@ Appl. Gen. Topol. 20, no. 2 (2019), 431-447 doi:10.4995/agt.2019.11645 c© AGT, UPV, 2019 Balleans, hyperballeans and ideals Dikran Dikranjana, Igor Protasovb, Ksenia Protasovab and Nicolò Zavaa a Department of Mathematical, Computer and Physical Sciences, Udine University, 33 100 Udine, Italy (dikran.dikranjan@uniud.it, nicolo.zava@gmail.com) b Department of Computer Science and Cybernetics, Kyiv University, Volodymyrska 64, 01033, Kyiv, Ukraine (i.v.protasov@gmail.com, ksuha@freenet.com.ua) Communicated by M. Sanchis Abstract A ballean B (or a coarse structure) on a set X is a family of subsets of X called balls (or entourages of the diagonal in X × X) defined in such a way that B can be considered as the asymptotic counterpart of a uniform topological space. The aim of this paper is to study two con- crete balleans defined by the ideals in the Boolean algebra of all subsets of X and their hyperballeans, with particular emphasis on their con- nectedness structure, more specifically the number of their connected components. 2010 MSC: 54E15. Keywords: balleans; coarse structure; coarse map; asymorphism; balleans defined by ideals; hyperballeans. 1. Introduction 1.1. Basic definitions. A ballean is a triple B = (X,P,B) where X and P are sets, P 6= ∅, and B : X ×P →P(X) is a map, with the following properties: (i) x ∈ B(x,α) for every x ∈ X and every α ∈ P ; (ii) symmetry, i.e., for any α ∈ P and every pair of points x,y ∈ X, x ∈ B(y,α) if and only if y ∈ B(x,α); Received 13 May 2018 – Accepted 15 July 2018 http://dx.doi.org/10.4995/agt.2019.11645 D. Dikranjan, I. Protasov, K. Protasova and N. Zava (iii) upper multiplicativity, i.e., for any α,β ∈ P , there exists a γ ∈ P such that, for every x ∈ X, B(B(x,α),β) ⊆ B(x,γ), where B(A,δ) = ⋃ {B(y,δ) | y ∈ A}, for every A ⊆ X and δ ∈ P . The set X is called support of the ballean, P – set of radii, and B(x,α) – ball of centre x and radius α. This definition of ballean does not coincide with, but it is equivalent to the usual one (see [11] for details). A ballean B is called connected if, for any x,y ∈ X, there exists α ∈ P such that y ∈ B(x,α). Every ballean (X,P,B) can be partitioned in its connected components : the connected component of a point x ∈ X is QX(x) = ⋃ α∈P B(x,α). Moreover, we call a subset A of a ballean (X,P,B) bounded if there exists α ∈ P such that, for every y ∈ A, A ⊆ B(y,α). The empty set is always bounded. A ballean is bounded if its support is bounded. In particular, a bounded ballean is connected. Denote by [(X) the family of all bounded subsets of a ballean X. If B = (X,P,B) is a ballean and Y a subset of X, one can define the subballean B �Y = (Y,P,BY ) on Y induced by B, where BY (y,α) = B(y,α)∩Y , for every y ∈ Y and α ∈ P . A subset A of a ballean (X,P,B) is thin (or pseudodiscrete) if, for every α ∈ P, there exists a bounded subset V of X such that BA(x,α) = B(x,α) ∩ A = {x} for each x ∈ A\V . A ballean is thin if its support is thin. Bounded balleans are obviously thin. We note that to each ballean on a set X can be associated a coarse structure [12]: a particular family E of subsets of X×X, called entourages of the diagonal ∆X. The pair (X,E) is called a coarse space. This construction highlights the fact that balleans can be considered as asymptotic counterparts of uniform topological spaces. For a categorical look at the balleans and coarse spaces as “two faces of the same coin” see [4]. Definition 1.1 ([11, 5]). Let B = (X,P,B) be a ballean. A subset A of X is called: (i) large in X if there exists α ∈ P such that B(A,α) = X; (ii) thick in X if, for every α ∈ P , there exists x ∈ A such that B(x,α) ⊆ A; (iii) small in X if, for every α ∈ P , X \B(A,α) is large in X. Let BX = (X,PX,BX) and BY = (Y,PY ,BY ) be two balleans. Then a map f : X → Y is called (i) coarse if for every radius α ∈ PX there exists another radius β ∈ PY such that f(BX(x,α)) ⊆ BY (f(x),β) for every point x ∈ X; (ii) effectively proper if for every α ∈ PY there exists a radius β ∈ PX such that f−1(BY (f(x),α)) ⊆ BX(x,β) for every x ∈ X; c© AGT, UPV, 2019 Appl. Gen. Topol. 20, no. 2 432 Balleans, hyperballeans and ideals (iii) a coarse embedding if it is both coarse and effectively proper; (iv) an asymorphism if it is bijective and both f and f−1 are coarse or, equiv- alently, f is bijective and both coarse and effectively proper; (v) an asymorphic embedding if it is an asymorphism onto its image or, equiv- alently, if it is an injective coarse embedding; (vi) a coarse equivalence if it is a coarse embedding such that f(X) is large in BY . We recall that a family I of subsets of a set X is an ideal if A,B ∈I, C ⊆ A imply A∪B ∈I, C ∈I. In this paper, we always impose that X /∈I (so that I is proper) and I contains the ideal FX of all finite subsets of X. Because of this setting, a set X that admits an ideal I is infinite, as otherwise X ∈I. We consider the following two balleans with support X determined by I. Definition 1.2. (i) The I-ary ballean XI-ary = (X,I,BI-ary), with radii set I and balls defined by BI-ary(x,A) = {x}∪A, for x ∈ X and A ∈I; (ii) The point ideal ballean XI = (X,I,BI), where BI(x,A) = {{x} if x /∈ A, {x}∪A = A otherwise. The balleans XI-ary and XI are connected and unbounded. While XI is thin, XI-ary is never thin (this follows from Proposition 1.3 and results from [11] reported in Theorem 2.2). For every connected unbounded ballean B with support X one can define the satellite ballean XI, where I = [(X) is the ideal of all bounded subsets of X. Proposition 1.3. For every ideal I on a set X, the map idX : XI → XI-ary is coarse, but it is not effectively proper. Proof. Pick an arbitrary non-empty element F ∈I. Since I is a proper ideal, for every K ∈I, there exists xK ∈ X \ (F ∪K). Hence, in particular, BI-ary(xK,F) = {xK}∪F 6⊆ {xK} = BI(xK,K). � Let B = (X,P,B) be a ballean. Then the radii set P can be endowed with a preorder ≤B as follows: for every α,β ∈ P , α ≤B β if and only if B(x,α) ⊆ B(x,β), for every x ∈ X. A subset P ′ ⊆ P is cofinal if it is cofinal in this preorder (i.e., for every α ∈ P , there exists α′ ∈ P ′, such that α ≤B α′). If P ′ is cofinal, then B = (X,P ′,B′), where B′ = B �X×P′. If I is an ideal on a set X, then both the preorders ≤BI and ≤BI-ary on I coincide with the natural preorder ⊆ on I, defined by inclusion. c© AGT, UPV, 2019 Appl. Gen. Topol. 20, no. 2 433 D. Dikranjan, I. Protasov, K. Protasova and N. Zava Remark 1.4. Let (X,PX,BX) and (Y,PY ,BY ) be two balleans and f : X → Y be an injective map. We want to give some sufficient conditions that implies the effective properness of f. (i) Suppose that there exist two cofinal subsets of radii P ′X and P ′ Y of PX and PY , respectively, and a bijection ψ : P ′ X → P ′ Y such that, for every α ∈ P ′ X and every x ∈ X, (1.1) f(BX(x,α)) = BY (f(x),ψ(α)) ∩f(X). We claim that, under these hypothesis, f is a coarse embedding and then f : X → f(X) is an asymorphism. First of all, let us check that f is coarse. Fix a radius α ∈ PX and let α′ ∈ P ′X such that α ≤ α ′. Hence f(BX(x,α)) ⊆ f(BX(x,α′)) ⊆ BY (f(x),ψ(α′)), for every x ∈ X, where the last inclusion holds because of (1.1). As for the effective properness, since f is bijective, (1.1) is equivalent to BX(x,α) = f −1(BY (f(x),ψ(α))), for every x ∈ X, and this yelds to the thesis. In fact, for every β ∈ PY , there exists α′ ∈ P ′X such that β ≤ ψ(α ′) and thus, for every x ∈ X, f−1(BY (f(x),β)) ⊆ f−1(BY (f(x),ψ(α′))) = BX(x,α′). (ii) Note that f : X → Y is a coarse embedding if and only if f : X → f(X) is a coarse embedding, where f(X) is endowed with the subballean structure inherited by Y . Suppose that P ′X ⊆ PX and P ′′ f(X) ⊆ PY are cofinal subsets of radii in X and f(X), respectively, and ψ : P ′X → P ′′ f(X) is a bijection such that (1.1) holds for every x ∈ X. Then f is a coarse embedding. (iii) In notations of item (ii), in order to show that P ′′ f(X) is cofinal in f(X), it is enough to provide a cofinal subset of radii P ′Y ⊆ PY in Y and a bijection ϕ: P ′′ f(X) → P ′Y such that, for every y ∈ f(X) and every α ∈ P ′′ f(X) , BY (y,α)∩ f(X) = BY (y,ϕ(α)) ∩f(X). 1.2. Hyperballeans. Definition 1.5. Let B = (X,P,B) be a ballean. Define its hyperballean to be exp(B) = (P(X),P, exp B), where, for every A ⊆ X and α ∈ P , (1.2) exp B(A,α) = {C ∈P(X) | A ⊆ B(C,α), C ⊆ B(A,α)}. It is not hard to check that this defines actually a ballean. Another easy observation is the following: for every ballean (X,P,B), Qexp X(∅) = {∅} and, in particular exp B({∅},α) = {∅} for every α ∈ P , since B(∅,α) = ∅. Motivated by this, we shall consider also the subballean exp∗(X) = exp(X) \ {∅}. If B = (X,P,B) is a ballean, the subballean X[ of expB having as support the family of all non-empty bounded subsets of B was already defined and c© AGT, UPV, 2019 Appl. Gen. Topol. 20, no. 2 434 Balleans, hyperballeans and ideals studied in [10]. Note that, B is connected (resp., unbounded) if and only if B[ is connected (resp., unbounded). In the sequel we focus our attention on four hyperballeans defined by an ideal I on a set X. In particular, we investigate exp XI and exp XI-ary, as well as their subballeans X[I-ary and X [ I. So exp XI = (P(X),I, exp BI), and, according to (1.2), for A ⊆ X and K ∈I one has (1.3) exp BI(A,K) = {{(A\K) ∪Y | ∅ 6= Y ⊆ K} if A∩K 6= ∅, {A} otherwise. In fact, fix C ∈ exp BI(A,K). If A∩K = ∅, then C = A (as BI(A,K) = A and, for every A′ ⊆ A, BI(A′,K) = A′). Otherwise, C ⊆ BI(A,K) = A∪K. Moreover, A ⊆ BI(C,K) if and only if C ∩K 6= ∅ and A ⊆ C ∪K. In other words, exp BI(A,K) = {Z ∈P(X) | A\K ( Z ⊆ A∪K}, if A∩K 6= ∅. Let us now compute the balls in exp XI-ary = (P(X),I, exp BI-ary). As mentioned above, exp BI-ary({∅},K) = {∅} for every K ∈I. Fix now a non- empty subset A of X and a radius K ∈ I. Then a non-empty subset C ⊆ X belongs to exp BI-ary(A,K) if and only if C ⊆ BI-ay(A,K) = A∪K and A ⊆ BI-ary(C,K) = C ∪K, since both A and C are non-empty. Hence exp BI-ary(A,K) = {(A\K) ∪Y | Y ⊆ A, (A\K) ∪Y 6= ∅} = = {Z ∈P(X) | A\K ⊆ Z ⊆ A∪K, Z 6= ∅} for every ∅ 6= A ⊆ X and K ∈I. By putting all together, one obtains that, for every A ⊆ X and every K ∈I, (1.4) exp BI-ary(A,K) = { {Z ∈P(X) | A\K ⊆ Z ⊆ A∪K, Z 6= ∅} if A 6= ∅, {A} otherwise A = ∅. Remark 1.6. Denote by CX = {0, 1}X the Boolean ring of all function X → {0, 1} = Z2 and for f ∈ CX let supp f = {x ∈ X | f(x) = 1}. Then one has a ring isomorphism  = X : P(X) → CX, sending A ∈P(X) to its characteristic function χA ∈ CX, so (∅) = 0, the zero function. Using , one can transfer the ball structure from exp BI-ary to CX: for 0 6= f ∈ {0, 1}X and A ∈ I one has (exp BI-ary( −1(f),A)) = {g | g(x) = f(x), x ∈ X \A} = {g | g �X\A= f �X\A}.(1.5) c© AGT, UPV, 2019 Appl. Gen. Topol. 20, no. 2 435 D. Dikranjan, I. Protasov, K. Protasova and N. Zava While, according to (1.3) and (1.4), the empty set is “isolated” in both balleans exp XI and XI-ary, the set {g | g(x) = 0, x ∈ X \A} = {g | g[X \A] = {0}} (i.e., the functions g with supp g ⊆ A), still makes sense and seems a more natural candidate for a ball of radius A centered at the zero function. Taking into account this observation, we modify the ballean structure on CX, denoting by C(X,I) the new ballean, with balls defined by the unique formula suggested by (1.5): (1.6) BC(X,I)(f,A) = {g | g(x) = f(x), x ∈ X \A} = {g | g �X\A= f �X\A}, where A ∈ I; when no confusion is possible, we shall write shortly BC(f,A). In this way (1.7)  �exp∗(XI-ary): exp ∗(XI-ary) → C(X,I) is an asymorphic embedding. The ballean C(X,I), as well as its subballean M(X,I), having as support the ideal {g ∈ CX | supp g ∈ I} of the ring CX, will play a prominent role in the paper (note that M(X,I)\{∅} coincides with (X[I−ary)). 1 If X = N and I = FN, then M(X,I) is the Cantor macrocube defined in [10]. Motivated by this, the ballean M(X,I), for an ideal I on a set X, will be called the I-macrocube (o, shortly, a macrocube) in the sequel. Remark 1.7. One of the main motivations for the above definitions comes from the study of topology of hyperspaces. For an infinite discrete space X, the set P(X) admits two standard non-discrete topologizatons via the Vietoris topology and via the Tikhonov topology. In the case of the Vietoris topology, the local base at the point Y ∈ P(X) consists of all subsets of X of the form {Z ∈ P(X) | K ⊆ Z ⊆ Y}, where K runs over the family of all finite subsets of Y . The Tikhonov topology arises after identification of P(X) with {0, 1}X via the characteristic functions of subsets of X. Given an ideal I on X, the point ideal ballean XI can be considered as one of the possible asymptotic versions of the discrete space X, see Section 2. With these observations, one can look at exp XI as a counterpart of the Vietoris hyperspace of X, and the Tikhonov hyperspace of X has two counterparts C(X,I) and exp XI-ary. These parallels are especially evident in the case of the ideal FX of finite subsets of X. 1.3. Main results. In this paper we focus on hyperballeans of balleans defined by means of ideals, these are the point ideal balleans and the I-ary balleans. It is known ([11]) that the point ideal balleans are precisely the thin balleans. Inspired by this fact, in §2, we give some further equivalent properties (Theorem 2.2). 1Sometimes we refer to C(X,I) as the I-Cartesian ballean. Its ballean structure makes both ring operations on C(X,I) coarse maps, while exp(XI-ary) fails to have this property. c© AGT, UPV, 2019 Appl. Gen. Topol. 20, no. 2 436 Balleans, hyperballeans and ideals As already anticipated the main objects of study will be the hyperballeans exp(XI), exp(XI-ary) and C(X,I), where I is an ideal of a set X. By re- striction, we will gain also knowledge of their subballeans X[I, X [ I-ary and the I-macrocube M(X,I). Since XI, XI-ary and C(X,I) are pairwise different, it is natural to ask whether their hyperballeans are different or not. Section §3 is devoted to answering this question, comparing these three balleans from various points of view. In particular, we prove that exp(XI) and exp(XI-ary) are different (Corollary 3.2), although they have asymorphic subballeans (The- orem 3.3), the same holds for the pair exp XI-ary and C(X,I). Moreover, we show that C(X,I) (and in particular, exp∗(XI-ary)) is coarsely equivalent to a subballean of exp(XI); so M(X,I) (and in particular, X[I-ary) is coarsely equivalent to a subballean of X[I The final part of the section is dedicated to a special class of ideals defined as follows. For a cardinal κ and λ ≤ κ consider the ideal Kλ = {F ⊆ κ | |F| < λ} of κ. A relevant property of this ideal is homogeneity (i.e., it is invariant under the natural action of the group Sym(κ) by permutations of κ 2). For the sake of brevity denote by K the ideal Kκ of κ. Theorem 3.5 pro- vides a bijective coarse embedding of a subballean of exp(κK) into exp(κK-ary) and, under the hypothesis of regularity of κ, also exp(κK) itself asymorphically embeds into exp(κK-ary). To measure the level of disconnectedness of a ballean B, one can consider the number dsc(B) of connected components of B. Although the two hyper- balleans exp(XI) and exp(XI-ary) are different, they have the same connected components and in particular, dsc(exp(XI)) = dsc(exp(XI-ary)). Moreover, this cardinal coincides with dsc(C(X,I)) + 1 (Proposition 4.1). The main goal of Section §4 is to compute the cardinal number dsc(C(X,I)). To this end we use a compact subspace I∧ of the Stone-Čech remained βX\X of the discrete space X. In this terms, dsc(C(X,I)) = w(I∧). 2. Characterisation of thin connected balleans Let B = (X,P,B) be a bounded ballean. Then B is thin. Moreover, [(X) = P(X), while every proper subset of X is non-thick and the only small subset is the empty set. Hence, we now focus on unbounded balleans. It is known ([11]) that a connected unbounded ballean B is thin if and only if the identity mapping of X defines an asymorphism between B and its satellite ballean. It was also shown that these properties are equivalent to having all functions f : X → {0, 1} being slowly oscillating (such a function is called slowly oscillating if, for every α ∈ P there exists a bounded subset V such that |f(B(x,α))| = 1 for 2Consequently, all these permutations become automatically asymorphisms, once we en- dow κ with the point ideal ballean or the Kλ-ary ideal structure. One can easily see that these are the only homogeneous ideals of κ. c© AGT, UPV, 2019 Appl. Gen. Topol. 20, no. 2 437 D. Dikranjan, I. Protasov, K. Protasova and N. Zava each x ∈ X\V ; this is a specialisation (for {0, 1}-valued functions) of the usual more general notion, [11]). Theorem 2.2 provides further equivalent properties. For a ballean B = (X,P,B), we define a mapping C : X →P(X) by C(x) = X \{x}. Lemma 2.1. Let B = (X,P,B) be a connected unbounded ballean. If Y is a subset of X, then C(Y ) is bounded in exp(B) if and only if there exists α ∈ P such that |B(y,α)| > 1, for every y ∈ Y . Proof. (→) Since C(Y ) is bounded in exp(B), there exists α ∈ P such that, for every x,y ∈ Y with x 6= y, C(y) ∈ exp B(C(x),α). Hence y ∈ X \{x} ⊆ B(X\{y},α) and x ∈ X\{y}⊆ B(X\{x},α), in particular, y ∈ B(Y \{y},α) and x ∈ B(Y \{x},α), from which the conclusion descends. (←) Since, for every y ∈ Y , there exists z ∈ Y \{y} such that y ∈ B(z,α), C(y) ∈ exp B(X,α). Hence C(Y ) ⊆ exp B(X,α), and the latter is bounded. � If B is a ballean, denote by BM the subballean of expB whose support is the family of all non-thick non-empty subsets of X. If B is unbounded, then so it is BM. Moreover, B[ is a subballean of BM. This motivation for the choice of M comes from the fact that non-thick subsets3 were called meshy in [5] (this term will not be adopted here). Theorem 2.2. Let B = (X,P,B) be an unbounded connected ballean. Then the following properties are equivalent: (i) B is thin; (ii) B = BI, where I = [(X), i.e., B coincides with its satellite ballean; (iii) if A ⊆ X is not thick, then A is bounded; (iv) BM is connected; (v) the map C : X →P(X) is an asymorphism between X and C(X); (vi) every function f : X →{0, 1} is slowly oscillating. Proof. The implication (iii)→(iv) is trivial, since item (iii) implies that BM = B[ and the latter is connected. Furthermore, (i)↔(ii) and (i)↔(vi) have already been proved in [11]. (iv)→(iii) Assume that A ⊆ X is not thick. Fix arbitrarily a point x ∈ X. The singleton {x} is bounded, hence non-thick. By our assumption, BM is connected and both A and {x} are non-thick, so there must be a ball centred at x and containing A. Therefore, A is bounded. (v)→(i) If B is not thin then there is an unbounded subset Y of X satis- fying Lemma 2.1. Since C(Y ) is bounded in expB, we see that C is not an asymorphism. (ii)→(v) On the other hand, suppose that B = BI. Fix a radius V ∈ I. Without loss of generality, suppose that V has at least two elements. Now, 3or, equivalently, those subsets whose complement is large c© AGT, UPV, 2019 Appl. Gen. Topol. 20, no. 2 438 Balleans, hyperballeans and ideals pick an arbitrary point x ∈ X. If x ∈ V , then C(BI(x,V )) = {X \{y} | y ∈ V}={A ∈ C(X) | X \ (V ∪{x}) ( A ⊆ X} = expBI(C(x),V ) ∩C(X). If, otherwise, x /∈ V , then C(BI(x,V )) = {X \{x}}={A ∈ C(X) | (X \{x}) \V ( A ⊆ X \{x}} = expBI(C(x),V ) ∩C(X). (i)→(iii) Suppose that B is thin and A is an unbounded subset of X. We claim that A is thick. Fix a radius α ∈ P and let V ⊆ X be a bounded subset of X such that B(x,α) = {x}, for every x /∈ V . Since A is unbounded, there exists a point xα ∈ A\V . Hence B(xα,α) = {xα}⊆ A, which shows that A is thick. (iii)→(vi) Assume that X does not satisfy (vi), i.e, X has a non-slowly- oscillating function f : X → {0, 1}. Take a radius α such that, for every bounded V , there exists x ∈ X \ V such that |f(B(x,α))| = 2. Hence A = {x ∈ X | |f(B(x,α))| = 2} is unbounded. Decompose A as the disjoint union of A0 = {x ∈ A | f(x) = 0} and A1 = {x ∈ A | f(x) = 1}. Since A = A0 ∪A1, either A0 or A1 is unbounded. Moreover, for every x ∈ A, both A0 ∩ B(x,α) 6= ∅ and A1 ∩ B(x,α) 6= ∅ and thus A0 and A1 are not thick. � Remark 2.3. (i) Let us see that one cannot weaken item (iii) in the above theo- rem by replacing “non-thick” by the stronger property “small”. In other words, a ballean need not be thin provided that all its small subsets are bounded. To this end consider the ω-universal ballean (see [11, Example 1.4.6]): an infinite countable set X, endowed with the radii set P = {f : X → [X]<∞ | x ∈ f(x), {y ∈ X | x ∈ f(y)}∈ [X]<∞, ∀x ∈ X}, and B(x,f) = f(x), for every x ∈ X and f ∈ P . Since it is maximal (i.e., it is connected, unbounded and every properly finer ballean structure is bounded) by [11, Example 10.1.1], then every small subset is finite (by application of [11, Theorem 10.2.1]), although it is not thin. (ii) Let B be an unbounded connected ballean and X be its support. Con- sider the map CB : B[ → expB such that CB(A) = X \A, for every bounded A. It is trivial that C = CB �X, where X is identified with the family of all its singletons. Hence, if CB is an asymorphic embedding, then C is an asymorphic embedding too, and thus B is thin, according to Theorem 2.2. However, we claim that CB is not an asymorphic embedding if B is thin and then item (v) in Theorem 2.2 cannot be replaced with this stronger property. Since B is thin, we can assume that B coincides with its satellite BI (Theorem 2.2). Fix a radius V ∈I of exp XI and suppose, without loss of generality, that V has at least two elements. For every radius W ∈ I of X[I, pick an element c© AGT, UPV, 2019 Appl. Gen. Topol. 20, no. 2 439 D. Dikranjan, I. Protasov, K. Protasova and N. Zava AW ∈I such that AW ⊆ X \(W ∪V ). Hence, CB−1(exp BI(CB(AW ),V )) 6⊆ B[I(AW ,W) = {AW}, which implies that CB is not effectively proper. In fact, since AW ∪V ∈I, exp BI(CB(AW ),V ) = {Z ⊆ X | X \ (AW ∪V ) ( Z ⊆ X \AW}⊆ CB(X[I), and thus |exp BI(CB(AW ),V ) ∩CB(X[I)| > 1. A characterization of thin (and coarsely thin) balleans in terms of asymp- totically isolated balls can be found in [8, Theorems 1, 2]. 3. Further properties of exp(XI), exp(XI-ary) and C(X,I) Let f : X → Y be a map between sets. Then there is a natural definition for a map exp f : P(X) → P(Y ), i.e., exp f(A) = f(A), for every A ⊆ X. If f : X → Y is a map between two balleans such that f(A) ∈ [(Y ), for every A ∈ [(Y ) (e.g., a coarse map), then the restriction f[ = exp f �X[ : X[ → Y [ is well-defined. The following proposition can be easily proved. Proposition 3.1. Let BX = (X,PX,BX) and BY = (Y,PY ,BY ) be two bal- leans and let f : X → Y be a map between them. Then: (i) f : BX → BY is coarse if and only if exp f : expBX → expBY is coarse if and only if f[ : B[X →B [ Y is well-defined and coarse; (ii) f : BX → BY is a coarse embedding if and only if exp f : expBX → expBY is a coarse embedding if and only if f[ : B[X →B [ Y is well-defined and a coarse embedding. For the sake of simplicity, throughout this section, for every ideal I of a set X, the ballean C(X,I) will be identified with −1(C(X,I)) (where  is defined in Remark 1.6), whose support is P(X). Hence, by this identification, if A ⊆ X and K ∈I, BC(A,K) = {Y | A\K ⊆ Y ⊆ A∪K}. Corollary 3.2. For every ideal I on X, the following statements hold: (i) j = exp idX : exp XI → exp XI-ary is coarse, but it is not an asymor- phism; (ii) j : exp XI-ary → C(X,I) is coarse, but it is not an asymorphism; (iii) the same holds for the restriction i = id[X : X [ I → X [ I-ary. Proof. Since idX : XI → XI-ary is coarse, but it is not effectively proper (Proposition 1.3), items (i) and (iii) follow from Propositions 3.1. Item (ii) descends from the fact that exp(XI-ary) �P(X)\{∅}= C(X,I) �P(X)\{∅}, and Qexp XI-ary (∅) = {∅}, while QC(X,I)(∅) = I. � In spite of Corollary 3.2, we show now that a cofinal part of exp(XI) asy- morphically embeds in exp(XI-ary). c© AGT, UPV, 2019 Appl. Gen. Topol. 20, no. 2 440 Balleans, hyperballeans and ideals For every ideal I on X and x ∈ X consider the families Ux = {U ⊆ X | x ∈ U}, the principal ultrafilter of P(X) generated by {x}, and Ix = Ux ∩I = {F ∈I | x ∈ F}. Theorem 3.3. For every ideal I on X and x ∈ X, the following statements hold: (i) if j : exp(XI) → exp(XI-ary) is the map defined in Corollary 3.2(i), its restriction j �Ux is an asymorphism between the corresponding subbal- leans; (ii) C(X,I) and, in particular, exp∗(XI-ary) are coarsely equivalent to the subballean of exp(XI) with support Ux, witnessed by the map j �Ux : exp(XI) �Ux→ exp ∗(XI-ary) ⊆ C(X,I). Proof. (i) For every C ∈Ux and A ∈Ix, we have exp BI(C,A) ∩Ux = {(C \A) ∪Y | x ∈ Y ⊆ A} = exp BI-ary(C,A) ∩Ux, since C ∩A 6= ∅. Hence the conclusion follows by Remark 1.4(i), since Ix is a cofinal subset of radii of I. (ii) In view of item (i), it remains to see that j(Ux) is large in C(X,I). Indeed, for every A ∈Ux, BC(A,{x}) = {A,A\{x}}, and so BC(j(Ux),{x}) = C(X,I), where {x}∈I. � Since X[I, X [ I-ary, and M(X,I) are subballeans of exp(XI), exp(XI-ary), and C(X,I) respectively, by taking by restrictions we obtain the following immediate corollary. Corollary 3.4. For every ideal I on X and x ∈ X, the following statements hold: (i) j �Ix is an asymorphism between the corresponding subballeans of X [ I and X[I-ary; (ii) M(X,I) and, in particular, X[I-ary are coarsely equivalent to the sub- ballean of X[I with support Ix, witnessed by the map j �Ix : X [ I �Ix→ X[I-ary ⊆ M(X,I). 3.1. C(κ,K) and the hyperballeans exp(κK) and exp(κK-ary). Now we fo- cus our study on some more specific ideals. For an infinite cardinal κ and for its ideal K = [κ]<κ = {Z ⊂ κ | |Z| < κ}, consider the two balleans κK and κK-ary. Here we investigate some relationships between hyperballeans of those two balleans and the ballean C(κ,K). Furthermore, with κ as above, if x < κ, put U≥x = {A ⊆ κ | min A = x} and K≥x = U≥x ∩K = {A ∈K | min A = x}. For every pair of ordinals α ≤ β < κ, let [α,β] = {γ ∈ κ | α ≤ γ ≤ β}. Clearly, the cardinal κ is regular if and only if the family Pint = {[0,α] | α < κ} is c© AGT, UPV, 2019 Appl. Gen. Topol. 20, no. 2 441 D. Dikranjan, I. Protasov, K. Protasova and N. Zava cofinal in K. For a ballean B = (X,P,B), two subsets A,B of X are called close if A and B are in the same connected component of expB. Theorem 3.5. Let κ be an infinite cardinal and x < κ. Then: (i) the subballean U≥x of exp(κK) is asymorphic to C(κ,K), so exp∗(κK) is the disjoint union of κ pairwise close copies of C(κ,K); (ii) if κ is regular, then exp(κK) asymorphically embeds into exp(κK-ary). Proof. (i) Fix a bijection g : κ →A, where A is the family of all ordinals α such that x < α < κ. Define a map f : C(κ,K) → U≥x ⊆ exp(κK) such that, for every X ⊆ κ, f(X) = g(X)∪{x}. We claim that f is the desired asymorphism. In order to prove it, we want to apply Remark 1.4(ii). Fix a radius K ∈K (i.e., |K| < κ). Then, for every X ⊆ κ, f(BC(X,K)) = f({Y ⊆ κ | X \K ⊆ Y ⊆ X ∪K}) = = {f(g−1(Z)) | X \K ⊆ g−1(Z) ⊆ X ∪K} = = {Z ∪{x} | g(X) \g(K) ⊆ Z ⊆ g(X) ∪g(K)} = = {W ∈U≥x | (g(X) ∪{x}) \ (g(K) ∪{x}) ( W ⊆ ⊆ (g(X) ∪{x}) ∪ (g(K) ∪{x})} = = exp BK(g(X) ∪{x},g(K) ∪{x}) ∩U≥x = exp BK(f(X),g(K) ∪{x}) ∩U≥x. If we show that {g(K) ∪ {x} | K ∈ K} is cofinal in f(κ) = U≥x, then the conclusion follows, since we can apply Remark 1.4(ii) by putting ψ(K) = g(K)∪ {x}, for every K ∈K. It is enough to check that, for every X ⊆ κ and K ∈K, exp BK(X,K) ∩U≥x = exp BK(X,ψ(K)) ∩U≥x, which proves the cofinality of ψ(K) in f(κ), in virtue of Remark 1.4(iii). The last assertion of item (i) follows from the facts that the family {U≥x | x < κ} is a partition of exp(κK), and, for every x < y < κ, U≥x ∈ exp BK(U≥y, [x,y]), where [x,y] ∈K. (ii) Every ordinal α ∈ κ can be written uniquely as α = β + n, where β is a limit ordinal and n is a natural number. We say that α is even (odd) if n is even (odd). We denote by E the set of all odd ordinals from κ and fix a monotonically increasing bijection ϕ: κ → E. For each non-empty F ⊆ κ, let yF ∈ κ such that yF +1 = min ϕ(F) and define f(F) = {yF}∪ϕ(F). Moreover, we set f(∅) = ∅. Let S = f(exp(κK)). Hence the elements of S are the empty set and those subsets A of κ, consisting of odd ordinals and precisely one even ordinal α ∈ A such that α = min A. We claim that f : exp(κK) → S is an asymorphism. Since κ is regular, Pint ⊆K is a cofinal subset of radii. Now fix [0,α] ∈ Pint. Take an arbitrary subset A of κ. We can assume A to be non-empty, since in that case, there is nothing to be proved. The thesis follows, once we prove that c© AGT, UPV, 2019 Appl. Gen. Topol. 20, no. 2 442 Balleans, hyperballeans and ideals (3.1) f(exp BK(A, [0,α])) = exp BK-ary(f(A), [0,ϕ(α)]) ∩S, since we can apply Remark 1.4(i) if we define the bijection ψ([0,β]) = [0,ϕ(β)], for every β < κ, between cofinal subsets of radii. If A ∩ [0,α] = ∅, then also f(A) and [0,ϕ(α)] are disjoint, which implies that exp BK-ary(f(A), [0,ϕ(α)]) ∩S = {f(A)}. Otherwise, suppose that A and [0,α] are not disjoint. In particular yA ∈ [0,ϕ(α)]. We divide the proof of (3.1) in this case in some steps. First of all we claim that, for every ∅ 6= Z ⊆ κ, (3.2) if A\ [0,α] ⊆ Z ⊆ A∪ [0,α], then: Z 6= A\ [0,α] if and only if yZ ≤ ϕ(α). In fact, if Z = A \ [0,α], then min Z > α and so min ϕ(Z) > ϕ(α). Since ϕ(α) ∈ E, ϕ(Z) ⊆ E, and yZ /∈ E, we have that yZ > ϕ(α). Conversely, if Z 6= A \ [0,α], there exists z ∈ Z ∩ [0,α], since Z ⊆ A ∪ [0,α]. Hence min Z ≤ z ≤ α and thus yZ < min ϕ(Z) ≤ ϕ(α). Fix now a subset Z ⊆ κ. If f(Z) ∈ exp BK-ary(f(A), [0,ϕ(α)]), then, by applying the definitions, ϕ(A)\[0,ϕ(α)] = f(A)\[0,ϕ(α)] ⊆ ϕ(Z)∪{yZ}⊆ f(A)∪[0,ϕ(α)] = ϕ(A)∪[0,ϕ(A)]. Note that ϕ(Z) ⊆ E and yZ /∈ E. Hence ϕ(A\ [0,α]) = ϕ(A) \ϕ([0,α]) ⊆ ϕ(Z) ⊆ ϕ(A) ∪ϕ([0,α]) = ϕ(A∪ [0,α]) and yZ ≤ ϕ(α). Since ϕ is a bijection, we can apply (3.2) and obtain that A \ [0,α] ( Z ⊆ A∪ [0,α], which means that Z ∈ exp BK(A, [0,α]). Hence we have proved the inclusion (⊇) of (3.1). Since all the previous implications can be reverted, then (3.1) finally follows. � Corollary 3.6. Let κ be an infinite cardinal and x < κ. Then: (i) the subballean K≥x of κ[K is asymorphic to M(κ,K), so κ [ K is the disjoint union of κ pairwise close K-macrocubes M(κ,K); (ii) if κ is regular then κ[K asymorphically embeds into κ [ K-ary. The proof of item (ii), specified for κ = ω, can be found in [10]. 4. The number of connected components of exp(XI) Recall that dsc(B) denotes the number of connected components of a ballean B. Clearly, (4.1) dsc(exp(B)) = dsc(exp∗(B)) + 1 ≥ 2 for every non-empty ballean B. We begin with the following crucial observation. c© AGT, UPV, 2019 Appl. Gen. Topol. 20, no. 2 443 D. Dikranjan, I. Protasov, K. Protasova and N. Zava Proposition 4.1. For an ideal I on a set X, one has (i) the non-empty subsets Y,Z of X are close in exp(XI−ary) if and only if Y4Z ∈I; (ii) two functions f,g ∈ CX are close in C(X,I) if and only if supp f4supp g ∈ I; (iii) for every A ⊆ X, Qexp(XI)(A) = Qexp(XI−ary)(A), and in particular, (4.2) dsc(exp(XI)) = dsc(exp(XI-ary)) and dsc(exp∗(XI)) = dsc(exp ∗(XI-ary)) = dsc(C(X,I)) (iv) dsc(exp(XI)) = dsc(C(X,I)) + 1. Proof. (i) Two non-empty subsets Y and Z of exp(XI-ary) are close if and only if there exists K ∈I such that Y ∈ exp BI-ary(Z,K), i.e., (4.3) Y ⊆ Z ∪K and Z ⊆ Y ∪K. If (4.3) holds, then Y4Z = (Y \Z) ∪ (Z \Y ) ⊆ ((Z ∪K) \Z) ∪ ((Y ∪K) \Y ) = K ∈I. Conversely, if Y4Z ∈I, then K = Y4Z trivially satisfies (4.3). (ii) Let Y = supp f and Z = supp g. If both f,g are non-zero, then Y,Z are non-empty and the assertion follows from (i) and the asymorphism between exp∗(XI-ary) and C(X,I) \ {0}. If g = 0, then f is close to g if and only if f ∈ (I), i.e., Y = −1(f) ∈I. As Z = ∅, this proves the assertion in this case as well. (iii) Fix a subset A of X. The inclusion Qexp(XI)(A) ⊆ Qexp(XI-ary)(A) follows from Corollary 3.2(i). Let us check the inclusion Qexp(XI)(A) ⊇ Qexp(XI−ary)(A). If A = ∅, the claim is trivial, since Qexp Y (∅) = {∅}, for every ballean Y . Otherwise, fix an element x ∈ A. Let C ∈ exp BI-ary(A,K), for some K ∈ I. Then C 6= ∅, so we can fix also a point y ∈ C and let K′ = K ∪{x,y}∈I. Then C ⊆ BI-ary(A,K) = A∪K = BI(A,K′) and A ⊆ BI-ary(C,K) = C ∪K = BI(C,K′), which shows that C ∈ exp BI(A,K′). Hence, C ∈Qexp(XI)(A). This proves the equality Qexp(XI)(A) = Qexp(XI−ary)(A). It implies the first as well as the second equality in (4.2). To prove the last equality in (4.2), it suffices to note that QC(X,I)(0) = I, by virtue of (ii). Hence, dsc(C(X,I)) = dsc(C(X,I) \{0}). To conclude, use the fact that exp∗(XI-ary) is asymorphic to C(X,I) \{0}. Item (iv) follows from (iii) and (4.1) applied to B = exp(XI). � Proposition 4.1 allows us to reduce the computations of the number of con- nected components of all hyperballeans involved to the computation of the car- dinal dsc(C(X,I). In the sequel we simply identify CX with the Boolean ring c© AGT, UPV, 2019 Appl. Gen. Topol. 20, no. 2 444 Balleans, hyperballeans and ideals P(X). So that functions f ∈ CX are identifies with their support and the ideals I of X are simply the proper ideals of the Boolean ring CX = {0, 1}X = ZX2 , containing ⊕ X Z2. According to Proposition 4.1, the connected components of C(X,I) are pre- cisely the cosets f +I of the ideal I, therefore, dsc(C(X,I)) coincides with the cardinality of the quotient ring CX/I: (4.4) dsc(C(X,I)) = |CX/I| = |P(X)/I|. In particular, for every infinite set X and an ideal I of X one has dsc(C(X,I)) = 2 if and only if I is a maximal ideal. This is an obvious consequence of (4.4) as |CX/I| = 2 if and only if the ideal I is maximal. Remark 4.2. The cardinality |CX/I| is easy to compute in some cases, or to get at least an easily obtained estimate for |CX/I| from above as we see now. To this end let ι(I) = min{λ | I is an intersection of λ maximal ideals}. Then (4.5) |CX/I|≤ min{2ι(I), 2|X|}. Indeed, if I = ⋂ {mi | i < ι(I)}, where mi are maximal ideals of B, then B/I embeds in the product ∏ i<ι(I) CX/mi having size ≤ 2 ι(I) as CX/mi ∼= Z2 for all i. To conclude the proof of (4.5) it remains to note that obviously |CX/I|≤ |CX| = 2|X|. If ι(I) = n is finite, then dsc(C(X,I)) = 2n. Indeed, now I = m1 ∩·· ·∩mn is a finite intersection of maximal ideals and the Chinese Remainder Theorem, applied to the Boolean ring CX and the maximal ideals m1, . . . ,mn, provides a ring isomorphism CX/I ∼= n∏ i=1 C/mi ∼= Zn2 . In particular, |CX/I| = |Zn2 |= 2n. By (4.4), we deduce (4.6) dsc(C(X,I)) = 2n. Let us conclude now with another example. For every infinite set X and the ideal I = FX one has dsc(exp(XFX )) = dsc(C(X,FX) = 2 |X|. This follows from (4.4) and |FX| = |X| < 2|X|, which implies |CX/FX| = |CX| = 2|X|. In order to obtain some estimate from below for |CX/I|, we need a deeper insight on the spectrum Spec CX of CX. Since CX is a Boolean ring, Spec CX coincides with the space of all maximal ideals of CX, which can be identified with the Stone–Čech compactification βX when we endow X with the discrete topology. As usual, c© AGT, UPV, 2019 Appl. Gen. Topol. 20, no. 2 445 D. Dikranjan, I. Protasov, K. Protasova and N. Zava • we identify the Stone–Čech compactification βX with the set of all ultrafilters on X; • the family {A | A ⊆ X}, where A = {p ∈ βX | A ∈ p}, forms the base for the topology of βX; and • the set X is embedded in βX by sending x ∈ X to the principal ultra- filter generated by x. For a filter ϕ on X, define a closed subset ϕ of βX as follows: ϕ = ⋂ {A | A ∈ ϕ}. An ultrafilter p ∈ βX belongs to ϕ if and only if p contains the filter ϕ. In other words, (4.7) ϕ = ⋂ {u | u ∈ ϕ}. For an ideal I on X, we consider the filter ϕI = {X\A | A ∈ I}, and we simply write ϕ when there is no danger of confusion. Similarly, for a filter ϕ we define the ideal Iϕ = {X\A | A ∈ ϕ} and we simply write I when there is no danger of confusion. Finally, let I∧ = ϕI, and note that all ultrafilters in I∧ are non-fixed, i.e., I∧ ⊆ βX \X, as ϕ is contained in the Fréchet filter of all co-finite sets on X (since I ⊇ FX). Moreover, for a subset A of X one has (4.8) A ∈I if and only if A 6∈ u for all u ∈ ϕI = I∧. As pointed out above, for any X the compact space βX coincides with the spectrum Spec CX of the ring CX. For an ideal I on X, ϕI is the set of ultrafilters on X containing ϕ. For u ∈ ϕI the ideal Iu is maximal and contains I. More precisely, I = ⋂ u∈I∧ Iu. The maximal ideals Iu, when u runs over ϕ, bijectively correspond to the maximal ideals of the quotient CX/I; in particular, |I∧| = |Spec(CX/I)|. Along with Remark 4.2, this gives: Proposition 4.3. Let I be an ideal on set X. If |I∧| = n is finite, then dsc(C(X,I)) = 2n. Otherwise, dsc(C(X,I)) = w(I∧). Here w(I∧) denotes the weight of the space I∧. The second assertion follows from (4.4) and the equality w(I∧) = |P(X)/I|, its proof can be found in [3, §2]. Corollary 4.4. Let I be an ideal on a countably infinite set set X such that I∧ is infinite. Then dsc(C(X,I)) = 2ω. Proof. Being an infinite compact subset of βX \X, I∧ contains a copy of βN. Therefore, w(I∧) = 2ω. Now Proposition 4.3 applies. � In this section we have thoroughly investigated the number of connected components of exp(XI), where I is an ideal of a set X. This leaves open the question to estimate dsc(exp(X)), where X is an arbitrary connected ballean. c© AGT, UPV, 2019 Appl. Gen. Topol. 20, no. 2 446 Balleans, hyperballeans and ideals Let Y be a subballean of X. In particular, dsc(exp(X)) ≥ dsc(exp(Y )). If Y is thin, we can apply Theorem 2.2 so that Y = Y[(Y ). The results from this section give a lower bound of dsc(exp(Y )), providing in this way also a lower bound for dsc(exp(X)), since (4.9) dsc(exp(X)) ≥ sup{dsc(exp(Z)) | Z is a thin subballean of X}. Unfortunately, (4.9) doesn’t provide any useful information in the case when every thin subballean of X is bounded. In fact, if Z is a non-empty bounded subballean, then exp∗(Z) is connected and so dsc(exp(Z)) = 2. Acknowledgements. The first named author thankfully acknowledges par- tial financial support via the grant PRID at the Department of Mathematical, Computer and Physical Sciences, Udine University. References [1] T. Banakh, I. Protasov, D. Repovs̆ and S. Slobodianiuk, Classifying homogeneous cel- lular ordinal balleans up to coarse equivalence, arxiv: 1409.3910v2. [2] T. Banakh and I. Zarichnyi, Characterizing the Cantor bi-cube in asymptotic categories, Groups, Geometry and Dynamics 5 (2011), 691–728. [3] W. Comfort and S. Negrepontis, The Theory of Ultrafilters, Grundlehren der math- ematischen Wissenschaften, Band 211, Springer–Verlag, Berlin-Heidelberg-New York, 1974. [4] D. Dikranjan and N. Zava, Some categorical aspects of coarse structures and balleans, Topology Appl. 225 (2017), 164–194. [5] D. Dikranjan and N. Zava, Preservation and reflection of size properties of balleans, Topology Appl. 221 (2017), 570–595. [6] A. Dow, Closures of discrete sets in compact spaces, Studia Sci. Math. Hungar. 42, no. 2 (2005), 227–234. [7] K. Kunen, Set theory. An introduction to independence proofs, Studies in Logic and Foundations of Math., vol. 102, North-Holland, Amsterdam-New York-Oxford, 1980. [8] O. Petrenko and I. Protasov, Balleans and filters, Mat. Stud. 38, no. 1 (2012), 3–11. [9] I. Protasov and T. Banakh, Ball Structures and Colorings of Groups and Graphs, Mat. Stud. Monogr. Ser 11, VNTL, Lviv, 2003. [10] I. Protasov and K. Protasova, On hyperballeans of bounded geometry, arXiv:1702.07941v1. [11] I. Protasov and M. Zarichnyi, General Asymptology, 2007 VNTL Publishers, Lviv, Ukraine. [12] J. Roe, Lectures on Coarse Geometry, Univ. Lecture Ser., vol. 31, American Mathemat- ical Society, Providence RI, 2003. [13] N. Zava, On F-hyperballeans, work in progress. c© AGT, UPV, 2019 Appl. Gen. Topol. 20, no. 2 447