superscript01𝜔X=\{0,1\}^{\omega}italic_X = { 0 , 1 } start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT. Then the following are equivalent:

  1. (i)

    W𝑊Witalic_W is open;

  2. (ii)

    Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT is an analytic P𝑃Pitalic_P-ideal;

  3. (iii)

    Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT is a P𝑃Pitalic_P-ideal.

Proof.

(i) \implies (ii). It follows by Theorem 4.3(iii).

(ii) \implies (iii). This is obvious.

(iii) \implies (i). Let us suppose that W𝑊Witalic_W is not open, i.e., Wcsuperscript𝑊𝑐W^{c}italic_W start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT is not closed. Since X𝑋Xitalic_X is metrizable, Wcsuperscript𝑊𝑐W^{c}italic_W start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT is not sequentially closed. Hence it is possible to pick a sequence 𝒚𝒚\bm{y}bold_italic_y taking values in Wcsuperscript𝑊𝑐W^{c}italic_W start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT which is convergent to some ηW𝜂𝑊\eta\in Witalic_η ∈ italic_W. Let us denote by d𝑑ditalic_d a compatible metric on X𝑋Xitalic_X, so that limkd(yk,η)=0subscript𝑘𝑑subscript𝑦𝑘𝜂0\lim_{k}d(y_{k},\eta)=0roman_lim start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT italic_d ( italic_y start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT , italic_η ) = 0. Since 𝒙𝒙\bm{x}bold_italic_x is dense, for each kω𝑘𝜔k\in\omegaitalic_k ∈ italic_ω there exists an infinite set Akωsubscript𝐴𝑘𝜔A_{k}\subseteq\omegaitalic_A start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ⊆ italic_ω such that lim𝒙Ak=yk𝒙subscript𝐴𝑘subscript𝑦𝑘\lim\bm{x}\upharpoonright A_{k}=y_{k}roman_lim bold_italic_x ↾ italic_A start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT = italic_y start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT. Considering that ykWcsubscript𝑦𝑘superscript𝑊𝑐y_{k}\in W^{c}italic_y start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ∈ italic_W start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT, it follows that L𝒙AkW=subscriptL𝒙subscript𝐴𝑘𝑊\mathrm{L}_{\bm{x}\upharpoonright A_{k}}\cap W=\emptysetroman_L start_POSTSUBSCRIPT bold_italic_x ↾ italic_A start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT end_POSTSUBSCRIPT ∩ italic_W = ∅, hence AkWsubscript𝐴𝑘subscript𝑊A_{k}\in\mathcal{I}_{W}italic_A start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ∈ caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT.

At this point, pick a set Aω𝐴𝜔A\subseteq\omegaitalic_A ⊆ italic_ω such that AkAFinsubscript𝐴𝑘𝐴FinA_{k}\setminus A\in\mathrm{Fin}italic_A start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ∖ italic_A ∈ roman_Fin for all kω𝑘𝜔k\in\omegaitalic_k ∈ italic_ω. Set for convenience b1:=0assignsubscript𝑏10b_{-1}:=0italic_b start_POSTSUBSCRIPT - 1 end_POSTSUBSCRIPT := 0 and define recursively a sequence of integers (bk:kω):subscript𝑏𝑘𝑘𝜔(b_{k}:k\in\omega)( italic_b start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT : italic_k ∈ italic_ω ) such that bksubscript𝑏𝑘b_{k}italic_b start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT is the smallest element of AAk𝐴subscript𝐴𝑘A\cap A_{k}italic_A ∩ italic_A start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT for which d(yk,xbk)d(yk,η)𝑑subscript𝑦𝑘subscript𝑥subscript𝑏𝑘𝑑subscript𝑦𝑘𝜂d(y_{k},x_{b_{k}})\leq d(y_{k},\eta)italic_d ( italic_y start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT , italic_x start_POSTSUBSCRIPT italic_b start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT end_POSTSUBSCRIPT ) ≤ italic_d ( italic_y start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT , italic_η ) and bk>bk1subscript𝑏𝑘subscript𝑏𝑘1b_{k}>b_{k-1}italic_b start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT > italic_b start_POSTSUBSCRIPT italic_k - 1 end_POSTSUBSCRIPT (note that this is well defined). It follows by the triangle inequality that

kω,d(xbk,η)d(xbk,yk)+d(yk,η)2d(yk,η).formulae-sequencefor-all𝑘𝜔𝑑subscript𝑥subscript𝑏𝑘𝜂𝑑subscript𝑥subscript𝑏𝑘subscript𝑦𝑘𝑑subscript𝑦𝑘𝜂2𝑑subscript𝑦𝑘𝜂\forall k\in\omega,\quad d(x_{b_{k}},\eta)\leq d(x_{b_{k}},y_{k})+d(y_{k},\eta% )\leq 2d(y_{k},\eta).∀ italic_k ∈ italic_ω , italic_d ( italic_x start_POSTSUBSCRIPT italic_b start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT end_POSTSUBSCRIPT , italic_η ) ≤ italic_d ( italic_x start_POSTSUBSCRIPT italic_b start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT end_POSTSUBSCRIPT , italic_y start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ) + italic_d ( italic_y start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT , italic_η ) ≤ 2 italic_d ( italic_y start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT , italic_η ) .

Therefore B:={bk:kω}assign𝐵conditional-setsubscript𝑏𝑘𝑘𝜔B:=\{b_{k}:k\in\omega\}italic_B := { italic_b start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT : italic_k ∈ italic_ω } is an infinite subset of A𝐴Aitalic_A such that lim𝒙B=η𝒙𝐵𝜂\lim\bm{x}\upharpoonright B=\etaroman_lim bold_italic_x ↾ italic_B = italic_η. Since ηW𝜂𝑊\eta\in Witalic_η ∈ italic_W, this proves that L𝒙AWsubscriptL𝒙𝐴𝑊\mathrm{L}_{\bm{x}\upharpoonright A}\cap W\neq\emptysetroman_L start_POSTSUBSCRIPT bold_italic_x ↾ italic_A end_POSTSUBSCRIPT ∩ italic_W ≠ ∅. Hence, by the definition of Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT, we conclude that AW𝐴subscript𝑊A\notin\mathcal{I}_{W}italic_A ∉ caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT. To sum up, (Ak:kω):subscript𝐴𝑘𝑘𝜔(A_{k}:k\in\omega)( italic_A start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT : italic_k ∈ italic_ω ) is a witnessing sequence of sets in Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT which fails the P𝑃Pitalic_P-property for the ideal Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT. ∎

As a consequence, we obtain that there are only three possibilities for the complexity of ideals Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT:

Corollary 4.6.

Let W𝑊Witalic_W be a subset of the Cantor space X={0,1}ω𝑋superscript01𝜔X=\{0,1\}^{\omega}italic_X = { 0 , 1 } start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT. Then exactly one of the following cases occurs:

  1. (i)

    Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT is a Σ20subscriptsuperscriptΣ02\Sigma^{0}_{2}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT-ideal;

  2. (ii)

    Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT is a Borel ideal which is not Σ30subscriptsuperscriptΣ03\Sigma^{0}_{3}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT;

  3. (iii)

    Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT is not a Borel ideal.

Proof.

First, if W=𝑊W=\emptysetitalic_W = ∅, then W=𝒫(ω)subscript𝑊𝒫𝜔\mathcal{I}_{W}=\mathcal{P}(\omega)caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT = caligraphic_P ( italic_ω ), which satisfies only (i). Hence suppose hereafter that W𝑊Witalic_W is nonempty. Suppose also that Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT is Borel, so that (iii) fails and Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT has the hereditary Baire property. If W𝑊Witalic_W is closed then Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT is Σ20subscriptsuperscriptΣ02\Sigma^{0}_{2}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT by Theorem 4.3, hence (i) holds. If W𝑊Witalic_W is not closed then W𝑊Witalic_W is Π30subscriptsuperscriptΠ03\Pi^{0}_{3}roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT-hard by Theorem 4.4, which is equivalent to be not Σ30subscriptsuperscriptΣ03\Sigma^{0}_{3}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT by [25, Theorem 22.10], hence (ii) holds. ∎

In light of Theorem 4.3, one may ask whether the third case in Corollary 4.6 really occurs. We answer in the affirmative with our last main result:

Theorem 4.7.

Let W𝑊Witalic_W be a Borel subset of the Cantor space X={0,1}ω𝑋superscript01𝜔X=\{0,1\}^{\omega}italic_X = { 0 , 1 } start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT which is not Σ20subscriptsuperscriptΣ02\Sigma^{0}_{2}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT. Then Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT is not analytic, hence not Borel.

Putting together the above results, we have the following consequence:

Corollary 4.8.

Let W𝑊Witalic_W be a Borel subset of the Cantor space X={0,1}ω𝑋superscript01𝜔X=\{0,1\}^{\omega}italic_X = { 0 , 1 } start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT. Then the following are equivalent:

  1. (i)

    W𝑊Witalic_W is Σ20subscriptsuperscriptΣ02\Sigma^{0}_{2}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT;

  2. (ii)

    Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT is a Π30subscriptsuperscriptΠ03\Pi^{0}_{3}roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT-ideal;

  3. (iii)

    Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT is a Borel ideal;

  4. (iv)

    Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT is an analytic ideal.

Proof.

(i) \implies (ii). It follows by Theorem 4.3(i).

(ii) \implies (iii) \implies (iv). They are obvious.

(iv) \implies (i). It follows by Theorem 4.7. ∎

We recall now some definition about trees, which will be needed in the proof of Theorem 4.7.

Denote by {0,1}<ωsuperscript01absent𝜔\{0,1\}^{<\omega}{ 0 , 1 } start_POSTSUPERSCRIPT < italic_ω end_POSTSUPERSCRIPT the set of finite {0,1}01\{0,1\}{ 0 , 1 }-sequences. We say that a=(a0,,an){0,1}<ω𝑎subscript𝑎0subscript𝑎𝑛superscript01absent𝜔a=(a_{0},\ldots,a_{n})\in\{0,1\}^{<\omega}italic_a = ( italic_a start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , … , italic_a start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT ) ∈ { 0 , 1 } start_POSTSUPERSCRIPT < italic_ω end_POSTSUPERSCRIPT is an extension of b=(b0,,bm){0,1}<ω𝑏subscript𝑏0subscript𝑏𝑚superscript01absent𝜔b=(b_{0},\ldots,b_{m})\in\{0,1\}^{<\omega}italic_b = ( italic_b start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , … , italic_b start_POSTSUBSCRIPT italic_m end_POSTSUBSCRIPT ) ∈ { 0 , 1 } start_POSTSUPERSCRIPT < italic_ω end_POSTSUPERSCRIPT, shortened as ab𝑎𝑏a\subseteq bitalic_a ⊆ italic_b, if nm𝑛𝑚n\leq mitalic_n ≤ italic_m and ak=bksubscript𝑎𝑘subscript𝑏𝑘a_{k}=b_{k}italic_a start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT = italic_b start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT for all kn𝑘𝑛k\leq nitalic_k ≤ italic_n. Given an infinite sequence z{0,1}ω𝑧superscript01𝜔z\in\{0,1\}^{\omega}italic_z ∈ { 0 , 1 } start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT, we define zn:=(z0,,zn){0,1}<ω𝑧𝑛assignsubscript𝑧0subscript𝑧𝑛superscript01absent𝜔z\upharpoonright n:=(z_{0},\ldots,z_{n})\in\{0,1\}^{<\omega}italic_z ↾ italic_n := ( italic_z start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , … , italic_z start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT ) ∈ { 0 , 1 } start_POSTSUPERSCRIPT < italic_ω end_POSTSUPERSCRIPT for each nω𝑛𝜔n\in\omegaitalic_n ∈ italic_ω. A tree on {0,1}01\{0,1\}{ 0 , 1 } is a subset T{0,1}<ω𝑇superscript01absent𝜔T\subseteq\{0,1\}^{<\omega}italic_T ⊆ { 0 , 1 } start_POSTSUPERSCRIPT < italic_ω end_POSTSUPERSCRIPT with the property that, aT𝑎𝑇a\in Titalic_a ∈ italic_T for all a,b{0,1}<ω𝑎𝑏superscript01absent𝜔a,b\in\{0,1\}^{<\omega}italic_a , italic_b ∈ { 0 , 1 } start_POSTSUPERSCRIPT < italic_ω end_POSTSUPERSCRIPT such that abT𝑎𝑏𝑇a\subseteq b\in Titalic_a ⊆ italic_b ∈ italic_T. The body of a tree T𝑇Titalic_T on {0,1}01\{0,1\}{ 0 , 1 }, denoted by [T]delimited-[]𝑇[T][ italic_T ], is the set of all its infinite branches, that is, the set of all sequences z{0,1}ω𝑧superscript01𝜔z\in\{0,1\}^{\omega}italic_z ∈ { 0 , 1 } start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT such that znT𝑧𝑛𝑇z\upharpoonright n\in Titalic_z ↾ italic_n ∈ italic_T for all nω𝑛𝜔n\in\omegaitalic_n ∈ italic_ω.

A tree T𝑇Titalic_T on {0,1}01\{0,1\}{ 0 , 1 } is said to be pruned if every aT𝑎𝑇a\in Titalic_a ∈ italic_T has a proper extension, that is, for all aT𝑎𝑇a\in Titalic_a ∈ italic_T there exists bT𝑏𝑇b\in Titalic_b ∈ italic_T such that ab𝑎𝑏a\neq bitalic_a ≠ italic_b and ab𝑎𝑏a\subseteq bitalic_a ⊆ italic_b, see [25, Definition 2.1]. Accordingly, we define

PTr2:={T{0,1}<ω:T is a pruned tree }.assignsubscriptPTr2conditional-set𝑇superscript01absent𝜔𝑇 is a pruned tree \mathrm{PTr}_{2}:=\{T\subseteq\{0,1\}^{<\omega}:T\text{ is a pruned tree }\}.roman_PTr start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT := { italic_T ⊆ { 0 , 1 } start_POSTSUPERSCRIPT < italic_ω end_POSTSUPERSCRIPT : italic_T is a pruned tree } .

Identifying a pruned tree with its characteristic function, the set PTr2subscriptPTr2\mathrm{PTr}_{2}roman_PTr start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT can be regarded as a closed subset of the Polish space {0,1}{0,1}<ωsuperscript01superscript01absent𝜔\{0,1\}^{\{0,1\}^{<\omega}}{ 0 , 1 } start_POSTSUPERSCRIPT { 0 , 1 } start_POSTSUPERSCRIPT < italic_ω end_POSTSUPERSCRIPT end_POSTSUPERSCRIPT, see [25, Exercise 4.32]. Thanks to Alexandrov’s theorem [25, Theorem 3.11], PTr2subscriptPTr2\mathrm{PTr}_{2}roman_PTr start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT is a Polish space.

With the above premises, let WF2subscriptsuperscriptWF2\mathrm{WF}^{\star}_{2}roman_WF start_POSTSUPERSCRIPT ⋆ end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT be the set of pruned trees on {0,1}01\{0,1\}{ 0 , 1 } for which every infinite branch contains finitely many ones, that is,

WF2:={TPTr2:z[T],n,zn=0},assignsubscriptsuperscriptWF2conditional-set𝑇subscriptPTr2formulae-sequencefor-all𝑧delimited-[]𝑇superscriptfor-all𝑛subscript𝑧𝑛0\mathrm{WF}^{\star}_{2}:=\left\{T\in\mathrm{PTr}_{2}:\forall z\in[T],\forall^{% \infty}n,\,\,z_{n}=0\,\right\},roman_WF start_POSTSUPERSCRIPT ⋆ end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT := { italic_T ∈ roman_PTr start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT : ∀ italic_z ∈ [ italic_T ] , ∀ start_POSTSUPERSCRIPT ∞ end_POSTSUPERSCRIPT italic_n , italic_z start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT = 0 } ,

see [25, Section 33.A]. It is known that WF2subscriptsuperscriptWF2\mathrm{WF}^{\star}_{2}roman_WF start_POSTSUPERSCRIPT ⋆ end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT is a Π11subscriptsuperscriptΠ11\Pi^{1}_{1}roman_Π start_POSTSUPERSCRIPT 1 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT-complete subset of PTr2subscriptPTr2\mathrm{PTr}_{2}roman_PTr start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT, hence it is conanalytic but not analytic, see [25, Exercise 27.3].

More generally, given a nonempty subset W𝑊Witalic_W of the Cantor space {0,1}ωsuperscript01𝜔\{0,1\}^{\omega}{ 0 , 1 } start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT, let 𝒯(W)𝒯𝑊\mathcal{T}(W)caligraphic_T ( italic_W ) be the set of pruned trees T𝑇Titalic_T on {0,1}01\{0,1\}{ 0 , 1 } which contains an infinite branch in W𝑊Witalic_W, that is,

𝒯(W):={TPTr2:[T]W}.assign𝒯𝑊conditional-set𝑇subscriptPTr2delimited-[]𝑇𝑊\mathcal{T}(W):=\left\{T\in\mathrm{PTr}_{2}:[T]\cap W\neq\emptyset\right\}.caligraphic_T ( italic_W ) := { italic_T ∈ roman_PTr start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT : [ italic_T ] ∩ italic_W ≠ ∅ } .

It is immediate to check that 𝒯(Wirr)𝒯subscript𝑊irr\mathcal{T}(W_{\mathrm{irr}})caligraphic_T ( italic_W start_POSTSUBSCRIPT roman_irr end_POSTSUBSCRIPT ) coincides with WF2subscriptsuperscriptWF2\mathrm{WF}^{\star}_{2}roman_WF start_POSTSUPERSCRIPT ⋆ end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT, where Wirrsubscript𝑊irrW_{\mathrm{irr}}italic_W start_POSTSUBSCRIPT roman_irr end_POSTSUBSCRIPT is the set of irrationals {z{0,1}ω:n,zn=1}conditional-set𝑧superscript01𝜔superscript𝑛subscript𝑧𝑛1\{z\in\{0,1\}^{\omega}:\exists^{\infty}n,\,z_{n}=1\}{ italic_z ∈ { 0 , 1 } start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT : ∃ start_POSTSUPERSCRIPT ∞ end_POSTSUPERSCRIPT italic_n , italic_z start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT = 1 }.

With the above premises, we are ready for the proof of Theorem 4.7.

Proof of Theorem 4.7.

We divide the proof in some intermediate steps. Hereafter, as in the statement of the result, W𝑊Witalic_W is a given Borel subset of the Cantor space which is not Σ20subscriptsuperscriptΣ02\Sigma^{0}_{2}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT.

Claim 5.

Fix nonempty subsets A,B{0,1}ω𝐴𝐵superscript01𝜔A,B\subseteq\{0,1\}^{\omega}italic_A , italic_B ⊆ { 0 , 1 } start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT such that AWBsubscriptW𝐴𝐵A\leq_{\mathrm{W}}Bitalic_A ≤ start_POSTSUBSCRIPT roman_W end_POSTSUBSCRIPT italic_B. Then

𝒯(A)W𝒯(B).subscriptW𝒯𝐴𝒯𝐵\mathcal{T}(A)\leq_{\mathrm{W}}\mathcal{T}(B).caligraphic_T ( italic_A ) ≤ start_POSTSUBSCRIPT roman_W end_POSTSUBSCRIPT caligraphic_T ( italic_B ) . (10)
Proof.

Since A𝐴Aitalic_A is Wadge reducible to B𝐵Bitalic_B, there exists a continuous map ϕ:{0,1}ω{0,1}ω:italic-ϕsuperscript01𝜔superscript01𝜔\phi:\{0,1\}^{\omega}\to\{0,1\}^{\omega}italic_ϕ : { 0 , 1 } start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT → { 0 , 1 } start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT such that ϕ1[B]=Asuperscriptitalic-ϕ1delimited-[]𝐵𝐴\phi^{-1}[B]=Aitalic_ϕ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ italic_B ] = italic_A, i.e., zA𝑧𝐴z\in Aitalic_z ∈ italic_A if and only if ϕ(z)Bitalic-ϕ𝑧𝐵\phi(z)\in Bitalic_ϕ ( italic_z ) ∈ italic_B. Let 𝒦𝒦\mathcal{K}caligraphic_K be the family of compact subsets of {0,1}ωsuperscript01𝜔\{0,1\}^{\omega}{ 0 , 1 } start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT, endowed with the Vietoris topology. It follows by [25, Exercise 4.32] that the map ψ:PTr2𝒦:𝜓subscriptPTr2𝒦\psi:\mathrm{PTr}_{2}\to\mathcal{K}italic_ψ : roman_PTr start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT → caligraphic_K defined by

TPTr2,ψ(T):=[T].formulae-sequencefor-all𝑇subscriptPTr2assign𝜓𝑇delimited-[]𝑇\forall T\in\mathrm{PTr}_{2},\quad\psi(T):=[T].∀ italic_T ∈ roman_PTr start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT , italic_ψ ( italic_T ) := [ italic_T ] .

is a well-defined homeomorphism. In addition, its inverse map is given by ψ1(K)={zn:zK,nω}superscript𝜓1𝐾conditional-set𝑧𝑛formulae-sequence𝑧𝐾𝑛𝜔\psi^{-1}(K)=\{z\upharpoonright n:z\in K,n\in\omega\}italic_ψ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT ( italic_K ) = { italic_z ↾ italic_n : italic_z ∈ italic_K , italic_n ∈ italic_ω } for all K𝒦𝐾𝒦K\in\mathcal{K}italic_K ∈ caligraphic_K. Thanks to [25, Exercise 4.29(vi)], the map φ:𝒦𝒦:𝜑𝒦𝒦\varphi:\mathcal{K}\to\mathcal{K}italic_φ : caligraphic_K → caligraphic_K defined by Φ(K):={ϕ(z):zK}assignΦ𝐾conditional-setitalic-ϕ𝑧𝑧𝐾\Phi(K):=\{\phi(z):z\in K\}roman_Φ ( italic_K ) := { italic_ϕ ( italic_z ) : italic_z ∈ italic_K } is continuous.

At this point, define the function f:PTr2PTr2:𝑓subscriptPTr2subscriptPTr2f:\mathrm{PTr}_{2}\to\mathrm{PTr}_{2}italic_f : roman_PTr start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT → roman_PTr start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT by

f:=ψ1Φψ,assign𝑓superscript𝜓1Φ𝜓f:=\psi^{-1}\circ\Phi\circ\psi,italic_f := italic_ψ start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT ∘ roman_Φ ∘ italic_ψ ,

so that f(T)={ϕ(z)n:z[T],nω}𝑓𝑇conditional-setitalic-ϕ𝑧𝑛formulae-sequence𝑧delimited-[]𝑇𝑛𝜔f(T)=\{\phi(z)\upharpoonright n:z\in[T],n\in\omega\}italic_f ( italic_T ) = { italic_ϕ ( italic_z ) ↾ italic_n : italic_z ∈ [ italic_T ] , italic_n ∈ italic_ω } for each pruned tree T𝑇Titalic_T. Now, it is sufficient to show that f𝑓fitalic_f is a witnessing map for the claimed Wadge reduction (10). For, note that f𝑓fitalic_f is a composition of three continuous maps, hence it is continuous. Moreover, the following chain of equivalences holds for each pruned tree T𝑇Titalic_T on {0,1}01\{0,1\}{ 0 , 1 }:

T𝒯(A) if and only if [T]A=, if and only if Φ([T])B=, if and only if [f(T)]B=, if and only if f(T)𝒯(B).formulae-sequence𝑇𝒯𝐴 if and only if formulae-sequencedelimited-[]𝑇𝐴 if and only if formulae-sequenceΦdelimited-[]𝑇𝐵 if and only if formulae-sequencedelimited-[]𝑓𝑇𝐵 if and only if 𝑓𝑇𝒯𝐵\begin{split}T\in\mathcal{T}(A)\quad&\text{ if and only if }\quad[T]\cap A=% \emptyset,\\ &\text{ if and only if }\quad\Phi([T])\cap B=\emptyset,\\ &\text{ if and only if }\quad[f(T)]\cap B=\emptyset,\\ &\text{ if and only if }\quad f(T)\in\mathcal{T}(B).\\ \end{split}start_ROW start_CELL italic_T ∈ caligraphic_T ( italic_A ) end_CELL start_CELL if and only if [ italic_T ] ∩ italic_A = ∅ , end_CELL end_ROW start_ROW start_CELL end_CELL start_CELL if and only if roman_Φ ( [ italic_T ] ) ∩ italic_B = ∅ , end_CELL end_ROW start_ROW start_CELL end_CELL start_CELL if and only if [ italic_f ( italic_T ) ] ∩ italic_B = ∅ , end_CELL end_ROW start_ROW start_CELL end_CELL start_CELL if and only if italic_f ( italic_T ) ∈ caligraphic_T ( italic_B ) . end_CELL end_ROW

Therefore f1[𝒯(B)]=𝒯(A)superscript𝑓1delimited-[]𝒯𝐵𝒯𝐴f^{-1}[\mathcal{T}(B)]=\mathcal{T}(A)italic_f start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ caligraphic_T ( italic_B ) ] = caligraphic_T ( italic_A ), which completes the proof. ∎

Claim 6.

WF2W𝒯(W)subscriptWsubscriptsuperscriptWF2𝒯𝑊\mathrm{WF}^{\star}_{2}\leq_{\mathrm{W}}\mathcal{T}(W)roman_WF start_POSTSUPERSCRIPT ⋆ end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ≤ start_POSTSUBSCRIPT roman_W end_POSTSUBSCRIPT caligraphic_T ( italic_W ).

Proof.

Thanks to [25, Theorem 22.10], the Borel set W𝑊Witalic_W is Π20subscriptsuperscriptΠ02\Pi^{0}_{2}roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT-hard, that is, AWWsubscriptW𝐴𝑊A\leq_{\mathrm{W}}Witalic_A ≤ start_POSTSUBSCRIPT roman_W end_POSTSUBSCRIPT italic_W for all AΠ20𝐴subscriptsuperscriptΠ02A\in\Pi^{0}_{2}italic_A ∈ roman_Π start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT. In particular, WirrWWsubscriptWsubscript𝑊irr𝑊W_{\mathrm{irr}}\leq_{\mathrm{W}}Witalic_W start_POSTSUBSCRIPT roman_irr end_POSTSUBSCRIPT ≤ start_POSTSUBSCRIPT roman_W end_POSTSUBSCRIPT italic_W. Hence by Claim 5, we conclude that WF2=𝒯(Wirr)W𝒯(W)subscriptsuperscriptWF2𝒯subscript𝑊irrsubscriptW𝒯𝑊\mathrm{WF}^{\star}_{2}=\mathcal{T}(W_{\mathrm{irr}})\leq_{\mathrm{W}}\mathcal% {T}(W)roman_WF start_POSTSUPERSCRIPT ⋆ end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT = caligraphic_T ( italic_W start_POSTSUBSCRIPT roman_irr end_POSTSUBSCRIPT ) ≤ start_POSTSUBSCRIPT roman_W end_POSTSUBSCRIPT caligraphic_T ( italic_W ). ∎

Claim 7.

𝒯(W)WWsubscriptW𝒯𝑊subscript𝑊\mathcal{T}(W)\leq_{\mathrm{W}}\mathcal{I}_{W}caligraphic_T ( italic_W ) ≤ start_POSTSUBSCRIPT roman_W end_POSTSUBSCRIPT caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT.

Proof.

We need to show that there exists a continuous map h:PTr2𝒫(ω):subscriptPTr2𝒫𝜔h:\mathrm{PTr}_{2}\to\mathcal{P}(\omega)italic_h : roman_PTr start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT → caligraphic_P ( italic_ω ) such that

T𝒯(W) if and only if h(T)W𝑇𝒯𝑊 if and only if 𝑇subscript𝑊T\in\mathcal{T}(W)\,\,\,\text{ if and only if }\,\,\,h(T)\in\mathcal{I}_{W}italic_T ∈ caligraphic_T ( italic_W ) if and only if italic_h ( italic_T ) ∈ caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT

for all pruned trees T𝑇Titalic_T on {0,1}01\{0,1\}{ 0 , 1 }.

First, define the map f:{0,1}<ωω:𝑓superscript01absent𝜔𝜔f:\{0,1\}^{<\omega}\to\omegaitalic_f : { 0 , 1 } start_POSTSUPERSCRIPT < italic_ω end_POSTSUPERSCRIPT → italic_ω as follows: for each s=(s0,,sk){0,1}<ω𝑠subscript𝑠0subscript𝑠𝑘superscript01absent𝜔s=(s_{0},\ldots,s_{k})\in\{0,1\}^{<\omega}italic_s = ( italic_s start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , … , italic_s start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ) ∈ { 0 , 1 } start_POSTSUPERSCRIPT < italic_ω end_POSTSUPERSCRIPT, let f(s)𝑓𝑠f(s)italic_f ( italic_s ) be the unique nonnegative integer such that

zf(s)=(s0,,sk,1,0,0,).subscript𝑧𝑓𝑠subscript𝑠0subscript𝑠𝑘100z_{f(s)}=(s_{0},\ldots,s_{k},1,0,0,\ldots).italic_z start_POSTSUBSCRIPT italic_f ( italic_s ) end_POSTSUBSCRIPT = ( italic_s start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , … , italic_s start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT , 1 , 0 , 0 , … ) .

Of course, f𝑓fitalic_f is injective. Recalling that x0=(0,0,)subscript𝑥000x_{0}=(0,0,\ldots)italic_x start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT = ( 0 , 0 , … ) is the unique sequence with empty support, it is easy to see the image of f𝑓fitalic_f is the set of positive integers.

At this point, define the map h:PTr2𝒫(ω):subscriptPTr2𝒫𝜔h:\mathrm{PTr}_{2}\to\mathcal{P}(\omega)italic_h : roman_PTr start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT → caligraphic_P ( italic_ω ) by h(T):={f(s):sT}assign𝑇conditional-set𝑓𝑠𝑠𝑇h(T):=\{f(s):s\in T\}italic_h ( italic_T ) := { italic_f ( italic_s ) : italic_s ∈ italic_T } for all TPTr2𝑇subscriptPTr2T\in\mathrm{PTr}_{2}italic_T ∈ roman_PTr start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT. Since T𝑇Titalic_T is an infinite and f𝑓fitalic_f is injective, then h(T)𝑇h(T)italic_h ( italic_T ) is infinite as well. It is also not difficult to show that hhitalic_h is continuous: indeed, a basic clopen set containing h(T)𝑇h(T)italic_h ( italic_T ) is of the type

V:={Sω:f(A)S and f(B)S=}assign𝑉conditional-set𝑆𝜔𝑓𝐴𝑆 and 𝑓𝐵𝑆V:=\{S\in\omega:f(A)\subseteq S\text{ and }f(B)\cap S=\emptyset\}italic_V := { italic_S ∈ italic_ω : italic_f ( italic_A ) ⊆ italic_S and italic_f ( italic_B ) ∩ italic_S = ∅ }

for some finite sets A,B{0,1}<ω𝐴𝐵superscript01absent𝜔A,B\subseteq\{0,1\}^{<\omega}italic_A , italic_B ⊆ { 0 , 1 } start_POSTSUPERSCRIPT < italic_ω end_POSTSUPERSCRIPT such that every restriction of each sA𝑠𝐴s\in Aitalic_s ∈ italic_A does not belong to B𝐵Bitalic_B. Then, the set U:={TPTr2:AT and BT=}assign𝑈conditional-set𝑇subscriptPTr2𝐴𝑇 and 𝐵𝑇U:=\{T\in\mathrm{PTr}_{2}:A\subseteq T\text{ and }B\cap T=\emptyset\}italic_U := { italic_T ∈ roman_PTr start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT : italic_A ⊆ italic_T and italic_B ∩ italic_T = ∅ } is an open set such that h[U]Vdelimited-[]𝑈𝑉h[U]\subseteq Vitalic_h [ italic_U ] ⊆ italic_V.

Next, we claim that

[T]=L𝒙h(T).delimited-[]𝑇subscriptL𝒙𝑇[T]=\mathrm{L}_{\bm{x}\upharpoonright h(T)}.[ italic_T ] = roman_L start_POSTSUBSCRIPT bold_italic_x ↾ italic_h ( italic_T ) end_POSTSUBSCRIPT . (11)

On the one hand, fix aL𝒙h(T)𝑎subscriptL𝒙𝑇a\in\mathrm{L}_{\bm{x}\upharpoonright h(T)}italic_a ∈ roman_L start_POSTSUBSCRIPT bold_italic_x ↾ italic_h ( italic_T ) end_POSTSUBSCRIPT. Then there exists a strictly increasing sequence of positive integers (pk)subscript𝑝𝑘(p_{k})( italic_p start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ) such that limkzpk=asubscript𝑘subscript𝑧subscript𝑝𝑘𝑎\lim_{k}z_{p_{k}}=aroman_lim start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT italic_z start_POSTSUBSCRIPT italic_p start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT end_POSTSUBSCRIPT = italic_a. For each kω𝑘𝜔k\in\omegaitalic_k ∈ italic_ω, set sk:=f1(pk)assignsuperscript𝑠𝑘superscript𝑓1subscript𝑝𝑘s^{k}:=f^{-1}(p_{k})italic_s start_POSTSUPERSCRIPT italic_k end_POSTSUPERSCRIPT := italic_f start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT ( italic_p start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ) and define kωsubscript𝑘𝜔\ell_{k}\in\omegaroman_ℓ start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ∈ italic_ω so that sk=(s0k,s1k,,skk)superscript𝑠𝑘subscriptsuperscript𝑠𝑘0subscriptsuperscript𝑠𝑘1subscriptsuperscript𝑠𝑘subscript𝑘s^{k}=(s^{k}_{0},s^{k}_{1},\ldots,s^{k}_{\ell_{k}})italic_s start_POSTSUPERSCRIPT italic_k end_POSTSUPERSCRIPT = ( italic_s start_POSTSUPERSCRIPT italic_k end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , italic_s start_POSTSUPERSCRIPT italic_k end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , … , italic_s start_POSTSUPERSCRIPT italic_k end_POSTSUPERSCRIPT start_POSTSUBSCRIPT roman_ℓ start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT end_POSTSUBSCRIPT ). Hence

zpk=zf(sk)=(s0k,s1k,,skk,1,0,0,).subscript𝑧subscript𝑝𝑘subscript𝑧𝑓superscript𝑠𝑘subscriptsuperscript𝑠𝑘0subscriptsuperscript𝑠𝑘1subscriptsuperscript𝑠𝑘subscript𝑘100z_{p_{k}}=z_{f(s^{k})}=(s^{k}_{0},s^{k}_{1},\ldots,s^{k}_{\ell_{k}},1,0,0,% \ldots).italic_z start_POSTSUBSCRIPT italic_p start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT end_POSTSUBSCRIPT = italic_z start_POSTSUBSCRIPT italic_f ( italic_s start_POSTSUPERSCRIPT italic_k end_POSTSUPERSCRIPT ) end_POSTSUBSCRIPT = ( italic_s start_POSTSUPERSCRIPT italic_k end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , italic_s start_POSTSUPERSCRIPT italic_k end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , … , italic_s start_POSTSUPERSCRIPT italic_k end_POSTSUPERSCRIPT start_POSTSUBSCRIPT roman_ℓ start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT end_POSTSUBSCRIPT , 1 , 0 , 0 , … ) .

Fix nω𝑛𝜔n\in\omegaitalic_n ∈ italic_ω. Then there exists k0ωsubscript𝑘0𝜔k_{0}\in\omegaitalic_k start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT ∈ italic_ω such that an=zpkn𝑎𝑛subscript𝑧subscript𝑝𝑘𝑛a\upharpoonright n=z_{p_{k}}\upharpoonright nitalic_a ↾ italic_n = italic_z start_POSTSUBSCRIPT italic_p start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT end_POSTSUBSCRIPT ↾ italic_n for all kk0𝑘subscript𝑘0k\geq k_{0}italic_k ≥ italic_k start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT. Since f𝑓fitalic_f is injective, there exists k1max{n,k0}subscript𝑘1𝑛subscript𝑘0k_{1}\geq\max\{n,k_{0}\}italic_k start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ≥ roman_max { italic_n , italic_k start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT } such that k1nsubscriptsubscript𝑘1𝑛\ell_{k_{1}}\geq nroman_ℓ start_POSTSUBSCRIPT italic_k start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT end_POSTSUBSCRIPT ≥ italic_n. It follows that

an=zf(sk1)n=(s0k1,s1k1,,snk1)=sk1n.𝑎𝑛subscript𝑧𝑓superscript𝑠subscript𝑘1𝑛subscriptsuperscript𝑠subscript𝑘10subscriptsuperscript𝑠subscript𝑘11subscriptsuperscript𝑠subscript𝑘1𝑛superscript𝑠subscript𝑘1𝑛a\upharpoonright n=z_{f(s^{k_{1}})}\upharpoonright n=(s^{k_{1}}_{0},s^{k_{1}}_% {1},\ldots,s^{k_{1}}_{n})=s^{k_{1}}\upharpoonright n.italic_a ↾ italic_n = italic_z start_POSTSUBSCRIPT italic_f ( italic_s start_POSTSUPERSCRIPT italic_k start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT end_POSTSUPERSCRIPT ) end_POSTSUBSCRIPT ↾ italic_n = ( italic_s start_POSTSUPERSCRIPT italic_k start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , italic_s start_POSTSUPERSCRIPT italic_k start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , … , italic_s start_POSTSUPERSCRIPT italic_k start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT end_POSTSUPERSCRIPT start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT ) = italic_s start_POSTSUPERSCRIPT italic_k start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT end_POSTSUPERSCRIPT ↾ italic_n .

Since T𝑇Titalic_T is a tree and sk1Tsuperscript𝑠subscript𝑘1𝑇s^{k_{1}}\in Titalic_s start_POSTSUPERSCRIPT italic_k start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT end_POSTSUPERSCRIPT ∈ italic_T, we obtain anT𝑎𝑛𝑇a\upharpoonright n\in Titalic_a ↾ italic_n ∈ italic_T. By the arbitrariness of n𝑛nitalic_n, it follows that a𝑎aitalic_a is an infinite branch of T𝑇Titalic_T. Therefore L𝒙h(T)[T]subscriptL𝒙𝑇delimited-[]𝑇\mathrm{L}_{\bm{x}\upharpoonright h(T)}\subseteq[T]roman_L start_POSTSUBSCRIPT bold_italic_x ↾ italic_h ( italic_T ) end_POSTSUBSCRIPT ⊆ [ italic_T ].

On the other hand, fix a[T]𝑎delimited-[]𝑇a\in[T]italic_a ∈ [ italic_T ], so that anT𝑎𝑛𝑇a\upharpoonright n\in Titalic_a ↾ italic_n ∈ italic_T for all nω𝑛𝜔n\in\omegaitalic_n ∈ italic_ω. It follows that f(an)h(T)𝑓𝑎𝑛𝑇f(a\upharpoonright n)\in h(T)italic_f ( italic_a ↾ italic_n ) ∈ italic_h ( italic_T ) and zf(an)=(a0,,an,1,0,0,)subscript𝑧𝑓𝑎𝑛subscript𝑎0subscript𝑎𝑛100z_{f(a\upharpoonright n)}=(a_{0},\ldots,a_{n},1,0,0,\ldots)italic_z start_POSTSUBSCRIPT italic_f ( italic_a ↾ italic_n ) end_POSTSUBSCRIPT = ( italic_a start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , … , italic_a start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT , 1 , 0 , 0 , … ) for all nω𝑛𝜔n\in\omegaitalic_n ∈ italic_ω. It is clear that limnzf(an)=asubscript𝑛subscript𝑧𝑓𝑎𝑛𝑎\lim_{n}z_{f(a\upharpoonright n)}=aroman_lim start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT italic_z start_POSTSUBSCRIPT italic_f ( italic_a ↾ italic_n ) end_POSTSUBSCRIPT = italic_a. Therefore the opposite inclusion [T]L𝒙h(T)delimited-[]𝑇subscriptL𝒙𝑇[T]\subseteq\mathrm{L}_{\bm{x}\upharpoonright h(T)}[ italic_T ] ⊆ roman_L start_POSTSUBSCRIPT bold_italic_x ↾ italic_h ( italic_T ) end_POSTSUBSCRIPT holds.

Recall that T𝒯(W)𝑇𝒯𝑊T\in\mathcal{T}(W)italic_T ∈ caligraphic_T ( italic_W ) if and only if [T]W=delimited-[]𝑇𝑊[T]\cap W=\emptyset[ italic_T ] ∩ italic_W = ∅. Thanks to (11), this is equivalent to L𝒙h(T)W=subscriptL𝒙𝑇𝑊\mathrm{L}_{\bm{x}\upharpoonright h(T)}\cap W=\emptysetroman_L start_POSTSUBSCRIPT bold_italic_x ↾ italic_h ( italic_T ) end_POSTSUBSCRIPT ∩ italic_W = ∅, that is, h(T)W𝑇subscript𝑊h(T)\in\mathcal{I}_{W}italic_h ( italic_T ) ∈ caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT. ∎

To conclude the proof, thanks to Claim 6, Claim 7, and the transitivity of WsubscriptW\leq_{\mathrm{W}}≤ start_POSTSUBSCRIPT roman_W end_POSTSUBSCRIPT, we get WF2WWsubscriptWsubscriptsuperscriptWF2subscript𝑊\mathrm{WF}^{\star}_{2}\leq_{\mathrm{W}}\mathcal{I}_{W}roman_WF start_POSTSUPERSCRIPT ⋆ end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ≤ start_POSTSUBSCRIPT roman_W end_POSTSUBSCRIPT caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT. Hence there exists a continuous map g:PTr2𝒫(ω):𝑔subscriptPTr2𝒫𝜔g:\mathrm{PTr}_{2}\to\mathcal{P}(\omega)italic_g : roman_PTr start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT → caligraphic_P ( italic_ω ) such that

g1[W]=WF2.superscript𝑔1delimited-[]subscript𝑊subscriptsuperscriptWF2g^{-1}[\mathcal{I}_{W}]=\mathrm{WF}^{\star}_{2}.italic_g start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT [ caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT ] = roman_WF start_POSTSUPERSCRIPT ⋆ end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT .

The conclusion follows by the facts that WF2subscriptsuperscriptWF2\mathrm{WF}^{\star}_{2}roman_WF start_POSTSUPERSCRIPT ⋆ end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT is Π11subscriptsuperscriptΠ11\Pi^{1}_{1}roman_Π start_POSTSUPERSCRIPT 1 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT-complete (hence, not analytic) and that analytic sets are closed under continuous preimages, see e.g. [31, Proposition 4.1.2]. ∎

It is possible to show that Wirrsubscriptsubscript𝑊irr\mathcal{I}_{W_{\mathrm{irr}}}caligraphic_I start_POSTSUBSCRIPT italic_W start_POSTSUBSCRIPT roman_irr end_POSTSUBSCRIPT end_POSTSUBSCRIPT is coanalytic, hence Π11subscriptsuperscriptΠ11\Pi^{1}_{1}roman_Π start_POSTSUPERSCRIPT 1 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT-complete. We leave as open question for the interested reader to check whether Wsubscript𝑊\mathcal{I}_{W}caligraphic_I start_POSTSUBSCRIPT italic_W end_POSTSUBSCRIPT is always coanalytic whenever W{0,1}ω𝑊superscript01𝜔W\subseteq\{0,1\}^{\omega}italic_W ⊆ { 0 , 1 } start_POSTSUPERSCRIPT italic_ω end_POSTSUPERSCRIPT is Borel and not Σ20subscriptsuperscriptΣ02\Sigma^{0}_{2}roman_Σ start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT.

5. Acknowledgments

The authors are grateful to Rafal Filipów and Adam Kwela (University of Gdansk, PL) for several discussions related to the results of the manuscript.

References

  • [1] M. Balcerzak, S. Głąb, and P. Leonetti, Another characterization of meager ideals, Rev. R. Acad. Cienc. Exactas Fís. Nat. Ser. A Mat. RACSAM 117 (2023), no. 2, Paper No. 90, 10.
  • [2] M. Balcerzak and P. Leonetti, On the relationship between ideal cluster points and ideal limit points, Topology Appl. 252 (2019), 178–190.
  • [3] P. Barbarski, R. Filipów, N. Mrożek, and P. Szuca, When does the Katĕtov order imply that one ideal extends the other?, Colloq. Math. 130 (2013), no. 1, 91–102.
  • [4] J. Calbrix, Classes de Baire et espaces d’applications continues, C. R. Acad. Sci. Paris Sér. I Math. 301 (1985), no. 16, 759–762.
  • [5] by same author, Filtres boréliens sur l’ensemble des entiers et espaces des applications continues, Rev. Roumaine Math. Pures Appl. 33 (1988), no. 8, 655–661.
  • [6] W. W. Comfort and S. Negrepontis, The theory of ultrafilters, Die Grundlehren der mathematischen Wissenschaften, Band 211, Springer-Verlag, New York-Heidelberg, 1974.
  • [7] I. Farah, Analytic quotients: theory of liftings for quotients over analytic ideals on the integers, Mem. Amer. Math. Soc. 148 (2000), no. 702, xvi+177.
  • [8] by same author, Luzin gaps, Trans. Amer. Math. Soc. 356 (2004), no. 6, 2197–2239.
  • [9] R. Filipów, K. Kowitz, and A. Kwela, Characterizing existence of certain ultrafilters, Ann. Pure Appl. Logic 173 (2022), no. 9, Paper No. 103157, 31.
  • [10] R. Filipow, A. Kwela, and P. Leonetti, Borel complexity of sets of ideal limit points, manuscript.
  • [11] D. H. Fremlin, Measure theory. Vol. 2, Torres Fremlin, Colchester, 2003, Broad foundations, Corrected second printing of the 2001 original.
  • [12] D. H. Fremlin, R. A. Johnson, and E. Wajch, Countable network weight and multiplication of Borel sets, Proc. Amer. Math. Soc. 124 (1996), no. 9, 2897–2903.
  • [13] J. A. Fridy, Statistical limit points, Proc. Amer. Math. Soc. 118 (1993), no. 4, 1187–1192.
  • [14] S. Garcia-Ferreira and O. Guzmán, More on MAD families and P𝑃Pitalic_P-points, Topology Appl. 305 (2022), Paper No. 107871, 10.
  • [15] R. W. Hansell, J. E. Jayne, and C. A. Rogers, K𝐾Kitalic_K-analytic sets, Mathematika 30 (1983), no. 2, 189–221 (1984).
  • [16] X. He, H. Zhang, and S. Zhang, The Borel complexity of ideal limit points, Topology Appl. 312 (2022), Paper No. 108061, 12.
  • [17] by same author, More on ideal limit points, Topology Appl. 322 (2022), Paper No. 108324.
  • [18] G. Hjorth and A. S. Kechris, New dichotomies for Borel equivalence relations, Bull. Symbolic Logic 3 (1997), no. 3, 329–346.
  • [19] M. Hrušák, Combinatorics of filters and ideals, Set theory and its applications, Contemp. Math., vol. 533, Amer. Math. Soc., Providence, RI, 2011, pp. 29–69.
  • [20] M. Hrušák and D. Meza-Alcántara, Comparison game on Borel ideals, Comment. Math. Univ. Carolin. 52 (2011), no. 2, 191–204.
  • [21] M. Hrušák, D. Meza-Alcántara, E. Thümmel, and C. Uzcátegui, Ramsey type properties of ideals, Ann. Pure Appl. Logic 168 (2017), no. 11, 2022–2049.
  • [22] M. Hrušák and J. L. Verner, Adding ultrafilters by definable quotients, Rend. Circ. Mat. Palermo (2) 60 (2011), no. 3, 445–454.
  • [23] W. Just and A. Krawczyk, On certain Boolean algebras 𝒫(ω)/I𝒫𝜔𝐼{\mathscr{P}}(\omega)/Iscript_P ( italic_ω ) / italic_I, Trans. Amer. Math. Soc. 285 (1984), no. 1, 411–429.
  • [24] V. Kadets, D. Seliutin, and J. Tryba, Conglomerated filters and statistical measures, J. Math. Anal. Appl. 509 (2022), no. 1, Paper No. 125955, 17.
  • [25] A. S. Kechris, Classical descriptive set theory, Graduate Texts in Mathematics, vol. 156, Springer-Verlag, New York, 1995.
  • [26] by same author, Rigidity properties of Borel ideals on the integers, vol. 85, 1998, 8th Prague Topological Symposium on General Topology and Its Relations to Modern Analysis and Algebra (1996), pp. 195–205.
  • [27] P. Leonetti and F. Maccheroni, Characterizations of ideal cluster points, Analysis (Berlin) 39 (2019), no. 1, 19–26.
  • [28] D. Meza-Alcántara, Ideals and filters on countable sets, (2009), PhD Thesis, UNAM México.
  • [29] C. A. Rogers, J. E. Jayne, C. Dellacherie, F. Topsoe, J. Hoffmann-Jorgensen, D. A. Martin, A. S. Kechris, and A. H. Stone, Analytic sets, Academic Press, London, 1980.
  • [30] S. Solecki, Analytic ideals, Bull. Symbolic Logic 2 (1996), no. 3, 339–348.
  • [31] S. M. Srivastava, A course on Borel sets, Graduate Texts in Mathematics, vol. 180, Springer-Verlag, New York, 1998.
  • [32] M. Staniszewski, On ideal equal convergence II, J. Math. Anal. Appl. 451 (2017), no. 2, 1179–1197.
  • [33] M. Talagrand, Compacts de fonctions mesurables et filtres non mesurables, Studia Math. 67 (1980), no. 1, 13–43.
  • [34] F. van Engelen, On Borel ideals, Ann. Pure Appl. Logic 70 (1994), no. 2, 177–203.