Completeness of invariant ideals in groups

We introduce and study various notions of completeness of translation-invariant ideals in groups.

Gespeichert in:
Bibliographische Detailangaben
Datum:2010
Hauptverfasser: Banakh, T. O., Lyaskovska, N., Банах, Т. О., Лясковська, Н.
Format: Artikel
Sprache:Englisch
Veröffentlicht: Institute of Mathematics, NAS of Ukraine 2010
Online Zugang:https://umj.imath.kiev.ua/index.php/umj/article/view/2934
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Назва журналу:Ukrains’kyi Matematychnyi Zhurnal
Завантажити файл: Pdf

Institution

Ukrains’kyi Matematychnyi Zhurnal
_version_ 1860508935224557568
author Banakh, T. O.
Lyaskovska, N.
Банах, Т. О.
Лясковська, Н.
author_facet Banakh, T. O.
Lyaskovska, N.
Банах, Т. О.
Лясковська, Н.
author_sort Banakh, T. O.
baseUrl_str https://umj.imath.kiev.ua/index.php/umj/oai
collection OJS
datestamp_date 2020-03-18T19:40:46Z
description We introduce and study various notions of completeness of translation-invariant ideals in groups.
first_indexed 2026-03-24T02:33:06Z
format Article
fulltext UDC 512 T. Banakh (Ivan Franko Nat. Univ. Lviv, Ukraine; Inst. Mat., Univ. Humanist.-Przyrod. Jana Kochanow- skiego Kielcach, Poland), N. Lyaskovska (Ivan Franko Nat. Univ. Lviv, Ukraine) COMPLETENESS OF INVARIANT IDEALS IN GROUPS ПОВНОТА IНВАРIАНТНИХ IДЕАЛIВ У ГРУПАХ We introduce and study various notions of completeness of translation-invariant ideals in groups. Введено та дослiджено рiзнi поняття повноти iнварiантних iдеалiв у групах. In this paper we introduce and study various notions of completeness of translation- invariant ideals in groups. Those completeness notions have topological, measure-theoretic, or packing nature. 1. The ideal of small subsets in a group. A proper family I ( P(G) of subsets of a group G is an ideal if I is closed under taking subsets and unions. An ideal I on G is translation-invariant (briefly, invariant) if xA ∈ I for each x ∈ G and A ∈ I. Each group G possesses the trivial ideal I0 = {∅} and this is a unique invariant ideal on a finite group. Each infinite group G possesses the invariant ideal F consisting of all finite subsets of G and this is the smallest invariant ideal containing a non-empty set. A less trivial example of an invariant ideal is the ideal of small sets. A subset A of a group G is called (i) large if FA = G for some finite set F ⊂ G; (ii) small if for each large set B ⊂ G the complement B \A is large. It follows from the definition that the family S of small subsets of G is an invariant ideal in G. To give a characterization of small subsets let us introduce the following combinatorial notion that has a topological flavour. For two subsets B,U ⊂ G the set IntBU = {x ∈ G : Bx ⊂ U} = G \ (B−1(G \ U)) is called the B-interior of U in G. Small sets admit a topological characterization. Theorem 1.1. For a subset A of a group G the following conditions are equiva- lent: (1) A is small; (2) for any finite set F ⊂ G the complement G \ FA is large; (3) for any finite set F ⊂ G there is a finite set B ⊂ G such that IntB(FA) = ∅; (4) A is nowhere dense with respect to some left-invariant totally bounded topology on G. A topology τ on a group G is totally bounded if each non-empty open set U ∈ τ is large. Proof. The equivalence (1)⇔ (2)⇔ (3) was proved in [1, 2], see also Theorem 2.1. (2)⇒ (4) Assuming that A ⊂ G satisfies (2), observe that the topology τ = {∅} ∪ { U ⊂ G : U ⊃ G \ FA for some finite F ⊂ G } c© T. BANAKH, N. LYASKOVSKA, 2010 1022 ISSN 1027-3190. Укр. мат. журн., 2010, т. 62, № 8 COMPLETENESS OF INVARIANT IDEALS IN GROUPS 1023 on G is left-invariant and totally bounded. The set A is closed in (G, τ) and has empty interior. Otherwise we could find a finite subset F ⊂ G with G \ FA ⊂ A. Then for the finite set Fe = F ∪ {e} we get FeA = FA ∪A = G, which contradicts (2). (4) ⇒ (2) Assume that a subset A ⊂ G is nowhere dense in some totally bounded left-invariant topology τ. Then for every finite F ⊂ G the set FA is nowhere dense in (G, τ) and hence the complement G \ FA contains some open set U ∈ τ. Since τ is totally bounded, U is large in G. Consequently, there is a finite subset B ⊂ G such that G = BU ⊂ B(G \ FA) ⊂ G, witnessing that (2) holds. By [3] or [4] (1.3, 9.4), on any group G there exists a totally bounded left-invariant topology τ, which is Hausdorff and extremally disconnected. The latter means that the closure of each open subset of (G, τ) is open. This result of I. Protasov implies the fol- lowing description of the ideal of small sets. Theorem 1.2. The ideal S of small subsets of any groupG coincides with the ideal of nowhere dense subsets of G endowed with some Hausdorff extremally disconnected left-invariant topology τs. Proof. According to [3] or [4] (1.3, 9.4), the group G admits a Hausdorff extremally disconnected totally bounded left-invariant topology τ. Let τs be the topology on G gen- erated by the base B = { U \A : U ∈ τ, A ∈ S } . Let us show that the topology τs is totally bounded. Given any basic set U \ A ∈ B, use the total boundedness of the topology τ in order to find a finite subset F ⊂ G such that G = FU. Since the set A is small, for the finite set F there is a finite set E ⊂ G such that E(G \ FA) = G. Now observe that EF (U \A) ⊃ E(FU \ FA) = E(G \ FA) = G, witnessing that the set U \A is large. Applying Theorem 1.1, we conclude that the idealM of nowhere dense subset of the totally bounded left-topological group (G, τs) lies in the ideal S of small sets. In order to prove the reverse inclusion S ⊂ M, fix any small set S ⊂ G. It follows from the definition of the topology τs that S is closed in (G, τs). We claim that S is nowhere dense. Assuming the converse, we would find a non-empty basic set U \A ⊂ S with U ∈ τ and A ∈ S. Then the set U ⊂ A ∪ S is small, which contradicts the total boundedness of the topology τ. This completes the proof of the equality S =M. Now we check that the topology τs is Hausdorff and extremally disconnected. Since the topology τ is Hausdorff, so is the topology τs ⊃ τ. In order to prove the extremal disconnectedness of the topology τs, take any open subset W ∈ τs and write it as the union W = ⋃ i∈I(Ui \ Ai) of basic sets Ui \ Ai ∈ B, i ∈ I. Consider the open set U = ⋃ i∈I Ui ∈ τ and its closure U in (G, τ). The extremal disconnectedness of the topology τ implies that U ∈ τ ⊂ τs. It remains to check that W is dense in U in the topology τs. In the opposite case, we would find a non-empty basic set V \ A ∈ B that meets U but is disjoint with W. The set V is open in τ and meets the closure U of the open set U ∈ τ. Consequently, V ∩ U 6= ∅ and there is an index i ∈ I such that V ∩ Ui 6= ∅. Then (Ui ∩ V \A) ∩ (Ui \Ai) ⊂ (Ui ∩ V \A) ∩W = ∅ ISSN 1027-3190. Укр. мат. журн., 2010, т. 62, № 8 1024 T. BANAKH, N. LYASKOVSKA implies that V ∩ Ui ⊂ A ∪ Ai. But this is impossible as V ∩ Ui ∈ τ is large while the set A ∪ Ai is small. This contradiction shows that the set W is dense in U ∈ τs in the topology τs. Consequently, the closure of the open set W in the topology τs in open and the left-topological group (G, τs) is extremally disconnected. Theorem 1.2 is proved. Problem 1.1. Is the ideal S of small subsets of the group Z equal to the ideal of nowhere dense subsets of Z with respect to some regular totally bounded left-invariant topology τ on Z? In light of this problem, let us remark that each countable Abelian group G contains a small subset which is dense with respect to any totally bounded group topology on G, see [2]. 2. S-complete ideals. In this section, given an invariant ideal I in a group G, we introduce an invariant ideal SI called the S-completion of I. For two subsets A,B ⊂ G we write A ⊂I B if A \B ∈ I and A =I B if A ⊂I B and B ⊂I A. We define a subset A of a group G to be (i) I-large if FA =I G for some finite subset F ⊂ G; (ii) I-small if for each I-large subset L ⊂ G the complement L \A is I-large. I-Small subsets admit the following characterization. Theorem 2.1. For a subset A ⊂ G the following conditions are equivalent: (1) A is I-small; (2) for every finite subset F ⊂ G the set G \ FA is I-large; (3) for every finite subset F ⊂ G there is a finite subset B ⊂ G such that IntB(FA) ∈ I. Proof. (1) ⇒ (2) Assume that A is I-small. Then for every finite set F ⊂ G the set FA is I-small (because the ideal SI is invariant) and then its complement G \ FA is I-large. (2) ⇒ (3) By (2), for every finite subset F ⊂ G the set G \ FA is I-large. Conse- quently, B(G \ FA) =I G for some finite set B ⊂ G. We can assume that B = B−1. It follows that the set I = G \ B(G \ FA) belongs to the ideal I. We claim that IntB(FA) ⊂ I. Indeed, for every x ∈ IntB(FA), we get Bx ⊂ FA and thus Bx∩ (G \ FA) = ∅. Then x ∈ G \B−1(G \ FA) = I. (3)⇒ (2) By (3), for every finite set F ⊂ G there is a finite set B = B−1 ⊂ G such that I = IntB(FA) ∈ I. Then for every x ∈ G \ I the set Bx meets G \ FA and hence G \ I ⊂ B(G \ FA), which means that B(G \ FA) is I-large. (2)⇒ (1) Given an I-large subset L ⊂ G we need to show that L\A is I-large. Find a finite subset F ⊂ G such that FL =I G. By (2), the set G \ FA is I-large and hence B(G\FA) =I G for some finite set B ⊂ G. Consider the finite set BF and observe that BF (L \A) ⊃ B(FL \ FA) =I B(G \ FA) =I G and hence BF (L \A) =I G witnessing that the set L \A is I-large. Theorem 2.1 is proved. It follows from the definition that the family SI of all I-small subsets is an invariant ideal that contains the ideal I. This ideal SI is called the S-completion of the ideal I. An invariant ideal I in a group G is called S-complete if SI = I. It is clear that for the smallest ideal I0 = {∅} its S-completion SI0 coincides with the ideal S of all small subsets of G. ISSN 1027-3190. Укр. мат. журн., 2010, т. 62, № 8 COMPLETENESS OF INVARIANT IDEALS IN GROUPS 1025 Theorem 2.2. For any invariant ideal I the ideal SI is S-complete. In particular, the ideal S = SI0 is S-complete. Proof. The S-completeness of the ideal SI will follow as soon as we check that each SI-small subset A ⊂ G is I-small. Given a finite subset F ⊂ G, we need to find a finite subset B ⊂ G such that IntB(FA) ∈ I. Since A is SI-small, we can apply Theorem 2.1 and find a finite subset B1 ⊂ G such that IntB1 (FA) ∈ SI . Applying this theorem once more, find a finite subset B2 ⊂ G such that IntB2IntB1(FA) ∈ I. Then for the finite set B = B1B2 ⊂ G we get IntB(FA) ⊂ IntB2 IntB1 FA ∈ I, which means that A is I-small by Theorem 2.1. Theorem 2.2 is proved. Corollary 2.1. For any invariant ideal I ⊂ S we get SI = SS = S. The definition of the I-small set implies that the S-completeness is preserved by arbitrary intersections. Proposition 2.1. For any S-complete ideals Iα, α ∈ A, in a group G the inter- section I = ⋂ α∈A Iα is an S-complete ideal in G. The following example shows that in general, the S-completion SI of an invariant ideal I is strictly larger that the smallest ideal generated by the union S ∪ I. Example 2.1. There is a set I ⊂ Z that generates an invariant ideal I such that the ideal SI contains an I-small subset A ⊂ Z that cannot be written as the union A = As ∪Ai of two subsets As ∈ S and Ai ∈ I. Proof. Choose a double sequence (In,m)n,m∈ω of pairwise disjoint intervals in R such that diamIn,m →∞ as n+m→∞. For every n ∈ ω let In = ⋃ m∈ω In,m. Let I = ⋃ n∈ω In ∩ n2Z and A = ⋃ n∈ω In ∩ n(2Z+ 1). The set I generates the invariant ideal I = {J ⊂ Z : ∃F ∈ F J ⊂ FI}. It is easy to check that the set A is I-small but for every J ∈ I the complement A \ J is not small. 3. The N -completion of an invariant ideal in an amenable group. It turns out that the S-completion SI of an invariant ideal I in an amenable groupG contains another interesting ideal NI called the N -completion of I. Let us recall that a group G is called amenable if G has a Banach measure, which is a left-invariant probability measure µ : P(G) → [0, 1] defined on the family of all subsets of G, see [5]. By the Følner condition [5] (0.7), a group G is amenable if and only if for every finite set F ⊂ G and every ε > 0 there is a finite set E ⊂ G such that |E4xE| < ε|E| for all x ∈ F. Here A4B = (A \ B) ∪ (B \ A) is the symmetric difference of two sets A,B. The class of amenable groups contains all locally finite and all abelian groups and is closed under many operations over groups, see [5] (0.16). It is clear that for each Banach measure µ on an amenable group G the family of µ-null sets Nµ = {A ⊂ G : µ(A) = 0} is an invariant ideal in G. The following theorem suggested to the authors by I. V. Protasov shows that the ideals Nµ form a cofinal subset in the family of all invariant ideals on an amenable group. ISSN 1027-3190. Укр. мат. журн., 2010, т. 62, № 8 1026 T. BANAKH, N. LYASKOVSKA Theorem 3.1. Each invariant ideal I in an amenable group G lies in the ideal Nµ for a suitable Banach measure µ on G. Proof. The Banach measure µ with Nµ ⊃ I will be constructed as a limit point of a net of probability measures (µd)d∈D indexed by elements of the directed set D = = F × N× I endowed with the partial order (F, n,A) ≤ (E,m,B) iff F ⊂ E, n ≤ m, and A ⊂ B. Here F is the family of finite subsets of the group G. To each triple d = (F, n,A) ∈ D we assign a probability measure µd on G in the following manner. Using the Følner criterion of amenability [5] (0.7), find a finite subset Fd ⊂ G such that |Fd4xFd| |Fd| < 1 n for all x ∈ F. Since A belongs to an invariant ideal, we can find a point yd ∈ G \ F−1d A. Now consider the probability measure µd : P(G)→ [0, 1] defined by µd(B) = |B ∩ Fdyd| |Fd| for B ⊂ G and observe that µd(A) = 0. Each measure µd, d ∈ D, being a function on the family P(G) of all subsets of the group G, is a point of the Tychonov cube [0, 1]P(G). By the compactness of [0, 1]P(G), the net of measures (µd)d∈D has a limit point µ ∈ [0, 1]P(G), see [6] (3.1.23). This is a function µ : P(G) → [0, 1] such that for each neighborhood O(µ) of µ in [0, 1]P(G) and every d0 ∈ D there is d ≥ d0 in D with µd ∈ O(µ). We claim that µ is a Banach measure on G with I ⊂ Nµ. We need to check the following conditions: (1) µ(G) = 1; (2) µ(A ∪B) = µ(A) + µ(B) for any disjoint sets A,B ⊂ G; (3) µ(xB) = µ(B) for every x ∈ G and B ⊂ G; (4) µ(B) = 0 for each B ∈ I. 1. Assuming that µ(G) < 1, consider the neighborhood O1(µ) = {η ∈ [0, 1]P(G) : η(G) < 1} and note that in contains no measure µd, d ∈ D. 2. Assuming that µ(A∪B) 6= µ(A)+µ(B) for some disjoint setsA,B ⊂ G, observe that O2(µ) = { η ∈ [0, 1]P(G) : η(A ∪ B) 6= η(A) + η(B) } is an open neighborhood of µ in [0, 1]P(G) containing no measure µd, d ∈ D. 3. Assume that µ(xB) 6= µ(B) for some x ∈ G and B ⊂ G. Find n ∈ N such that 3 n < |µ(xB) − µ(B)| and consider the element d0 = ({x−1}, n,∅) ∈ D. Since µ is a limit point of the net (µd)d∈D, the neighborhood O3(µ) = { η ∈ [0, 1]P(G) : max{|η(A)− µ(A)|, |η(xA)− µ(xA)|} < 1 n } of µ contains the measure µd for some d = (F,m,B) ∈ D with d ≥ d0. Since {x−1} ⊂ F, the definition of the set Fd guarantees that |Fd4x−1Fd| |Fd| < 1 m ≤ ≤ 1 n . Then ISSN 1027-3190. Укр. мат. журн., 2010, т. 62, № 8 COMPLETENESS OF INVARIANT IDEALS IN GROUPS 1027 |µd(B)− µd(xB)| = ∣∣|B ∩ Fdyd| − |B ∩ x−1Fdyd|∣∣ |Fd| ≤ ≤ |Fdyd4x −1Fdyd| |Fd| = |Fd4x−1Fd| |Fd| < 1 m ≤ 1 n . On the other hand, µd ∈ O(µ) implies |µd(B)− µd(xB)| ≥ ≥ |µ(B)− µ(xB)| − |µ(B)− µd(B)| − |µ(xB)− µd(xB)| > 1 n and this is a contradiction. 4. Take any A0 ∈ I and assume that µ(A0) 6= 0. Find n ∈ N such that µ(A0) > 1 n and consider the element d0 = (∅, n,A0) ∈ D. Since µ is a limit of the net (µd)d∈D, the neighborhood O(µ) = { η ∈ [0, 1]P(G) : η(A0) > 1 n } contains the measure µd for some d = (F,m,A) ≥ d0 in D. The choice of the point yd guarantees that Fdyd ∩ ∩A0 ⊂ Fdyd ∩A = ∅ and hence µd(A0) = |A0 ∩ Fdyd| |Fd| = 0 6> 1 n . Theorem 3.1 is proved. Theorem 3.1 implies that for an invariant ideal I in an amenable group G the inter- section NI = ⋂ Nµ⊃I Nµ is a well-defined ideal that contains I. In this definition µ runs over all Banach measures on G such that I ⊂ Nµ. The idealNI will be called theN -completion of the ideal I. An invariant ideal I is defined to be N -complete if I coincides with its N -completion NI . The N -completion N{∅} = ⋂ µNµ of the smallest ideal I = {∅} is denoted by N and called the ideal of absolute null sets. The ideal N is well-defined for each amenable group G. The following properties of N -complete ideals follows immediately from the defini- tion. Proposition 3.1. Let G be an amenable group. (1) For every Banach measure µ the ideal Nµ is N -complete. (2) For any N -complete ideals Iα, α ∈ A, in G the intersection ⋂ α∈A Iα is an N -complete ideal. (3) For each invariant ideal I in G the ideal NI is N -complete. (4) NI = N for each invariant ideal I ⊂ N . The S- and N -completions relate as follows. Theorem 3.2. For any invariant ideal I in an amenable group G we get I ⊂ ⊂ NI ⊂ SI . In particular, N ⊂ S. Proof. Assume conversely that some setA ∈ NI is not I-small. This means that there is a finite set F ⊂ G such that the complement G \ FA is not I-large. Consequently, the family I ∪ (G \ FA) generates an invariant ideal J = { J ⊂ G : J ⊂ I ∪ E(G \ FA) for some I ∈ I and E ∈ F } . ISSN 1027-3190. Укр. мат. журн., 2010, т. 62, № 8 1028 T. BANAKH, N. LYASKOVSKA By Theorem 3.1, there is a Banach measure µ onG such thatJ ⊂ Nµ. Then µ(G\FA) = = 0 and hence µ(FA) = 1 and µ(A) > 0. Now we see that I ⊂ NI ⊂ Nµ but A /∈ Nµ, which contradicts the choice of A ∈ NI . Theorem 3.2 is proved. Corollary 3.1. Each S-complete ideal in an amenable group in N -complete. 4. Packing indices. Let I be an invariant ideal of subsets of a group G. To each subset A of the group G we can assign the I-packing index I-pack(A) = sup { |B| : B ⊂ G is such that {bA}b∈B is disjoint modulo I } . An indexed family {Ab}b∈B of subsets of G is called disjoint modulo the ideal I if Ab ∩Aβ ∈ I for any distinct indices b, β ∈ B. If I = {∅} is the trivial ideal on G, then we write pack(A) instead of {∅}-pack(A). For example, the packing index pack(2Z) of the set A = 2Z of even numbers in the group G = Z is equal to 2. The same equality I-pack(2Z) = 2 holds for any ideal I on Z. It should be mentioned that in the definition of the I-packing index, the supremum cannot be replaced by the maximum: by [7], each infinite group G contains a subset A ⊂ G such that pack(A) ≥ ℵ0 but no infinite set B ⊂ G with disjoint {bA}b∈B exists. To catch the difference between sup and max, for a subset A ⊂ G let us consider a more informative cardinal characteristic I-Pack(A) = sup { |B|+ : B ⊂ G such that {bA}b∈B is disjoint modulo I } . It is clear that I-pack(A) ≤ I-Pack(A) and I-pack(A) = sup{κ : κ < I-Pack(A)}, so the value of I-pack(A) can be recovered from that of I-Pack(A). The packing indices {∅}-pack and {∅}-Pack were intensively studied in [7 – 10]. In fact, the I-packing indices I-pack(A) and I-Pack(A) are partial cases of the pack- ing indices I-packn(A) and I-Packn(A) defined for every cardinal number n ≥ 2 by the formulas: I-packn(A) = sup { |B| : B ⊂ G such that { ⋂ c∈C cA : C ∈ [B]n } ⊂ I } and I-Packn(A) = sup { |B|+ : B ⊂ G such that { ⋂ c∈C cA : C ∈ [B]n } ⊂ I } = = min { κ : ∀B ⊂ G ( |B| ≥ κ ⇒ ( ∃C ∈ [B]n with ⋂ c∈C cA /∈ I ))} , where [B]n stands for the family of all n-element subsets of B. It is clear that I-pack(A) = I-pack2(A) and I-Pack(A) = I-Pack2(A). Also, I-packn(A) ≤ I-packn+1(A) and I-Packn(A) ≤ I-Packn+1(A) for any finite n ≥ 2. ISSN 1027-3190. Укр. мат. журн., 2010, т. 62, № 8 COMPLETENESS OF INVARIANT IDEALS IN GROUPS 1029 If I = {∅}, then we shall write packn(A) and Packn(A) instead of {∅}-packn(A) and {∅}-Packn(A). The following example show that the difference between the packing indices pack2(A) and pack3(A) can be infinite. Example 4.1. The subset A = {n(n − 1)/2 : n ∈ N} of the group Z has pack2(A) = 1 and pack3(A) = ℵ0. The latter equality follows from the observation that the family {2m+A}m∈N is 3-disjoint in the sense that (2n+A)∩(2m+A)∩(2k+A) = ∅ for any pairwise distinct numbers n,m, k ∈ N. 5. Packing-complete ideals. It is clear that for each set A ∈ I and finite n ≥ 2 we get I-packn(A) = |G| and I-Packn(A) = |G|+. We shall be interested in ideals for which the converse implication holds. Definition 5.1. An invariant ideal I on a group G is called Packn-complete (resp. packn-complete) if I contains each set A ⊂ G with I-Packn(A) ≥ ℵ0 (resp. I-packn(A) ≥ ℵ0). Since Packn(A) ≤ Packn+1(A), each Packn+1-complete ideal is Packn-complete. Definition 5.2. An invariant ideal I on a group G is called Pack<ω-complete if I is Packn-complete for every n ≥ 2. For each ideal I in a group G and every n ≥ 2 we get the implications: Pack<ω-complete⇒ Packn-complete⇒ Pack2-complete⇒ pack2-complete. The simplest example of a Pack<ω-complete ideal is the ideal Nµ. Theorem 5.1. For any Banach measure µ on a group G the ideal Nµ = {A ⊂ ⊂ G : µ(A) = 0} is Pack<ω-complete. Proof. We need to show that the ideal Nµ is Packn-complete for every n ≥ 2. Take any subset A ⊂ G with Nµ-Packn(A) ≥ ℵ0. This means that for any natural number m there is a subset Bm of size m such that µ(b1A ∩ · · · ∩ bnA) = 0 for any distinct b1, . . . bn ∈ Bm. It follows that 1 ≥ µ ( ⋃ b∈Bm bA ) ≥ 1 n ∑ b∈Bm µ(bA) = 1 n ∑ b∈Bm µ(A) and hence µ(A) ≤ n |Bm| = n m . Since this equality holds for any m we conclude that µ(A) = 0 and hence A ∈ Nµ. Theorem 5.1 is proved. Since the intersection of Pack<ω-complete ideals is Pack<ω-complete, Theorems 5.1 and 3.2 imply Corollary 5.1. An invariant ideal I in an amenable group G is Pack<ω-complete provided I is N -complete or S-complete. Thus for each ideal I in an amenable group G and every n ≥ 2 we get the implica- tions: S-complete ⇒ N -complete ⇒ Pack<ω-complete ⇒ pack2-complete. The amenability assumption is essential in Corollary 5.1. Proposition 5.1. If a group G contains an isomorphic copy of the free group F2 with two generators, then G = A ∪ B is the union of two subsets with infinite pack2- index. Consequently, no ideal of G is pack2-complete. In particular, the ideal S of small subsets of G is not pack2-complete. ISSN 1027-3190. Укр. мат. журн., 2010, т. 62, № 8 1030 T. BANAKH, N. LYASKOVSKA Proof. Let a, b be the generators of the free subgroup F2 ⊂ G. Choose a subset S ⊂ G that meets each coset F2 ·g, g ∈ G, at a single point. We shall additionally assume that the singleton S ∩ F2 = {e} contains the neutral element of G. Each element of F2 can be uniquely written as an irreducible word in the alphabet {a, a−1, b, b−1}. Let Fa be the set of irreducible words that start with a or a−1. It is clear that F2 = Fa ∪ (F2 \ Fa) and thusG = A∪B whereA = FaS andB = G\A = (F2 \Fa)S. It remains to observe that the sets A, B have infinite pack2-index. Assuming that G contains a pack2-complete ideal I, we conclude that A,B ∈ I and hence G = A ∪B ∈ I, which is a contradiction. Proposition 5.1 is proved. 6. Pack<ω-completion of ideals. For an ideal I in a group G let Pack<ω(I) be the intersection of all invariant Pack<ω-complete ideals J ⊂ P(G) that contain I. If no such an ideal J exists, then we put Pack<ω(I) = P(G). The family Pack<ω(I) is called the Pack<ω-completion on I. If Pack<ω(I) 6= P(G), then Pack<ω(I) is a Pack<ω-complete ideal in G. By analogy, for every n ≥ 2 we can define the Packn-completion Packn(I) of I. Corollary 5.1 guarantees that for an invariant ideal I in an amenable group G its packing completions Pack<ω(I) and Packn(I) are ideals lying in the Pack<ω-complete ideal NI . On the other hand, for a group G containing a copy of the free group F2, the packing completions Packn(I) and Pack<ω(I) coincide with P(G). The following theorem describes the inner structure of the Pack<ω-completion. A subset A ⊂ P(G) is called additive if A ∪B for any sets A,B ∈ A. Proposition 6.1. The Pack<ω-completion Pack<ω(I) of an invariant ideal I on a group G is equal to the union I<ω1 = ⋃ α<ω1 Iα where I0 = I and Iα is the smallest additive family containing all subsets A ⊂ G with infinite index Iβ-Packn(A) for some β < α and n < ω. Proof. The inclusion Pack<ω(I) ⊃ I<ω1 follows from the fact that each Pack<ω- complete ideal which contains Iβ for all β < α also contains Iα. To show the equality we need to prove that I<ω1 is a Pack<ω-complete ideal if I<ω1 6= P(G). First we show that I<ω1 is Packn-complete for each n ≥ 2. Let A be subset of G with I<ω1 -Packn(A) ≥ ℵ0. It means that there is countable sequence (Bm)m∈ω of subsets Bm ∈ [G]m such that for any subset C ∈ [Bm]n the intersection ⋂ c∈C cA ∈ ∈ I<ω1 and thus ⋂ c∈C cA ∈ Iα(m,C) for some countable ordinal α(m,C) < ω1. Since the sequence (Bm) is countable and the sets Bm and C ∈ [Bm]n are finite, there is a countable ordinal α such that α > α(m,C) for each m and each C. For this ordinal α we get Iα-Packn(A) ≥ ℵ0. According to the definition of Iα+1 the set A ∈ Iα+1 ⊂ I<ω and thus I<ω1 is Packn-complete for each n ≥ 2. Proposition 6.1 is proved. By analogy we can describe the Packn-completion Packn(I) of I. Proposition 6.2. The Packn-completion Packn of an invariant ideal I on a group G is equal to the union I<ω1 = ⋃ α<ω1 Iα ISSN 1027-3190. Укр. мат. журн., 2010, т. 62, № 8 COMPLETENESS OF INVARIANT IDEALS IN GROUPS 1031 where I0 = I and Iα is the smallest additive family that contains all subsets A ⊂ G with infinite index Iβ-Packn(A) for some β < α. 7. Some open problems. For an invariant ideal I in an amenable group G and every n ≥ 2 we get the following chain of ideals: I ⊂ Pack2(I) ⊂ · · · ⊂ Packn(I) ⊂ Packn+1(I) ⊂ · · · ⊂ Pack<ω(I) ⊂ NI ⊂ SI . For the ideal I = S all these ideals coincide. What happens for the smallest ideal I0 = = {∅}? Problem 7.1. Are the ideals Packn(I0), n ≥ 2, in the group Z pairwise distinct? Do they differ from the ideal Pack<ω(I0)? Is Pack<ω(I0) 6= N? It should be mentioned thatN 6= S for each countable amenable group, see [11] (5.3). Problem 7.2. Give a combinatorial description of subsets lying in the ideals Pack<ω(I0) and Packn(I0), n ≥ 2. Each ideal of subsets of the group Z can be considered as a subspace of the Cantor cube {0, 1}Z. So, we can speak about topological properties of ideals. Problem 7.3. Describe the Borel complexity of the ideals Pack<ω(I0) and Packn(I0) for n ≥ 2. Are they coanalytic? Is the ideal N of absolute null sets co- analytic? Borel? Let us recall that a subspace C of a Polish space X is coanalytic if its complement X \C is analytic. The latter means that X \C is the continuous image of a Polish space, see [12]. It easy to show that the ideal S of small subsets of a countable group G is an Fσδ-subset of P(G). By Corollary 5.1, for the smallest ideal I0 = {∅} in an amenable group G its packing completion Pack<ω(I0) ⊂ S. On the other hand, Proposition 5.1 implies that Pack<ω(I0) = P(G) if G contains a copy of the free group F2. Problem 7.4 (I. Protasov). Is a group G amenable if Pack<ω(I0) ⊂ S? 8. Acknowledgment. The authors would like to thank Igor Protasov for careful and creative reading the manuscript and many valuable remarks and suggestions that substan- tially improved both the results and presentation of the paper. 1. Bella A., Malykhin V. I. On certain subsets of a group // Quest. Answers Gen. Topology. – 1999. – 17, № 2. – P. 183 – 197. 2. Bella A., Malykhin V. I. On certain subsets of a group II // Ibid. – 2001. – 19, № 1. – P. 81 – 94. 3. Protasov I. V. Maximal topologies on groups // Sib. Mat. Zh. – 1998. – 39, № 6. – P. 1368 – 1381. 4. Protasov I. V. Algebra in the Stone-Cech compactification: applications to topologies on groups // Algebra Discrete Math. – 2009. – № 1. – P. 83 – 110. 5. Paterson A. Amenability. – Providence, RI: Amer. Math. Soc., 1988. 6. Engelking R. General topology // Sigma Ser. Pure Math. – Berlin: Heldermann, 1989. – 6. 7. Banakh T., Lyaskovska N. Weakly P -small not P -small subsets in groups // Int. J. Algebra Comput. – 2008. – 18, № 1. – P. 1 – 6. 8. Banakh T., Lyaskovska N. Weakly P -small not P -small subsets in Abelian groups // Algebra Discrete Math. – 2006. – № 3. – P. 29 – 34. 9. Banakh T., Lyaskovska N., Repovs D. Packing index of subsets in Polish groups // Notre Dame J. Formal Logic. – 2009. – 50, № 4. – P. 453 – 468. 10. Lyaskovska N. Constructing subsets of a given packing index in Abelian groups // Acta Univ. carol. Math. et phys. – 2007. – 48, № 2. – P. 69 – 80 (ArXiv:0901.1151). 11. Lutsenko Ie., Protasov I. V. Sparse, thin and other subsets of groups // Int. J. Algebra Comput. – 2009. – 19, № 4. – P. 491 – 510. 12. Kechris A. Classical descriptive set theory. – Berlin: Springer, 1995. Received 09.03.10 ISSN 1027-3190. Укр. мат. журн., 2010, т. 62, № 8
id umjimathkievua-article-2934
institution Ukrains’kyi Matematychnyi Zhurnal
keywords_txt_mv keywords
language English
last_indexed 2026-03-24T02:33:06Z
publishDate 2010
publisher Institute of Mathematics, NAS of Ukraine
record_format ojs
resource_txt_mv umjimathkievua/c3/62aaef116461fe5287df2d158993bcc3.pdf
spelling umjimathkievua-article-29342020-03-18T19:40:46Z Completeness of invariant ideals in groups Повнота інваріантних ідеалів у групах Banakh, T. O. Lyaskovska, N. Банах, Т. О. Лясковська, Н. We introduce and study various notions of completeness of translation-invariant ideals in groups. Введено та досліджено різні поняття повноти інваріантних ідеалів у групах. Institute of Mathematics, NAS of Ukraine 2010-08-25 Article Article application/pdf https://umj.imath.kiev.ua/index.php/umj/article/view/2934 Ukrains’kyi Matematychnyi Zhurnal; Vol. 62 No. 8 (2010); 1022–1031 Український математичний журнал; Том 62 № 8 (2010); 1022–1031 1027-3190 en https://umj.imath.kiev.ua/index.php/umj/article/view/2934/2618 https://umj.imath.kiev.ua/index.php/umj/article/view/2934/2619 Copyright (c) 2010 Banakh T. O.; Lyaskovska N.
spellingShingle Banakh, T. O.
Lyaskovska, N.
Банах, Т. О.
Лясковська, Н.
Completeness of invariant ideals in groups
title Completeness of invariant ideals in groups
title_alt Повнота інваріантних ідеалів у групах
title_full Completeness of invariant ideals in groups
title_fullStr Completeness of invariant ideals in groups
title_full_unstemmed Completeness of invariant ideals in groups
title_short Completeness of invariant ideals in groups
title_sort completeness of invariant ideals in groups
url https://umj.imath.kiev.ua/index.php/umj/article/view/2934
work_keys_str_mv AT banakhto completenessofinvariantidealsingroups
AT lyaskovskan completenessofinvariantidealsingroups
AT banahto completenessofinvariantidealsingroups
AT lâskovsʹkan completenessofinvariantidealsingroups
AT banakhto povnotaínvaríantnihídealívugrupah
AT lyaskovskan povnotaínvaríantnihídealívugrupah
AT banahto povnotaínvaríantnihídealívugrupah
AT lâskovsʹkan povnotaínvaríantnihídealívugrupah