@ Appl. Gen. Topol. 18, no. 1 (2017), 31-44 doi:10.4995/agt.2017.4469 c© AGT, UPV, 2017 Non metrizable topologies on Z with countable dual group Daniel de la Barrera Mayoral Department of Mathematics, IES Los Rosales, Madrid, Spain (dbarrera@ucm.es) Communicated by E. Induráin Abstract In this paper we give two families of non-metrizable topologies on the group of the integers having a countable dual group which is isomorphic to a infinite torsion subgroup of the unit circle in the complex plane. Both families are related to D-sequences, which are sequences of natural numbers such that each term divides the following. The first family consists of locally quasi-convex group topologies. The second consists of complete topologies which are not locally quasi-convex. In order to study the dual groups for both families we need to make numerical considerations of independent interest. 2010 MSC: Primary: 22A05; 55M05; Secondary: 20K45. Keywords: locally quasi-convex topology; D-sequence; continuous charac- ter; infinite torsion subgroups of T. 1. Introduction and notation The most important class of topological abelian groups are the locally com- pact groups. In this class, the most interesting theorems can be proved. In this paper we consider a wider class of groups, which is the class of locally quasi- convex groups. Clearly, locally compact groups are locally quasi-convex, since they can be considered as a dual group (in fact, a locally compact abelian group is isomorphic to its bidual group) and dual groups are locally quasi-convex. This paper is part of a project about duality, which intends to solve, among others, the Mackey Problem (first stated in [8]). This problem asks whether Received 28 December 2015 – Accepted 13 September 2016 http://dx.doi.org/10.4995/agt.2017.4469 D. de la Barrera Mayoral there exists a maximum element in the family of all locally quasi-convex com- patible topologies (see Definition 1.12). We present two different kinds of topologies on the group of the integers, having as common feature that topologies in both families have countable dual group, which is isomorphic to an infinite torsion subgroup of the unit circle in the complex plane. On the one hand, the topologies presented in Section 3 are metrizable, non- complete and locally quasi-convex and the results contained are the natural further steps of the ones given in [2]. On the other hand, the topologies presented in Section 4 are complete, non- metrizable but they are not locally quasi-convex. These topologies, introduced by Graev, were thoroughly studied by Protasov and Zelenyuk. Precisely, in [16], a neighborhood basis for this topology is given and this allows us to study these topologies on Z from a numerical point of view. The fact that these topologies have countable dual group is surprising due to the belief that the more open sets the topology has, the more continuous characters it should have. All the groups considered will be abelian and the notation will be additive. In order to state properly the Mackey Problem, we need first to recall some notation on duality: Definition 1.1. We will consider the unit circle as T = R/Z = { x + Z : x ∈ ( − 1 2 , 1 2 ]} . We will consider in T the following neighborhood basis: Tm := { x + Z : x ∈ [ − 1 4m , 1 4m ]} . We set T+ := { x + Z : x ∈ [ − 1 4 , 1 4 ]} = T1. Definition 1.2. Let (G,τ) be a topological group. Denote by G∧ the set of continuous homomorphisms (or continuous characters) CHom(G,T). This set is a group when we consider the operation (f +g)(x) = f(x)+g(x). In addition, if we consider the compact open topology on G∧ it is a topological group. The topological group G∧ is called the dual group of G. Definition 1.3. The natural embedding αG : G → G∧∧, is defined by x 7→ αG(x) : G∧ → T, where χ 7→ χ(x). Proposition 1.4. αG is a homomorphism. c© AGT, UPV, 2017 Appl. Gen. Topol. 18, no. 1 32 Non metrizable topologies on Z with countable dual group Proof. αG(x+y)(χ) = χ(x+y) = χ(x)χ(y) = αG(x)(χ)αG(y)(χ) = (αG(x)αG(y))(χ), for all χ ∈ G∧. Thus, αG(x + y) = αG(x)αG(y). � Definition 1.5. A topological group (G,τ) is reflexive if αG is a topological isomorphism. Definition 1.6. Let τ,ν be two group topologies on a group G. We say that τ and ν are compatible if algebraically (G,τ)∧ = (G,ν)∧. Now we include the definition of quasi-convex set, which emulates the Hahn- Banach Theorem from Functional Analysis: Definition 1.7 ([17]). Let (G,τ) be a topological group and let M ⊂ G be a subset. We say that M is quasi-convex if for every x ∈ G\M, there exists χ ∈ G∧ such that χ(M) ⊂ T+ and χ(x) /∈ T+. We say that τ is locally quasi- convex if there exists a neighborhood basis for τ consisting of quasi-convex subsets. We denote by Lqc the class of all locally quasi-convex topological groups. Definition 1.8. Let τ be a group topology. We define the locally quasi-convex modification of τ as the finest locally quasi-convex group topology (τ)lqc among all those satisfying that they are coarser than τ. For a deeper insight into the locally quasi-convex modification of a topolog- ical group, see [10]. Lemma 1.9. Let (G,τ) be a topological group. Then (G,τ)∧ = (G, (τ)lqc) ∧. Definition 1.10. Let (G,τ) a topological group. The polar of a subset S ⊂ G is defined as S. := {χ ∈ G∧ | χ(S) ⊂ T+}. Remark 1.11. It is clear that (G,τ)∧ = ⋃ V∈B V ., where B is a neighborhood basis of 0 for the topology τ. Definition 1.12 ([12]). Let G be a class of abelian topological groups and let (G,τ) be a topological group in G. Let CG(G,τ) denote the family of all G-topologies ν on G compatible with τ. We say that µ ∈ CG(G,τ) is the G- Mackey topology for (G,G∧) (or the Mackey topology for (G,τ) in G) if ν ≤ µ for all ν ∈CG(G,τ). If G is the class of locally quasi-convex groups Lqc, we will simply say that the topology is Mackey. The problem whether the Mackey Topology for a topological group (G,τ) exists was first posed in [8]. This problem is called the Mackey Problem. During recent years several mathematicians have tried to solve the Mackey Problem: de Leo [9], Aussenhofer, Dikranjan, Mart́ın-Peinador [4], Dı́az-Nieto c© AGT, UPV, 2017 Appl. Gen. Topol. 18, no. 1 33 D. de la Barrera Mayoral [10], Gabriyelyan [11]. A survey including similarities and differences between the Mackey topology in spaces and the Mackey Problem for groups has recently appeared [14]. In [3], the Mackey Problem for groups of finite exponent is reduced to the problem of finding a top element in the family of compatible linear topologies (that is, topologies having a neighborhood basis consisting in open subgroups). In order to solve the Mackey Problem (in the negative) we try to find all the locally quasi-convex group topologies which are compatible with a non-discrete linear topology on Z and study if the supremum of all these topologies is again compatible. So far, it is not known if the supremum of compatible topologies is again compatible. Now we give the basic definitions on D-sequences, which are sequences of natural numbers that encode the relevant information of non-discrete linear topologies on Z. Definition 1.13. A sequence of natural numbers b = (bn)n∈N0 ∈ NN0 is called a D-sequence if it satisfies: (1) b0 = 1, (2) bn 6= bn+1 for all n ∈ N0, (3) bn divides bn+1 for all n ∈ N0. The following notation will be used in the sequel: • D := {b : b is a D − sequence}. • D∞ := {b ∈D : bn+1 bn →∞}. • D∞(b) := {c : c is a subsequence of b and c ∈D∞}. and for an arbitrary (fixed) natural number ` we define: • D`∞ := {b ∈D : bn+` bn →∞}. • D`∞(b) := {c : c is a subsequence of b and c ∈D`∞}. Our interest in D-sequences stems from the fact that several compatible topologies on Z can be associated to D-sequences. This family of topologies is interesting due to the fact that we don’t know in general, if the supremum of compatible topologies is again compatible. This is the main difference be- tween the study of the Mackey Topology in spaces and the Mackey Problem for groups. Definition 1.14. Let b be a D-sequence. Define (qn)n∈N by qn := bn bn−1 . Definition 1.15. Let b be a D-sequence. We say that (a) b has bounded ratios if there exists a natural number N, satisfying that qn = bn bn−1 ≤ N. (b) b is basic if qn is a prime number for all n ∈ N. Definition 1.16. Let b be a D-sequence. We write Z(bn) := { k bn + Z : k = 0, 1, . . . ,bn − 1 } and Z(b∞) := ⋃ n∈N0 Z(bn). c© AGT, UPV, 2017 Appl. Gen. Topol. 18, no. 1 34 Non metrizable topologies on Z with countable dual group As quotient groups, we can write Z(bn) = 〈{ 1 bn }〉 /Z and Z(b∞) = 〈{ 1 bm : m ∈ N0 }〉 /Z. In other words, Z(bn) consists of the elements whose order divides bn and Z(b∞) is generated by the elements having order bn for some natural number n ∈ N0. Proposition 1.17. Let b be a D-sequence. Suppose that qj+1 6= 2 for infinitely many j. For each integer number L ∈ Z, there exists a natural number N = N(L) and unique integers k0, . . . ,kN , such that: (1) L = N∑ j=0 kjbj. (2) ∣∣∣∣∣∣ n∑ j=0 kjbj ∣∣∣∣∣∣ ≤ bn+12 for all n. (3) kj ∈ ( − qj+1 2 , qj+1 2 ] , for 0 ≤ j ≤ N. The proof of Proposition 1.17 can be found in [5, Proposition 2.2.1] and [6, Proposition 1.4] Definition 1.18. The family Bb := {bnZ : n ∈ N0} is a neighborhood basis of 0 for a linear (that is, it has a neighborhood basis consisting in open subgroups) group topology on Z, which will be called b-adic topology and is denoted by λb. This topology is precompact. In [2], it is proved that (Z,λb)∧ = Z(b∞) and the following Lemma 1.19. Let b be a D-sequence. Then bn → 0 in λb. The topologies of uniform convergence on the group Z are given by differ- ent families in Hom(Z,T). Identifying Hom(Z,T) with T, we simply have to consider families of subsets in T. In this paper, we concentrate on families of cardinality 1, that is we consider the topology of uniform convergence on {B} for B ⊂ T. We shall write τB instead of τ{B}. The topology τB admits the following neighborhood basis of 0: VB,m := {z ∈ Z : χ(z) + Z ∈ Tm for all χ ∈ B} = ⋂ χ∈B χ−1(Tm). A particular case of these topologies is the following. Definition 1.20. A D-sequence b in Z induces in a natural way a sequence in T. Namely, for a D-sequence, b = (bn)n∈N0 , define b := ( 1 bn + Z : n ∈ N0 ) ⊂ T. c© AGT, UPV, 2017 Appl. Gen. Topol. 18, no. 1 35 D. de la Barrera Mayoral Denote by τb the topology of uniform convergence on b. 2. Killing λb null sequences. In this section we prove that for a D sequence, b with bounded ratios and any λb convergent sequence, (an), there exists a strictly finer compatible topol- ogy (constructed as a topology of uniform convergence on a suitable sequence in D∞(b)) such that the fixed sequence (an) is no longer convergent. This will allow us to find a locally quasi-convex topology which has no convergent sequences, but it is still compatible. Theorem 2.1. Let b be a basic D-sequence with bounded ratios and let (xn) ⊂ Z be a non-quasiconstant (that is, no element in (xn) is repeated infinitely many times) sequence such that xn λb→ 0. Then there exists a metrizable locally quasi-convex compatible group topology τ(= τc for some subsequence c of b) on Z satisfying: (a) τ is compatible with λb. (b) xn τ9 0. (c) λb < τ. Proof. We define τ as the topology of uniform convergence on a subsequence c of b, which we construct by means of the claim: Claim 2.2. Since b is a D-sequence with bounded ratios, there exists L ∈ N such that qn+1 ≤ L for all n ∈ N. We can construct inductively two sequences (nj), (mj) ⊂ N such that nj+1 −nj ≥ j and xmj bnj+1 + Z /∈ TL. Proof of the claim: j = 1. Choose n1 ∈ N such that there exists xm1 , satisfying bn1 | xm1 but bn1+1 - xm1 . By Proposition 1.17, we write xm1 = N(xm1 )∑ i=0 kibi. The condition bn1 | xm1 implies ki = 0 if i < n1 and bn1+1 - xm1 implies kn1 6= 0. Hence, xm1 bn1+1 + Z = N(xm1 )∑ i=0 kibi bn1+1 + Z = n1∑ i=0 kibi bn1+1 + Z = kn1 qn1+1 + Z. We have that 1 2 ≥ |kn1| qn1+1 ≥ 1 L . c© AGT, UPV, 2017 Appl. Gen. Topol. 18, no. 1 36 Non metrizable topologies on Z with countable dual group Therefore, xm1 bn1+1 + Z /∈ TL = [ − 1 4L , 1 4L ] + Z. j ⇒ j + 1. Suppose we have nj,mj satisfying the desired conditions. Let nj+1 ≥ nj + j be a natural number such that there exists xmj+1 satisfying bnj+1 | xmj+1 and bnj+1+1 - xmj+1 . By Proposition 1.17, we write xmj+1 = N(xmj+1 )∑ i=0 kibi. The condition bnj+1 | xmj+1 implies that ki = 0 if i < nj+1 and bnj+1+1 - xmj+1 implies knj+1 6= 0. Then, xmj+1 bnj+1+1 + Z = N(xmj+1 )∑ i=0 kibi bnj+1+1 + Z = nj+1∑ i=0 kibi bnj+1+1 + Z = knj+1 qnj+1+1 + Z. From 1 2 ≥ ∣∣knj+1∣∣ qnj+1+1 ≥ 1 qnj+1+1 ≥ 1 L , we get that xmj+1 bnj+1+1 + Z /∈ TL. This ends the proof of the claim. We continue the proof of Theorem 2.1. Consider now c = ( bnj+1 ) j∈N. Let c = { 1 bnj+1 + Z : j ∈ N } and let τ = τc be the topology of uniform convergence on c. Then: (1) By Proposition [2, Remark 3.3], τ is metrizable and locally quasi- convex. By [2, Proposition 3.7], we have λb < τ. (2) By the claim, we have proved that xmj j→∞ 9 0 in τ. This implies that xn 9 0 in τ. (3) Since nj+1 ≥ nj +j, we get that nj+1−nj ≥ j. Thus, bnj+1 bnj+1+1 ≤ 1 2j → 0. By [2, Theorem 4.4], we have that (Z,τ)∧ = Z(b∞). � c© AGT, UPV, 2017 Appl. Gen. Topol. 18, no. 1 37 D. de la Barrera Mayoral 3. The topology Γb. In Section 2, we use a subfamily of D∞(b) to eliminate all λb convergent sequences. Unfortunately, this subfamily depends on the choices made in Theo- rem 2.1. This is why we introduce a new topology Γb, which is the supremmum of the topologies of uniform convergence on sequences in D∞(b). The topology Γb is well-defined (in the sense that it does not depend on the choices made in Theorem 2.1). Definition 3.1. Let b be a basic D-sequence with bounded ratios. We define on Z the topology Γb := sup{τc : c ∈D∞(b)}. Now we set some results on Γb: Remark 3.2. Since D∞(b) is non-empty, we have that λb < Γb. Theorem 3.3. Let b a basic D-sequence with bounded ratios. Then the topol- ogy Γb has no nontrivial convergent sequences. Hence, it is not metrizable. Proof. Since every Γb-convergent sequence is λb-convergent, we consider only λb-convergent sequences. Let (xn) ⊆ Z be a sequence such that xn λb→ 0. By Theorem 2.1, there exists a subsequence (bnk)k satisfying that τ(bnk ) is a locally quasi-convex topology, (bnk) ∈D∞(b) and xn τbnk9 0. Since τbnk ≤ Γb, we know that xn Γb9 0. Hence, the only convergent sequences in Γb are the trivial ones. Since Γb has no nontrivial convergent sequences, it is not metrizable. � Corollary 3.4. Let p be a prime number and set p = (pn). Then Γp has no nontrivial convergent sequences. Lemma 3.5. Let b be a D-sequence. Then (Z, Γb)∧ = Z(b∞). For the proof of Lemma 3.5 we need the following definition of [5]: Definition 3.6 ([5, Definition 4.4.4]). Let b be a basic D-sequence with bounded ratios. We define on Z the topology δb := sup{τc : c ∈D`∞(b)}. Proof of Lemma 3.5. Since λb < Γb, it is clear that Z(b∞) ≤ (Z, Γb)∧. By definition, it is clear that Γb ≤ δb. Hence (Z, Γb)∧ ≤ (Z,δb)∧ = Z(b∞). 2 Now we set some interesting questions related to the Mackey Problem and Γb: Open question 1. Is Γb = δb? Open question 2. Is Γb (or δb, if different) the Mackey topology for (Z,λb)? The following lemmas are easy to prove, even for a topological space instead of a topological group. c© AGT, UPV, 2017 Appl. Gen. Topol. 18, no. 1 38 Non metrizable topologies on Z with countable dual group Lemma 3.7. Let K be a countably infinite compact subset of a Hausdorff space. Then K is first countable, and, hence, every accumulation point in K is the limit point of some sequence. Proof. Let K = {xn,n ∈ N}. Fix a ∈ K. Consider for each xn ∈ K a closed neighborhood of a (this is possible because K is regular) say Vn ⊂ K\xn. The set {Vn,n ∈ N} and its finite intersections constitute a neighborhood basis for a ∈ K. In fact, if W is any open neighborhood of a, we have a family of closed sets given by {K \W,Vn,n ∈ N} with empty intersection. Since K is compact, the mentioned family cannot have the finite intersection property. Thus, there is a finite subfamily {Vj,j ∈ F} such that (K \W) ∩i∈F Vi = ∅. � Lemma 3.8. If a countable topological group G has no nontrivial convergent sequences, then it does not have infinite compact subsets either. Proof. Suppose, by contradiction, that K ⊂ G is an infinite compact subset. Since K is first countable and Hausdorff, any accumulation point is the limit of a sequence in K. SinceK is infinite and compact, it has an accumulation point, contradicting the fact that there does not exist convergent sequences in G (nor in K). Hence any compact subset must be finite. � Proposition 3.9. Let b be a basic D-sequence with bounded ratios, then the group G = (Z, Γb) is not reflexive. In fact, the bidual G∧∧ can be identified with Z endowed with the discrete topology. Proof. The dual group of (Z, Γb) has supporting set Z(b∞). The Γb-compact subsets of Z are finite by Lemma 3.8, therefore, the dual group carries the pointwise convergence topology. Thus, G∧ is exactly Z(b∞) with the topology induced by the euclidean of T. By [1, 4.5], the group Z(b∞) has the same dual group as T, namely, Z with the discrete topology. We conclude that the canonical mapping αG is an open non-continuous isomorphism. � Corollary 3.10. Since (Z, Γb)∧∧ is discrete, the homomorphism α(Z,Γb) is not continuous. However, it is an open algebraic isomorphism. Proposition 3.11. Let b ∈ D`∞ and Gγ := (Z,τb). Then αGγ is a non- surjective embedding from Gγ into G ∧∧ γ . Proof. Since τb is metrizable, αGγ is continuous. The fact that τb is locally quasi-convex and Z(b∞) separates points of Z imply that αGγ is injective and open in its image [1, 6.10]. On the other hand, αGγ is not onto, for otherwise Gγ would be reflexive. However, a non-discrete countable metrizable group cannot be reflexive. Indeed, the dual of a metrizable group is a k-space ([1, 7]) and the dual group of a k-space is complete. Hence, the original group must be complete as well. By Baire Category Theorem, the only metrizable complete group topology on a countable group is the discrete one. � c© AGT, UPV, 2017 Appl. Gen. Topol. 18, no. 1 39 D. de la Barrera Mayoral 4. The complete topology T{bn} In this section we consider some special families of topologies which were introduced by Graev and deeply studied in the group of the integers by Protasov and Zelenyuk ([15, 16]). We need first some definitions about sequences. Definition 4.1. Let G be a group and g = (gn) ⊂ G a sequence of elements of G. We say that g is a T-sequence if there exists a Hausdorff group topology τ such that gn τ→ 0. Since the topology must be Hausdorff, we can consider that gn 6= gm if n 6= m Lemma 4.2. Let g ⊂ G be a T -sequence, then there exists the finest group topology T{gn} satisfying that gn T{gn}→ 0. Since we can describe T{gn} we are interested in finding a suitable neighbor- hood basis for this topology. Definition 4.3 ([16]). Let G be a group and let a = (an) be a T-sequence in G and (ni)i∈N a sequence of natural numbers. We define: • A∗m := {±an|n ≥ m}∪{0G}. • A(k,m) := {g0 + · · · + gk|gi ∈ A∗m i ∈{0, . . . ,k}}. • [n1, . . . ,nk] := {g1 + · · · + gk : gi ∈ A∗ni, i = 1, . . . ,k}. • V(ni) = ∞⋃ k=1 [n1, . . . ,nk]. Proposition 4.4. The family {V(ni) : (ni) ∈ N N} is a neighborhood basis of 0G for a group topology T{an} on G, which is the finest among all those satisfying an → 0. The symbol G{an} will stand for the group G endowed with T{an}. Next, we include a theorem of great interest from [16]: Theorem 4.5 ([16, Theorem 2.3.11]). Let (gn) be a T -sequence on a group G. Then, the topology T{gn} is complete. Corollary 4.6. Let b be a D-sequence and c ∈D`∞(b). Then τc 6= T{bn}. Applying the Baire Category Theorem, we can obtain the following corollary: Corollary 4.7. Let G be a countable group. For any T -sequence g, the topology T{gn} is not metrizable. Since we are mainly interested on topologies on the group of integers, we include the following results: Proposition 4.8 ([16, Theorem 2.2.1]). If limn→∞ an+1 an = ∞ then (an) is a T -sequence in Z. This proposition provides a condition to prove that the following sequences are T-sequences: (bn) = (2 n2 ), (ρn) = (p1 · · · · ·pn), where (pn) is the sequence of prime numbers. c© AGT, UPV, 2017 Appl. Gen. Topol. 18, no. 1 40 Non metrizable topologies on Z with countable dual group Proposition 4.9 ([16, Theorem 2.2.3]). If limn→∞ an+1 an = r and r is a tran- scendental number then (an) is a T -sequence in Z. Let b be a D-sequence. It is clear (by Lemma 1.19) that b is a T-sequence. Hence, the topology T{bn} can be constructed on Z. By Theorem 4.5, it is a complete topology. The following properties of Z{an} are interesting by themselves. Remark 4.10. Let (an) be a T-sequence on Z. Then: (1) Z{an} is sequential, but not Frechet-Urysohn ([15, Theorem 7 and The- orem 6]). (2) Z{an} is a kω-group ([16, Corollary 4.1.5] ). We prove now that, for a D-sequence b with bounded ratios, the locally quasi-convex modification of T{bn} is λb. This fact is also proved in [11, Propo- sition 2.5], but we give here a direct proof. First we study some numerical results in the group T: Lemma 4.11. Let k ∈ T and m ∈ N satisfy k, 2k,. . . ,mk ∈ T+. If mk ∈ Tn, then k ∈ Tnm. Proof. Write k = z + y where z ∈ Z and y ∈ ( −1 2 , 1 2 ] . Since k ∈ T+, we have |y| ≤ 1 4 . Suppose that k /∈ Tnm. This means that 14nm < |y| ≤ 1 4 . Now, mk = mz + my. Since 1 4n < |my| ≤ m 4 and mk ∈ Tn, we have that |my| ≥ 1 − 14n. Hence, there exists N ≤ k such that 1 4 < |Ny| < 3 4 . Consequently, Nk /∈ T+. � Lemma 4.12. Let k ∈ T and let m1,m2, . . . ,mn ∈ N\{1}. If k, 2k, 3k,. . . ,m1k, 2m1k,. . . ,m1m2k,. . . ,( n−1∏ i=1 mi ) k, 2 ( n−1∏ i=1 mi ) k,. . . , ( n∏ i=1 mi ) k ∈ T+, then k ∈ T∏n i=1 mi . Proof. Since (∏n−1 i=1 mi ) k, 2 (∏n−1 i=1 mi ) k,. . . ( ∏n i=1 mi) k ∈ T+ and ( ∏n i=1 mi) k ∈ T1, by Lemma 4.11, we have that (∏n−1 i=1 mi ) k ∈ Tmn. Since (∏n−2 i=1 mi ) k, 2 (∏n−2 i=1 mi ) k,. . . (∏n−1 i=1 mi ) k ∈ T+ and (∏n−1 i=1 mi ) k ∈ Tmn, by Lemma 4.11, we have that (∏n−2 i=1 mi ) k ∈ Tmnmn−1 . Repeating the argument n times, we obtain that k ∈ T∏n i=1 mi. � Lemma 4.13. Let k ∈ T and let (mi) ∈ NN a sequence of natural numbers satisfying mi > 1 for all i ∈ N. If k, 2k, 3k,. . . ,m1k, 2m1k,. . . ,m1m2k,. . . ,( n−1∏ i=1 mi ) k, 2 ( n−1∏ i=1 mi ) k,. . . , ( n∏ i=1 mi ) k, · · · ∈ T+ c© AGT, UPV, 2017 Appl. Gen. Topol. 18, no. 1 41 D. de la Barrera Mayoral then k = 0 + Z. Proof. By Lemma 4.12, we have that k ∈ ⋂ N>0 T∏Ni=1 mi. Hence, k = 0+Z. � Now we can study the polar of the neighborhoods in the topology T{bn}. Proposition 4.14. Let b be a D-sequence with bounded ratios and V(ni) a basic neighborhood of T{bn}. Then, there exists N ∈ N such that A := {bN, 2bN, . . . ,bN+1, 2bN+1, . . . ,bN+2, 2bN+2, . . .}∈ V(ni). As a consequence V . (ni) are finite subsets of T, for any sequence (ni). Proof. Since b has bounded ratios, there exists L ∈ N such that qn ≤ L for all n ∈ N. Let n1,n2, . . . ,nL be the indexes satisfying that ni ≤ nj if i ∈ {1, 2, . . . ,L} and j /∈ {1, 2, . . . ,L}. That is, the terms n1, . . . ,nL are the L smallest ones in (ni). Let N := max{n1,n2, . . . ,nL}. Then bN,bN+1, · · · ∈ A∗ni for i ∈ {1, 2, . . . ,L} and A ⊂{bN, 2bN, . . . ,LbN,bN+1, 2bN+1, . . . ,LbN+1, . . .}⊂ [n1,n2, . . . ,nL] ⊂ V(ni). Let χ ∈ V . (ni) . Since A ⊂ V(ni), it is clear that V . (ni) ⊂ A.. Define k := χ(bN ). Then χ(A) = {k, 2k,. . .qNk, 2qNk,. . . ,qNqN+1k,. . .}. Since χ ∈ A., we have χ(A) ⊂ T+. By Lemma 4.13, the fact k = χ(bN ) = 0 + Z follows. Hence χ ∈ Z(bN ) and V .(ni) is finite. � Corollary 4.15. Let b be a D-sequence with bounded ratios. Then Z∧{bn} = Z(b∞). Proof. Since Z{bn} = ⋃ (ni) V . (ni) , we have that Z∧{bn} ≤ Z(b ∞). Since b → 0 in λb, λb ≤T{bn} and Z ∧ {bn} ≥ Z(b ∞). � Finally, we can prove that for a D-sequence, b, with bounded ratios, the locally quasi-convex modification of T{bn} is λb. Theorem 4.16. Let b a D-sequence with bounded ratios. Then the locally quasi-convex modification of T{bn} is λb. Proof. Since the polar sets of the basic neighborhoods of T{bn} are finite, the equicontinuous subsets in the dual group are finite as well. Hence the lo- cally quasi-convex modification of T{bn} is precompact. Since T{bn} and its locally quasi-convex modification are compatible topologies, it follows that( T{bn} ) lqc = λb. � Corollary 4.17. Let b a D-sequence with bounded ratios. Then T{bn} is not locally quasi-convex. Proof. If T{bn} were locally quasi-convex, then ( T{bn} ) lqc = T{bn}, but ( T{bn} ) lqc is not complete and T{bn} is complete. � c© AGT, UPV, 2017 Appl. Gen. Topol. 18, no. 1 42 Non metrizable topologies on Z with countable dual group Remark 4.18. Although the examples in Theorem 4.16 are not locally quasi- convex, there exist examples of Graev-type topologies on Z which are locally quasi-convex. Indeed, in [13] it is proved that there exist Graev-type topologies on Z which are reflexive (hence locally quasi-convex). Remark 4.19. Corollary 4.17 gives a family of examples of: • Complete topologies whose locally quasi-convex modifications are not complete. • Non-metrizable topologies whose locally quasi-convex modifications are metrizable. Proposition 4.20. The dual group of Z{bn} coincides algebraically and topo- logically with the dual group of (Z,λb); that is, the dual group of Z{bn} is Z(b ∞) endowed with the discrete topology. Proof. From Corollary 4.15 the dual group of Z{bn} is Z ∧ {bn} = Z(b ∞). Item (2) in Remark 4.10 implies that Z∧{bn} is metrizable and complete. By Baire Category Theorem and since Z∧{bn} is countable, we get that it is discrete. Hence, it coincides topologically with (Z,λb)∧. � Now we answer the question whether Z{bn} is MAP-Mackey. Proposition 4.21. The topology Z{bn} is not MAP-Mackey. Proof. If T{bn} were Mackey, the condition Γb ≤ T{bn} should hold. But, the topology Γb has no convergent sequences and bn → 0 in T{bn}. � Open question 3. Let (gn) a T -sequence in Z. If T{gn} is locally quasi-convex, is T{gn} the Mackey Topology? Acknowledgements. The author is thankful to the referee for the kind and deep review. References [1] L. Außenhofer, Contributions to the duality theory of abelian topological groups and to the theory of nuclear groups, Dissertationes Math. (Rozprawy Mat.) 384, (1999). [2] L. Außenhofer and D. de la Barrera Mayoral, Linear topologies on Z are not Mackey topologies, Journal of Pure and Applied Algebra 216, no. 6 (2012), 1340–1347. [3] L. Außenhofer, D. de la Barrera Mayoral, D. Dikranjan and E. Mart́ın-Peinador, ’Varopoulos paradigm’: Mackey property versus metrizability in topological groups, Revista Matemática Complutense 30, no. 1 (2017), 167-184. [4] L. Außenhofer, D. Dikranjan and E. Mart́ın-Peinador, Locally quasi-convex compatible topologies on a topological group, Axioms 4 (2015), 436–458. [5] D. de la Barrera, Duality on abelian topological groups: The Mackey Problem, PhD Thesis, Universidad Complutense de Madrid, (2015). c© AGT, UPV, 2017 Appl. Gen. Topol. 18, no. 1 43 D. de la Barrera Mayoral [6] D. de la Barrera Mayoral, Q is not Mackey group, Topology Appl. 178 (2014), 265–275. [7] M. J. Chasco, Pontryagin duality for metrizable groups, Arch. Math. (Basel) 70, no. 1 (1998), 22–28. [8] M. J. Chasco, E. Mart́ın-Peinador and V. Tarieladze, On Mackey topology for groups, Stud. Math. 132, no. 3 (1999), 257–284. [9] L. de Leo, Weak and strong topologies in topological abelian groups, PhD Thesis, Uni- versidad Complutense de Madrid, July 2008. [10] J. M. Dı́az Nieto and E. Mart́ın-Peinador, Characteristics of the Mackey topology for Abelian topological groups. J. C. Ferrando and M. López-Pellicer (eds.), Descriptive Topology and Functional Analysis, Springer Proceedings in Mathematics & Statistics 80 (doi: 10.1007/978-3-319-05224-3 7). [11] D. Dikranjan, S. S. Gabriyelyan and V. Tarieladze, Characterizing sequences for pre- compact group topologies, J. Math Anal. Appl. 412 (2014), 505–519. [12] D. Dikranjan, E. Mart́ın-Peinador and V. Tarieladze, A class of metrizable locally quasi- convex groups which are not Mackey, Forum Math. 26 (2014), 723 –757 [13] S. S. Gabriyelyan, Groups of quasi-invariance and the Pontryagin duality, Topology Appl. 157, no. 18, (2010), 2786–2802. [14] E. Mart́ın-Peinador and V. Tarieladze, Mackey topology on locally convex spaces and on locally quasi-convex groups. Similarities and historical remarks, RACSAM 110, no. 2 (2016), 667-679. [15] I. Protasov and E. Zelenyuk, Topologies on abelian groups, Math. USSR Izvestiya, 37, no. 2 (1991). [16] I. Protasov and E. Zelenyuk, Topologies on groups determined by sequences, Mathe- matical Studies, Monograph Series Vol. 4, Lviv, 1999. [17] N. Ya. Vilenkin, The theory of characters of topological Abelian groups with bounded- ness given, Izvestiya Akad. Nauk SSSR. Ser. Mat. 15 (1951), 439–162. c© AGT, UPV, 2017 Appl. Gen. Topol. 18, no. 1 44