<<

arXiv:2105.01810v1 [math.FA] 5 May 2021 otenMteaia nttt fteRsinAaeyof Academy Russian the of Institute Mathematical Southern 3 eateto ahmtc,Mdl atTcnclUnivers Technical East Middle Mathematics, of Department 1 The eioeaosbtenvco lattices vector between operators Levi called pcsaespoe ob el prtr ier vector linear, Archimedean. operators lattices real, vector vector and be Hausdorff, all to topologies stated) supposed otherwise are (unless spaces paper present the preliminaries In and Introduction 1 prope domination convergence, erator order lattice, Banach tice, o atc falodrbuddlna ucinl on functionals linear bounded order all of lattice tor Dom studied. investigation. are under operators of operator classes la an these vector of for topological range properties a and of domain properties the lie order T lattices vector and to lattices. topological related vector notions to of versions lattices operator solid spac vector locally topological from and operators lattices solid spaces; vector locally topological between to lattices vector from erators o n etrlattice vector any For MSC2020: keywords: the investigate We aa Alpay Safak oτ n oooia etrspaces vector topological and re dual order ooe nttt fMteais ooiis,Russia Novosibirsk, Mathematics, of Institute Sobolev 2 cniuu,Lbsu,K,and KB, Lebesgue, -continuous, etrltie oooia etrsae oal oi la solid locally space, vector topological lattice, vector 64,4B2 47L05 46B42, 46A40, 1 dadEmelyanov Eduard , oτ of cniuu/one/opc n eegeop- Lebesgue and -continuous/bounded/compact X a ,2021 6, May n sdntdby denoted is and Abstract X Russia h eeidcmlt vec- complete Dedekind the , 1 1 , 2 vtaaGorokhova Svetlana , X nredistributing in s h Boperators KB the ∼ s n h Levi the and es; cecs Vladikavkaz, Sciences, emi daof idea main he t,ajitop- adjoint rty, h order The . t,Akr,Turkey Ankara, ity, tc between ttice ination X 3 is t- ∼ ∼ continuous part Xn is a of X . Some vector lat- tices may have trivial order duals, for example X∼ = {0} whenever X = Lp[0, 1] with 0

d) pre-Lebesgue if 0 ≤ xn ↑≤ x in X implies that xn is a ς-Cauchy sequence in X.

2 e) Levi/σ-Levi if every increasing ς-bounded net/sequen- ce in X+ has a supremum in X. f) Fatou/σ-Fatou if the topology ς has a base at zero consisting of order/σ-order closed solid sets.

The assumption xα ↓ 0 on the net xα in c) can be replaced o by xα −→ 0. A Lebesgue lattice (X, ς) is Dedekind complete iff order intervals of X are ς-complete [29, Prop.3.16]. Ev- ery Lebesgue lattice is pre-Lebesgue [2, Thm.3.23] and Fa- tou [2, Lem.4.2]. Furthermore, (X, ς) is pre-Lebesgue iff the topological completion (X,ˆ ςˆ) of (X, ς) is Lebesgue [2, Thm.3.26] iff (X,ˆ ςˆ) is pre-Lebesgue and Fatou [2, Thm.4.8]. By [2, Thm.3.22], d) is equivalent to each of the following two conditions: ′ d ) if 0 ≤ xα ↑≤ x holds in X, then xα is a ς-Cauchy net; d′′) every order bounded disjoint sequence in X is ς- convergent to zero. The next well known fact follows directly from d′). Proposition 1.1. Every ς-complete pre-Lebesgue lattice (X, ς) is Dedekind complete. A normed lattice (X, k·k) is called Kantorovich-Banach or KB-space if every norm bounded upward directed set in X+ converges in the norm. Each KB-space is a Levi lattice with order continuous complete norm; each order contin- uous Levi normed lattice is Fatou; and each Levi normed lattice is Dedekind complete. Lattice-normed versions of KB-spaces were recently studied in [8, 9, 11]. g) We call a locally solid lattice (X, ς) by a KB/σ-KB lattice if every increasing ς-bounded net/sequence in X+ is ς-convergent. Clearly, each KB/σ-KB lattice is Levi/σ-Levi and each Levi/σ-Levi lattice is Dedekind complete/σ-complete. Recall that a continuous operator T : h) between two Banach spaces is said to be Dunford- Pettis if T takes weakly null sequences to norm null

3 sequences. It is well known that every weakly on L1(µ) is Dunford-Pettis and that an op- erator is Dunford-Pettis iff it takes weakly Cauchy se- quences to norm convergent sequences [3, Thm.5.79]. i) from a X to a Y is called M-weakly compact if kT xnkY → 0 holds for every norm bounded disjoint sequence xn in X. j) from a Banach space Y to a Banach lattice X is called L-weakly compact whenever kxnkX → 0 holds for ev- ery disjoint sequence xn in the solid hull of T (UY ), where UY is the closed unit ball of Y . An operator T from a vector lattice X to a (Y, τ) is called τ k) oτ-continuous/σoτ-continuous if T xα −→ 0 for every o net/sequence xα such that xα −→ 0 [27]. Replacement of o-null nets/sequences by uo-null ones above gives the definitions of uoτ-continuous/σuoτ-continuous op- erators. L-/M-weakly compact operators are weakly compact (cf. [3, Thm.5.61]) and the norm limit of a sequence of L-/M- weakly compact operators is again L-/M-weakly compact (cf. [3, Thm.5.65]). For further unexplained terminology and notions, we refer to [2, 3, 4, 5, 25, 28, 33, 34]. Various versions of Banach lattice properties like a prop- erty to be a KB-space were investigated recently (see, e.g., [1, 6, 7, 8, 9, 12, 13, 14, 17, 18, 20, 21, 22, 23, 25, 27, 29, 30, 31]). In the present paper we continue the study of operator versions of several topological/order properties, focusing on locally solid lattices. The main idea behind operator versions consists in a redistribution of topolog- ical and order properties of a topological vector lattice between the domain and range of the operator under in- vestigation (like in the case of Dunford-Pettis and L-/M- weakly compact operators). As the is not topological in general [13, 26], the most important op-

4 erator versions emerge when both o- and ς-convergences are involved simultaneously. Definition 1.1. Let T be an operator from a vector lattice X to a topological vector space (Y, τ). We say that: τ (a) T is τ-Lebesgue/στ-Lebesgue if T xα −→ 0 for every net/ sequence xα such that xα ↓ 0; T is quasi τ- Lebesgue/ στ-Lebesgue if T xα is τ-Cauchy for every net/sequence xα in X+ satisfying xα ↑≤ x ∈ X. If there is no confusion with the choice of the topology τ on Y , we call τ-Lebesgue operators by Lebesgue etc. (b) T is oτ-bounded/oτ-compact if T [0, x] is a τ-bounded/ τ-totally bounded subset of Y for each x ∈ X+. If additionally X =(X, ς) is a locally solid lattice, (c) T is KB/σ-KB if, for every ς-bounded increasing net/sequence xα in X+, there exists (not necessarily τ unique) x ∈ X such that T xα −→ T x. (d) T is quasi KB/quasi σ-KB if T takes ς-bounded increasing nets/sequences in X+ to τ-Cauchy nets. If X and Y are vector lattices with (X, ς) locally solid, (e) T is Levi/σ-Levi if, for every ς-bounded increasing net/sequence xα in X+, there exists (not necessarily o unique) x ∈ X such that T xα −→ T x. (f) T is quasi Levi/quasi σ-Levi if T takes ς-bounded increasing nets/sequences in X+ to o-Cauchy nets. Replacement of decreasing o-null nets/sequences by uo- null ones in (a) and o-convergent (o-Cauchy) nets/sequen- ces by (uo-Cauchy) uo-convergent ones in (e) and (f) above gives the definitions of uoτ-continuous and of(quasi) uo-Levi operators respectively. In our approach, we focus on: ∗ modification of nets/sets of operators domains in (a), (b), (d), (e), and (f); ∗∗ information which operators provide about conver- gences in their domains/ranges in (c), (e), and (f).

5 Remark 1.1. a) The identity operator in a locally solid lattice (Y, τ) is Lebesgue/KB/Levi iff (Y, τ) is Lebes- gue/KB/Levi. This motivates the terminology. Cle- arly, every oτ-continuous/σoτ-continuous operator is τ-Lebesgue/στ-Lebesgue. By Lemma 2.1, any regu- lar operator to a locally solid lattice is Lebesgue/σ- Lebesgue iff T is oτ-continuous/σoτ-continuous; and every regular Lebesgue operator to a locally solid lat- tice is quasi Lebesgue by Theorem 2.2. b) Each regular operator from a vector lattice X to a lo- cally solid lattice (Y, τ) is oτ-bounded by Proposition 2.1. In the case of normed range space Y , oτ-bounded operators are also known as interval-bounded (cf. [28, Def.3.4.1]). Like in [28, Lem.3.4.2], each oτ- T from a vector lattice X to a topological vector space (Y, τ) possesses adjoint T ◦ : Y ′ → X∼ given by T ◦y′ = y′ ◦ T for all y′ ∈ Y ′. c) In the case when X is also a topological vector lat- tice, the τ-continuity of operator T is not assumed in (b) of Definition 1.1. For example, the rank one dis- T x ∞ x e c , continuous operator := (Pk=1 k) 1 in ( 00 k·k) is oτ-compact and oτ-continuous yet not compact. Each continuous operator T from a discrete Dedekind com- plete locally convex Lebesgue lattice to a topological vector space is oτ-compact by the Kawai theorem (cf. [2, Cor.6.57)]. Every Dunford-Pettis operator from a Banach lattice to a Banach space is o-weakly-compact (cf. [3, Thm 5.91]). d) Each KB-operator is quasi KB; each quasi (σ-) KB- operator is quasi (σ-) Lebesgue; and each continu- ous operator from a KB-space to a topological vector space is KB. It is well known that the identity op- erator I in a Banach lattice is KB iff I is σ-KB iff I is quasi KB. Proposition 1.2 shows that the no- tions of quasi KB and quasi σ-KB operator coincide. The most important reason for using the term KB-

6 operator for (c) of Definition 1.1 is existence of limits of ς-bounded increasing nets. Some authors (see, e.g., [7, 12, 31]) use the term “KB-operator” for (d) of Definition 1.1 which is slightly confusing because it says nothing about existence of limits of topologically bounded increasing nets and all continuous finite-rank operators in every Banach lattice satisfy (d). Each or- der bounded operator from a Banach lattice to a KB- space is quasi KB. However, if we take X =(c00, k·kl) and Y =(c00, k·kp) with l,p ∈ [1, ∞], the identity op- erator I is quasi KB iff l ≤ p< ∞. e) It is clear that every compact operator T from a Banach lattice X to a Banach space Y is oτ-compact. However, compact operator need not to be Lebesgue (cf. Example 3.1). In particular, oτ-compact operator need not to be oτ-continuous.

Example 1.1. Let (c(R), k·k∞) be the Banach lattice of all R-valued functions on R such that for every f ∈ c(R) there exists af ∈ R for which the set {r ∈ R : |f(r) − af | ≥ ε} is finite for each ε > 0. Then the identity operator I in c(R) is σoτ-continuous and quasi σ-Lebesgue yet neither Lebesgue nor quasi Lebesgue. Example 1.2. Recall that, for a nonempty set H, the vec- tor space c00(H) of all finitely supported R-valued functions on H is a Dedekind complete vector lattice. Furthermore, any vector space X is linearly isomorphic to c00(H), where H is a Hamel basis for X. As each order interval of c00(H) lies in a finite-dimensional subspace of c00(H), every op- erator from c00(H) to any topological vector space (Y, τ) is oτ-continuous, oτ-bounded, and oτ-compact. On the other hand, the identity operator in (c00, k·k∞) is neither quasi σ-KB nor quasi σ-Levi.

Remark 1.2. It is well known that, if a net yα of a topo- logical vector space (Y, τ) is not τ-Cauchy, then there ex- ist U ∈ τ(0) and an increasing sequence αn such that yαn+1 − yαn 6∈ U for each n (see, e.g., [2, Lem.2.5]).

7 Proposition 1.2. An operator T from a locally solid lat- tice (X, ς) to a topological vector space (Y, τ) is quasi KB iff T is quasi σ-KB. Proof. The necessity is trivial. For the sufficiency, sup- pose T is not quasi KB. Then there exists a ς-bounded increasing net xα in X+ such that T xα is not τ-Cauchy in Y . It follows from Remark 1.2 that for some increasing sequence αn and some U ∈ τ(0) N T xαn+1 − T xαn 6∈ U (∀n ∈ ) (1)

Since the sequence xαn is increasing and ς-bounded, the condition (1) implies that T is not quasi σ-KB. Corollary 1.1. Every ς-complete σ-KB lattice (X, ς) is a KB lattice. Proof. The identity operator I on X is σ-KB and hence quasi σ-KB. By Proposition 1.2, I is quasi KB. Then every ς-bounded increasing net in X+ is ς-Cauchy, and hence is ς-convergent. Remark 1.3. The following fact (cf., e.g., [10, Prop.1.1]) is well known:

a) ρ(yα,y) → 0 in a metric space (Y, ρ) iff, for every

subnet yαβ of the net yα, there exists a (not necessary increasing) sequence βk of indices with ρ(yα ,y) → 0. βk Application of a) to Cauchy nets/sequences in (Y, ρ) gives that:

b) a net/sequence yα of a metric space (Y, ρ) is Cauchy

iff, for every subnet/subsequence yαβ of yα, there ex- ists a sequence βk of indices such that the sequence yα is ρ-Cauchy. βk Application of b) and Proposition 1.2 gives that, for an operator T from a locally solid lattice (X, ς) to a metric vector space (Y, ρ), the following are equivalent: (i) T is quasi KB;

8 (ii) Every ς-bounded increasing sequence xn in X+ has a

subsequence xnk such that T xnk is ρ-Cauchy in Y . Another application of b) to an operator T from a locally solid lattice (X, ς) to a metric vector space (Y, ρ) gives the equivalence of the following conditions: (i)′ T is KB; ′ (ii) For any ς-bounded increasing net xα in X+, there exist an element x ∈ X and a sequence βk of indices such that ρ(T xα , T x) → 0. βk In many cases, like for (quasi) KB or Levi operators, the lattice structure in the domain/range of operator can be relaxed to the ordered space structure [7, 11, 19, 21], or combined with the lattice-norm structure [1, 8, 9, 11]. Such generalizations are not included in the present pa- per. In Section 2 we investigate operators whose domains and/or ranges are locally solid lattices. Section 3 is de- voted to operators between Banach lattices.

2 Operators between locally solid lattices

In this section, we study mostly operators whose domains and/or ranges are locally solid lattices. Observe first that the sets LLeb(X,Y ), Loτ (X,Y ), Loτb(X,Y ), and Loτc(X,Y ) of Lebesgue, oτ-continuous, oτ-bounded, and oτ-compact operators respectively from a vector lattice X to a topo- logical vector space (Y, τ) are vector spaces and:

(∗) Loτ (X,Y ) ⊆ LLeb(X,Y );

(∗∗) Loτc(X,Y ) ⊆ Loτb(X,Y ) since every totally bounded subset in Y is bounded. Theorem 2.1. Let X be a vector lattice and (Y, τ) be a locally convex Lebesgue lattice which is either Dedekind complete or τ-complete. Then Loτc(X,Y ) T Lr(X,Y ) is a band of the lattice Lr(X,Y ) of all regular operators from X to Y .

9 Proof. Since every Lebesgue lattice is pre-Lebesgue by [2, Thm.3.23], and every topologically complete pre-Lebesgue lattice is Dedekind complete by Proposition 1.1, the lat- tice (Y, τ) is Dedekind complete in either case. By [3, Thm.5.10], for each x ∈ X+, the set

C(x)= {T ∈ Lb(X,Y ): T [0, x] is τ-totally bounded} is a band of the Dedekind complete lattice Lb(X,Y ) of all order bounded operators from X to Y . Since Lr(X,Y )= L (X,Y ), the set L (X,Y ) L (X,Y )= C(x) is b oτc T r Tx∈X+ also a band of Lr(X,Y ) as desired. The following two propositions might be known. We include their elementary proofs as we did not find appro- priate references. Proposition 2.1. Each regular operator T from a vector lattice X to a locally solid lattice (Y, τ) is oτ-bounded. Proof. Without lost of generality, assume T ≥ 0. Let x ∈ X+, U ∈ τ(0), and V be a solid τ-neighborhood of zero of Y with V ⊆ U. Then T x ∈ λV for all λ ≥ λ0 for some λ0 > 0 and hence T [0, x] ⊆ [0, T x] ⊆ λV ⊆ λU for all λ ≥ λ0. Since U ∈ τ(0) was arbitrary, T is oτ- bounded. Proposition 2.2. An order continuous regular operator T from a vector lattice X to a locally convex-solid lattice (Y, τ) is Lebesgue iff T is weakly Lebesgue. Proof. Without lost of generality, we can suppose T ≥ 0. The necessity is trivial. For the sufficiency, assume that T is weakly Lebesgue (i.e., T : X → Y is Lebesgue with ′ respect to the σ(Y,Y ) on Y ). Let xα ↓ 0 ′ in X. Then T xα ↓ 0 in Y . Since T is σ(Y,Y )-Lebesgue, σ(Y,Y ′) T xα −−−−→ 0. By using Dini-type arguments (cf. [3, τ Thm.3.52]), we conclude T xα −→ 0, as desired.

10 We do not know any example of an order continuous oτ- bounded weakly Lebesgue operator from a vector lattice to a locally solid lattice that is not Lebesgue. Proposition 2.3. A weakly continuous regular operator T from a Lebesgue (σ-Lebesgue) lattice (X, ς) to a locally solid lattice (Y, τ) is Lebesgue (σ-Lebesgue). Proof. As arguments are similar, we restrict ourselves to the Lebesgue case. Without lost of generality, assume T ≥ ς 0. Let xα ↓ 0 in X. Since (X, ς) is Lebesgue, xα −→ 0 and σ(X,X′) hence xα −−−−→ 0. The weak continuity of T implies ′ σ(Y,Y ) τ T xα −−−−→ 0. Since T xα ↓ 0 in Y , it follows T xα −→ 0, as in the proof of Proposition 2.2. Theorem 2.2. Each regular Lebesgue operator T from a vector lattice X to a locally solid lattice (Y, τ) is quasi Lebesgue. Proof. Without lost of generality, we can suppose T ≥ 0. Let xα ↑≤ x ∈ X, so uα := x − xα ↓≥ 0. Therefore (uα − z)α;z∈L ↓ 0, where L = {z ∈ X :(∀α ∈ A)[z ≤ uα]} is the upward directed set of lower bounds of the net uα (cf. [16, Thm.1.2]). Since T is Lebesgue and T ≥ 0, τ (Tuα − T z)α;z∈L −→ 0 & (Tuα − T z)α;z∈L ↓≥ 0. (2)

It follows from (2), that (Tuα − T z)α;z∈L ↓ 0. Then Tuα and hence T xα is τ-Cauchy, as desired. We do not know any example of an oτ-bounded Lebesgue operator which is not quasi Lebesgue. In contrast to The- orem 3.24 of [2], the converse of Theorem 2.2 is false even if Y = R due to Example 3.1. Recall that an Archimedean vector lattice X is called laterally (σ-) complete whenever every (countable) subset of pairwise disjoint vectors of X+ has a supremum. Every laterally complete vector lattice X contains a weak and every band of such an X is a principal band (cf. [2, Thm.7.2]). Furthermore, by the Veksler – Geiler

11 theorem (cf. [2, Thm.7.4]), if a vector lattice X is laterally (σ-) complete, then X satisfies the (principal projection) projection property. A vector lattice that is both laterally and Dedekind (σ- ) complete is referred to as universally (σ-) complete. It follows from [2, Thm.7.4] that a vector lattice X is univer- sally complete iff X is Dedekind σ-complete and laterally complete iff X is uniformly complete and laterally com- plete (cf. [2, Thm.7.5]). Similarly, a laterally σ-complete vector lattice X is Dedekind σ-complete iff X is uniformly complete. A universal completion of a vector lattice X is a later- ally and Dedekind complete vector lattice Xu which con- tains X as an order dense sublattice. Every vector lattice has unique universal completion (cf. [2, Thm.7.21]). A laterally complete vector lattice X is discrete iff X is lattice isomorphic to RS for some nonempty set S iff X admits a Hausdorff locally convex-solid Lebesgue topol- ∼ ogy iff the space Xn of order continuous functionals on X separates X (cf. [2, Thm.7.48]); in each of these cases X = Xu. Definition 2.1. A topological vector lattice (Y, τ) is called τ-laterally (σ-) complete whenever every τ-bounded (count- able) subset of pairwise disjoint vectors of X+ has a supre- mum. Every laterally (σ-) complete topological vector lattice (Y, τ) is τ-laterally (σ-) complete. Every Dedekind com- plete AM-space X with an order unit is τ-laterally com- plete with respect to the norm. Example 2.1. Let X be a vector lattice of real functions on R such that each f ∈ X may differ from a constant say af on a countable subset of R and f − af IR ∈ ℓ1(R) for each f ∈ X. (i) The vector lattice X is not Dedekind σ-complete, as fn := IR\{1,2,...,n} ↓≥ 0 yet inf fn does not exist in X. n∈N

12 (ii) The vector lattice X is a Banach space with respect to the norm kfk := |af | + kf − af IRk1. (iii) The vector lattice X is not τ-laterally σ-complete with respect to the norm topology on X. Indeed, the norm bounded countable set of pairwise disjoint orths en = I{n} of X+ has no supremum in X. (iv) The identity operator I on X is not σ-Lebesgue, as 1 I 1 I n+1 n R\{1,2,...,n} ↓ 0 in X yet k n R\{1,2,...,n}k = n 6→ 0. Proposition 2.4. Any Levi (σ-Levi) lattice is τ-laterally (σ-) complete and Dedekind (σ-) complete. Proof. We consider the Levi case only, because the σ-Levi case is similar. The Dedekind completeness of X is trivial. Let D be a τ-bounded subset of pairwise disjoint positive vectors of a Levi lattice (Y, τ). Then the collection D∨ of suprema of finite subsets of D forms an increasing τ- bounded net indexed by the set Pfin(D) of all finite subsets of D directed by inclusion. Since (Y, τ) is Levi, D∨ ↑ d for some d ∈ Y . It follows from sup D = sup D∨ = d that (Y, τ) is τ-laterally complete. Example 2.2. The locally solid lattice (RS, τ), where τ is the product topology on the vector lattice RS of real func- tions on a set S is locally convex, Lebesgue, Levi, and universally complete. Proposition 2.5. Each regular operator T from a later- ally σ-complete vector lattice X to a σ-Lebesgue lattice (Y, τ) is σ–Lebesgue. Proof. Without lost of generality, we can suppose T ≥ 0. Let xn ↓ 0 in X. By Theorem 7.8 of [2], T is σ-order continuous, and hence T xn ↓ 0 in Y . Since (Y, τ) is σ– τ Lebesgue, T xn −→ 0. We do not know whether or not every positive operator T from a laterally complete vector lattice X to a Lebesgue lattice (Y, τ) is Lebesgue.

13 Each laterally σ-complete locally solid lattice is σ-Le- besgue and pre-Lebesgue [2, Thm.7.49]. Proposition 2.6. Let T be a continuous operator from a laterally σ-complete locally solid lattice (X, ς) to a topolog- ical vector space (Y, τ). Then T is σ-Lebesgue in each of the following cases: (i) X is discrete. (ii) (X, ς) is metrizable.

Proof. Let xn ↓ 0 in X. In case (i), (X, ς) is σ-Lebesgue by [2, Thm.7.49]. The result follows because T is ςτ- continuous. In case (ii), the result follows from (i) as every infinite dimensional laterally σ-complete metrizable locally solid lattice is lattice isomorphic to RN. Since (X,uς) is pre-Lebesgue iff (X, ς) is pre-Lebesgue [30, Cor.4.3], the statement of Proposition 2.6 is also true when (X, ς) is replaced by (X,uς). Remark 2.1. Let T be a positive operator from a locally solid lattice (X, ς) to a laterally σ-complete locally solid lattice (Y, τ) and let 0 ≤ xα ↑≤ x in X. Since (Y, τ) is pre-Lebesgue by [2, 7.49], it follows from 0 ≤ T xα ↑≤ T x that T xα is τ-Cauchy in Y and hence T is quasi Lebesgue. Remark 2.2. Let T be a continuous operator from a lat- erally σ-complete locally solid lattice (X, ς) to a locally solid lattice (Y, τ).

a) Suppose T ≥ 0 and 0 ≤ xα ↑≤ x in X. Since (X, ς) is pre-Lebesgue by [2, 7.49] then xα is ς-Cauchy in X and hence T xα is τ-Cauchy in Y meaning that T is quasi Lebesgue.

b) Suppose (X, ς) is Fatou, and xα ↓ 0 in X. Then by [2, ς τ 7.50] (X, ς) is Lebesgue, so xα −→ 0 and hence T xα −→ 0 meaning that T is Lebesgue. c) Suppose ς is a metrizable locally solid topology and xα ↓ 0 in X. Then (X, ς) is Lebesgue by [2, Thm.7.55].

14 ς τ Thus xα −→ 0 and hence T xα −→ 0 meaning that T is Lebesgue. d) If(X, ς) is a Fr´echet lattice then ς is uniquely defined and every regular T : (X, ς) → (Y, τ) is continuous (cf. [2, Thm.5.19 and Thm.5.21]). In particular, every regular operator from a laterally σ-complete Fr´echet lattice X(ς) to a locally solid lattice (Y, τ) is Lebesgue. Proposition 2.7. Any continuous operator T from a lo- cally solid lattice (X, ς) to a topological vector space (Y, τ) is oτ-bounded.

Proof. Let x ∈ X+, U ∈ τ(0), and V be a solid ς-neighbor- hood of zero of X with V ⊆ T −1U. Then there exists λ0 > 0 such that x ∈ λV for all λ ≥ λ0 and hence [0, x] ⊆ −1 λV ⊆ λT (U) for all λ ≥ λ0. Thus T [0, x] ⊆ λU for all λ ≥ λ0, which shows that T is oτ-bounded. It is worth mentioning here that an oτ-bounded operator T from Dedekind σ-complete vector lattice X to a normed lattice Y is σ-Lebesgue iff T is order-weakly compact (cf. [27, Cor.1]). The following lemma can be considered as an extension of [27, Lem.1] to locally solid lattices. Lemma 2.1. A regular operator T from a vector lattice X to a locally solid lattice (Y, τ) is Lebesgue/σ-Lebesgue iff T is oτ-continuous/σoτ-continuous. Proof. Without lost of generality, we suppose T ≥ 0. As the σ-Lebesgue case is similar, we consider only Lebesgue operators. The sufficiency is routine. For the necessity, o assume T is Lebesgue and xα −→ 0 in X. Take a net zβ in X with zβ ↓ 0 such that, for each β, there exists αβ with |xα| ≤ zβ for α ≥ αβ. It follows from T ≥ 0 that |T xα| ≤ T |xα| ≤ T zβ for α ≥ αβ. As T is Lebesgue, τ τ T zβ −→ 0, which implies T xα −→ 0, because the topology τ is locally solid. Corollary 2.1. An order bounded operator T from a vec- tor lattice to a Dedekind complete locally solid lattice is

15 Lebesgue/σ-Lebesgue iff T is oτ-continuous/σoτ-continu- ous. Now we investigate some properties of adjoint opera- tors. Recall that, for a locally solid lattice (X, ς), the absolute weak topology |σ|(X∼,X) on X∼ is the locally convex-solid topology generated by the collection of Riesz seminorms {ρx(f) = |f|(|x|): x ∈ X}. The locally solid lattice (X∼, |σ|(X∼,X)) is Lebesgue, Levi, and Fatou [24, Prop.81C]. We begin with the following technical lemma. Lemma 2.2. Let T :(X, ς) → (Y, τ) be an order bounded continuous operator from a locally solid lattice (X, ς) to a Dedekind complete locally solid lattice (Y, τ). Then the topological adjoint T ′ :(Y ′, |σ|(Y ∼,Y )) → (X′, |σ|(X∼,X)) is also continuous. ∼ |σ|(Y ,Y ) ∼ Proof. Indeed, for fα −−−−−−→ 0 in Y , it follows from

′ ′ ′ ′ ρx(T fα)= h|T fα|, |x|i ≤ h|T ||fα|, |x|i ≤ h|T | |fα|, |x|i =

h|fα|, |T ||x|i = |fα|(|T ||x|)= ρ|T ||x|(fα) → 0 (3) ∼ ′ |σ|(X ,X) that T fα −−−−−−→ 0. The second inequality in (3) fol- lows from the existence of the modulus of T and from the observation ±T ≤ |T | ⇒ ±T ′ ≤ |T |′ (cf. [3, p.67]) which implies |T ′|≤|T |′. Theorem 2.3. Let T : X → Y be an order bounded oper- ator from a vector lattice X to a Dedekind complete vector lattice Y , and T ′ : (Y ∼, |σ|(Y ∼,Y )) → (X∼, |σ|(X∼,X)) the correspondent topological adjoint of T . Then T ′ is oτ- bounded, oτ-continuous, Levi, and KB. Proof. The oτ-boundedness of T ′ follows from Lemma 2.2 by Proposition 2.7. Since T is regular, without lost of generality, we can suppose T ≥ 0 (and hence T ′ ≥ 0). o ∼ ′ ′ o Let fα −→ 0 in Y . As T is order continuous, T fα −→ 0 in X∼, and since (X∼, |σ|(X∼,X)) is a Lebesgue lattice, ∼ ′ |σ|(X ,X) ′ we obtain T fα −−−−−−→ 0 and hence T is oτ-continuous.

16 ∼ Let fα be a positive |σ|(Y ,Y )-bounded increasing net ∼ ∼ ∼ in Y . Since (Y , |σ|(Y ,Y )) is Levi, fα ↑ f for some ∼ ′ ′ ′ ∼ f ∈ Y . As T is order continuous, T fα ↑ T f in X , in particular T ′ is Levi. Since the locally solid lattice ∼ ∼ ∼ ′ |σ|(X ,X) ′ (X , |σ|(X ,X)) is Lebesgue, T fα −−−−−−→ T f and hence T ′ : (Y ∼, |σ|(Y ∼,Y )) → (X∼, |σ|(X∼,X)) is KB.

∼ ∼ The natural embedding J of a vector lattice X to (Xn )n ∼ is defined by (Jx)(f)= f(x) for all f ∈ Xn , x ∈ X. The mapping J is an order continuous lattice homomorphism (cf. e.g. [3, Thm.1.70]). By the Nakano theorem (cf. [3, Thm.1.71]), J is one-to-one and onto (i.e. X is perfect) ∼ iff Xn separates X and whenever a net xα of X+ satisfies ∼ xα ↑ and sup f(xα) < ∞ for each 0 ≤ f ∈ Xn , then there exists some x ∈ X satisfying xα ↑ x in X. Lemma 2.3. Let T be an order bounded operator from a vector lattice X to a vector lattice Y . The second order ∼∼ ∼ ∼ adjoint T , while restricted to (Xn ) , ∼∼ ∼ ∼ ∼ ∼ ∼ T ∼ ∼ ∼ ∼ ∼ ((Xn ) , |σ|((Xn ) ,Xn )) −−→ ((Y ) , |σ|((Yn ) ,Yn )) is continuous. ∼ ∼ ∼ ∼ ∼ |σ|((Xn ) ,Xn ) Proof. Let (Xn ) ∋ xα −−−−−−−−−→ 0. By use of the fact |T ∼∼|≤|T ∼|∼ it follows

∼∼ ∼ ∼ ∼ h|T xα|, |y|i≤ h|T | |xα|, |y|i≤ h|xα|, |T ||y|i (4)

∼ ∼ ∼ ∼ |σ|((Xn ) ,Xn ) for all y ∈ Yn . Since xα −−−−−−−−−→ 0, it follows from (4) ∼ ∼ ∼ ∼∼ |σ|((Yn ) ,Yn ) ∼∼ that T xα −−−−−−−−→ 0 and hence the operator T is ∼ ∼ ∼ ∼ ∼ ∼ |σ|((Xn ) ,Xn ) to |σ|((Yn ) ,Yn ) continuous. Theorem 2.4. Let T be an order bounded operator from a vector lattice X to a vector lattice Y . Then ∼∼ ∼ ∼ ∼ ∼ ∼ T ∼ ∼ ∼ ∼ ∼ ((Xn )n , |σ|((Xn )n ,Xn )) −−→ ((Yn )n , |σ|((Yn )n ,Yn )). is a continuous, Lebesgue, Levi, and KB operator.

17 Proof. The continuity follows from Lemma 2.3 by restrict- ∼∼ ∼ ∼ ing of the second order adjoint T to (Xn )n . ∼ ∼ ∼∼ Let xα ↓ 0 in (Xn )n . As T is order continuous, ∼∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼ T xα ↓ 0 in (Yn )n . Since ((Yn )n , |σ|((Yn )n ,Yn )) is ∼ ∼ ∼ ∼∼ |σ|((Yn )n ,Yn ) Lebesgue, T xα −−−−−−−−→ 0 showing that the opera- tor is Lebesgue. ∼ ∼ ∼ Let xα be a positive increasing |σ|((Xn )n ,Xn )-bounded ∼ ∼ ∼ ∼ ∼ ∼ ∼ net in (Xn )n . Since ((Xn )n , |σ|((Xn )n ,Xn )) is Levi, the ∼ ∼ net xα has a supremum in (Xn )n , say x. Thus xα ↑ x ∼∼ ∼∼ ∼∼ and T xα ↑ T x as T is order continuous. Thus the operator is Levi. ∼ ∼ ∼ ∼ ∼ ∼ ∼ Since ((Xn )n , |σ|((Xn )n ,Xn )) is Levi, xα ↑ x in (Xn )n . ∼∼ o ∼∼ Then T xα −→ T x by order continuity of the operator ∼∼ ∼ ∼ ∼ ∼ ∼ T . As ((Yn )n , |σ|((Yn )n ,Yn )) is a Lebesgue lattice, ∼ ∼ ∼ ∼∼ |σ|((Yn )n ,Yn ) ∼∼ T xα −−−−−−−−→ T x and the operator is KB. A vector sublattice Z of a vector lattice X is called regular, if the embedding of Z into X preserves arbitrary suprema and infima. Order ideals and order dense vector sublattices of a vector lattice X are regular [2, Thm.1.23]. ∼ ∼ Corollary 2.2. Let X = Xn , and Y be a vector lattice. ∼ Then each order bounded operator T : X → (Y, |σ|(Y,Yn )) is Lebesgue.

Proof. Let xα ↓ 0 in X. Since the natural embedding J : X → X∼∼ is one-to-one and J(X) is a regular sublattice ∼∼ ∼∼ of X (cf. [2, Thm.1.67]), then T = T ◦J and Jxα ↓ 0 ∼ ∼ ∼ ∼∼ ∼∼ |σ|((Xn )n ,Xn ) ∼∼ in X . Thus T xα = T (Jxα) −−−−−−−−−→ 0 since T is Lebesgue by Theorem 2.4.

We also mention the following question. Let T be Lebes- gue, oτ-continuous, oτ-bounded, oτ-compact, KB, or Levi. Does T ∼∼ satisfy the same property? The answer is nega- tive, when T maps a Banach lattice X to a Banach lattice Y , in the following cases:

18 (∗) for Lebesgue and for oτ-continuous operators, since the identity I : c0 → c0 is oτ-continuous yet its second order adjoint I : ℓ∞ → ℓ∞ is not even Lebesgue; (∗∗) for oτ-compact operators, since the natural embed- ∼∼ ∗∗ ding J : c0 → ℓ∞ is oτ-compact but J : ℓ∞ → ℓ∞ is not oτ-compact. Next we discuss the domination properties of positive operators. Let T and S be positive operators between vector lattices X and Y satisfying 0 ≤ S ≤ T . Question 2.1. When does the assumption that T is Lebes- gue, oτ-continuous, oτ-bounded, oτ-compact, KB, or Levi imply that S has the same property? Question 2.1 has positive answers in the following cases: Trivially, for Lebesgue; σ-Lebesgue; and oτ-bounded operators from a vector lattice to a locally solid lattice. For oτ-compact operators from a vector lattice X to a locally convex Lebesgue lattice (Y, τ) which is either Dedekind complete or τ-complete by Theorem 2.1. For o|σ|(Y,Y ′)-compact operators from a Banach lat- tice X to a Banach lattice Y [3, Thm.5.11]. For KB/σ-KB lattice homomorphisms between lo- cally solid lattices by Corollary 2.3. For quasi KB operators between locally solid lattices by Theorem 2.6. For quasi Levi operators from a locally solid lattice to a vector lattice by Theorem 2.7. We remind that the modulus |T | of an order bounded disjointness preserving operator T between vector lattices X and Y exists and satisfies |T ||x| = |T |x|| = |T x| for all x ∈ X [3, Thm.2.40]; there exist lattice homomorphisms R1, R2 : X → Y with T = R1 − R2 (cf. [3, Exer.1, p.130]. Theorem 2.5. Let T be an order bounded disjointness preserving KB/σ-KB operator between locally solid lat- tices (X, ς) and (Y, τ). If |S|≤|T | then S is KB/σ-KB.

19 Proof. Take a ς-bounded increasing net/sequence xα in τ X+. Then T (xα − x) −→ 0 for some x ∈ X. It follows from τ |S(xα − x)|≤|S||xα − x|≤|T ||xα − x| = |T xα − T x| −→ 0 τ that Sxα −→ Sx, as desired. Since 0 ≤ S ≤ T with a lattice homomorphism T im- plies that S is also a lattice homomorphism, the next result is a direct consequence of Theorem 2.5. Corollary 2.3. Let T be a KB/σ-KB lattice homomor- phism between locally solid lattices (X, ς) and (Y, τ). Then each S satisfying 0 ≤ S ≤ T is also a KB/σ-KB lattice homomorphism. Theorem 2.6. Let T be a positive quasi KB operator between locally solid lattices (X, ς) and (Y, τ). Then each operator S : X → Y with 0 ≤ S ≤ T is also quasi KB.

Proof. Let 0 ≤ S ≤ T and xα be an increasing ς-bounded net in X+. Since T ≥ 0 then T xα ↑. Since T is quasi KB then T xα is τ-Cauchy. Take a U ∈ τ(0) and a solid neighborhood V ∈ τ(0) with V − V ⊆ U. There exists α0 satisfying T (xα − xβ) ∈ V for all α, β ≥ α0. In particular,

T (xα − xα0 ) ∈ V for all α ≥ α0. Since 0 ≤ S ≤ T and V is solid, S(xα − xα0 ) ∈ V for all α ≥ α0. Thus, we obtain

Sxα −Sxβ = S(xα −xα0 )−S(xβ −xα0 ) ∈ V −V ⊆ U (5) for all α, β ≥ α0. Since U ∈ τ(0) was taken arbitrary, (5) implies that Sxα is also τ-Cauchy. Theorem 2.7. Let T be a positive quasi Levi operator from a locally solid lattice (X, ς) to a vector lattice Y . Then each operator S : X → Y with 0 ≤ S ≤ T is also quasi Levi.

Proof. Let T : X → Y be quasi Levi and xα be a positive ς-bounded increasing net in (X, ς). Then T xα is o-Cauchy in Y . So, there exists a net zβ ↓ in Y such that, for each

20 β, there exists αβ such that |T xα1 − T xα2 | ≤ zβ for all α1,α2 ≥ αβ. Choosing α1,α2 ≥ αβ for a fixed αβ we have

Sxα1 − Sxα2 ≤ S(xα1 − xαβ ) ≤ T (xα1 − xαβ ) ≤ zβ, and similarly

Sxα2 − Sxα1 ≤ S(xα2 − xαβ ) ≤ T (xα2 − xαβ ) ≤ zβ.

Thus |Sxα1 − Sxα2 | ≤ zβ for all α1,α2 ≥ αβ showing that the operator S is also quasi Levi. A locally solid lattice (X, ς) is a regular sublattice of its ς-completion Xˆ iff every ς-Cauchy net xα in X+ such that o ς xα −→ 0 in X satisfies xα −→ 0 [2, Thm.2.41]. Lemma 2.4. Let Z be an order dense and majorizing sub- lattice of a vector lattice X and T be an oτ-continuous op- erator from X to a topological vector space (Y, τ). Then the restriction T |Z : Z → Y is also oτ-continuous. Proof. Under the assumptions of the lemma, for each net o o zα in Z, zα −→ 0 in Z iff zα −→ 0 in X, by [25, Thm.2.8]. Therefore the result follows directly from the definition of an oτ-continuous operator. Situation is even better for uo-null nets. Let Z be a regular uo uo sublattice of X then xα −→ 0 in Z iff xα −→ 0 in X by [25, Thm.3.2], and hence, for a uoτ-continuous operator T from a vector lattice X to a topological vector space (Y, τ), the restriction T |Z : Z → Y is also uoτ-continuous. Lemma 2.5. Let T be a regular Lebesgue operator from a vector lattice X to a locally solid lattice (Y, τ). If Z is an order dense and majorizing sublattice of X, then the restriction T |Z : Z → Y is oτ-continuous. Proof. T is oτ-continuous by Lemma 2.1. It follows from Lemma 2.4 that T |Z : Z → Y is oτ-continuous. Since a vector lattice X is order dense and majorizing in Xδ, the next result follows immediately from Lemma 2.5.

21 Proposition 2.8. If an operator T from the Dedekind completion Xδ of a vector lattice X to a locally solid lattice (Y, τ) is Lebesgue then the operator T |X is oτ-continuous. Each order continuous lattice homomorphism T from a Fatou lattice (X, ς) to a Lebesgue lattice (Y, τ) is ςτ- continuous on order intervals of X by [2, Thm.4.23]. Since T is also Lebesgue under the given conditions, the follow- ing can be considered as a generalization of [2, Thm.4.23]. Theorem 2.8. Let (X, ς) be a Fatou lattice, (Y, τ) a Lebes- gue lattice, and T : X → Y a Lebesgue lattice homomor- phism. Then T is ςτ-continuous when restricted to order bounded subsets in X. Proof. Let N be a base of solid neighborhoods for τ(0). Since T is a Lebesgue lattice homomorphism, the family −1 {T (V )}V ∈N is a base of solid neighborhoods of zero for a Lebesgue (not necessary Hausdorff) topology ζ on X. Since ς is Fatou and ζ is Lebesgue, then on any order bounded subset of X the topology induced by ς is finer than the topology induced by ζ due to Theorem 4.21 of [2]. Clearly, T : X → Y is ζτ-continuous. Thus, T is also ςτ-continuous on order bounded subsets of X. Recall that, for a locally solid lattice (X, ς), a vector vˆ ∈ Xˆ+ is called an upper element of X whenever there ςˆ exists an increasing net uα in X+ such that uα → vˆ (cf. [2, Def.5.1]). In a metrizable locally solid lattice (X, ς) upper elements can be described by sequences: a vectorv ˆ ∈ Xˆ+ is an upper element of X iff there exists an increasing ςˆ sequence un in X+ such that un → vˆ (cf. [2, Thm.5.2]). For a metrizable σ-Lebesgue lattice (X, ς) its topological completion (X,ˆ ςˆ) is also σ-Lebesgue (cf. [2, Thm.5.36]). Theorem 2.9. Let (X, ς) be a metrizable locally solid lat- tice, (Y, τ) a σ-Lebesgue lattice, and T : (X, ς) → (Y, τ) a positive ςτ-continuous σ-Lebesgue operator. Then the unique extension Tˆ :(X,ˆ ςˆ) → (Y,ˆ τˆ) is positive, ςˆτˆ-conti- nuous, and σ-Lebesgue.

22 Proof. Clearly, Tˆ is positive andς ˆτˆ-continuous. To show that Tˆ is σ-Lebesgue, assumex ˆn ↓ 0 in (X,ˆ ςˆ). Let W ∈ τ(0). We have to show Tˆxˆn ∈ W for large enough n. Take a solid neighborhood W1 ∈ τ(0) with W1 + W1 ⊆ W and choose a neighborhood V ∈ ς(0) with T (V ) ⊆ W1. Let ∞ {Vn}n=1 be a base for ς(0) satisfying Vn+1 + Vn+1 ⊆ Vn for all natural n and also V1 + V1 ⊆ V . At this point we borrow the second paragraph of the proof of [2, Thm.5.36]. For each n, letv ˆn ∈ Xˆ be an upper element of X such thatx ˆn ≤ vˆn andv ˆn − xˆn ∈ V n; see [2, Thm.5.4]. Put n wˆn := ∧i=1vˆn for all n and note that eachw ˆn is an upper element of X,w ˆn ↓ andw ˆn − xˆn ∈ V n. Sincex ˆn ↓ 0, it follows from [2, Thm.2.21(e)] thatw ˆn ↓ 0 also holds in Xˆ. Thus, we can assume without loss in generality thatx ˆn is a sequence of upper elements of X. Take un ∈ X+ with un ≤ xˆn andx ˆn − un ∈ V n. Application of [2, Thm.2.21(e)] to the netsx ˆn ↓ 0 and n ςˆ zn := ∧i=1un ↓ withx ˆn − zn −→ 0 gives zn ↓ 0 in X. As T is τ σ-Lebesgue, T zn −→ 0. So, there exists some n1 such that

Tˆ zn = T zn ∈ W1 ⊆ W 1 (∀n ≥ n1). (6) Observe that n n n 0 ≤ xˆn−zn =x ˆn−∧i=1ui = ∨i=1(ˆxn−ui) ≤ ∨i=1(ˆxi−ui) ≤ n N X(ˆxi − ui) ∈ V 1 + V 2 + ... + V n ⊆ V (∀n ∈ ). (7) i=1

It follows from (6), (7), and Tˆ(V ) ⊆ W 1 that

Tˆ(ˆxn) ≤ Tˆ(zn)+Tˆ(ˆxn −zn) ∈ W 1 +W 1 ⊆ W (∀ n ≥ n1). Since W ∈ τ(0) was arbitrary, Tˆ is σ-Lebesgue. Since Theorem 5.36 of [2] follows from Theorem 2.9 when (X, ς) coincides with (Y, τ) and T is the identity operator on X, Theorem 2.9 can be also considered as a generaliza- tion of [2, Thm.5.36]. One more question deserves being mentioned. Let T be Lebesgue, oτ-continuous, oτ-bounded, oτ-compact, KB,

23 or Levi. Does |T | satisfy the same property? The answer is positive, when T maps a vector lattice X to a Dedekind complete locally solid lattice (Y, τ), in the following cases: (∗) for a Lebesgue/oτ-continuous order continuous op- erator T , when (Y, τ) is Lebesgue, since |T | is order o continuous by [3, Thm.1.56] and hence xα ↓ 0/xα −→ 0 o in X imply |T |xα −→ 0 and (since (Y, τ) is Lebesgue) τ |T |xα −→ 0; (∗∗) foran oτ-bounded operator T , by Proposition 2.1. Remark 2.3. Let (X, ς) and (X, τ) be locally solid lattices. a) Recall that (Y, τ) is said to be boundedly order-boun- ded (BOB) whenever increasing τ-bounded nets in Y+ are order bounded in Y (cf. [29, Def.3.19]). Assume (Y, τ) to be Dedekind complete and BOB, Then each regular quasi KB operator from (X, ς) to (Y, τ) is quasi Levi. Indeed, without lost of generality suppose T ≥ 0. Let xα be an increasing ς-bounded net in X+. Then T xα is a τ-Cauchy increasing net in Y+. Then T xα is τ-bounded and hence o-bounded since (Y, τ) is BOB. The Dedekind completeness of Y implies that T xα ↑ y for some y ∈ Y . Therefore T xα is o-Cauchy as desired. b) Recall that (X, ς) satisfies the B-property (cf. [29, Def.3.1]) whenever the identity operator I on X is quasi σ-KB. Let T : (X, ς) → (Y, τ) be a continu- ous operator and assume the topology ς in (X, ς) to be minimal. Then T is Lebesgue and quasi KB. The first part follows as (X, ς) is Lebesgue by [2, Thm.7.67] and hence T : (X, ς) → (Y, τ) is Lebesgue. The sec- ond part follows as (X, ς) satisfies the B-property [29, Prop.3.2] and hence T is σ-KB, and thus T is quasi KB by Proposition 1.2. Similarly, each regular operator from a vector lattice X to a locally solid lattice (Y, τ) with minimal topology τ is quasi Lebesgue by [29, Cor.3.3].

24 3 Operators between Banach lattices

In this section we investigate operators whose domains and ranges are Banach lattices. It is clear that every con- tinuous operator from an order continuous Banach lattice to a topological vector space is Lebesgue. The following proposition generalizes [8, Prop.2]. Proposition 3.1. An operator T from a σ-order contin- uous Banach lattice X to a metrizable topological vector space (Y, τ) is continuous iff T is σoτ-continuous. Proof. The necessity is trivial. k·kX For the sufficiency: let T be σoτ-continuous, xα −−→ x, and xαβ be a subnet of the net xα. By a) of Remark 1.3, there exists a sequence βn of indices with kxαβn −xkX → 0.

Since X is a Banach lattice, there is a subsequence xαβ nk o of xαβ satisfying xαβ −→ x in X (see [32, Thm.VII.2.1]). n nk τ Since T is σoτ-continuous then T xαβ −→ T x. Remark 1.3 nk τ implies T xα −→ T x, as required. In the next proposition we transfer the σ-condition from the domain X of T to the operator T . Proposition 3.2. Each σoτ-continuous operator T from a Banach lattice X to a Banach lattice Y is continuous.

Proof. It is sufficient to show that kT xnk → 0 for every norm-null sequence xn in X. Let kxnk → 0 in X. For proving kT xnk → 0, in view of b) of Remark 1.3, it is enough to show that, for each subsequence T xnk of T xn, there exists a sequence n of indices with kT x k → 0. ki nki

Let T xnk be a subsequence of T xn. By [32, Thm.VII.2.1], x has a further subsequence x with x −→o 0, and since nk nki nki T is σoτ-continuous, kT x k→ 0 as desired. nki Theorem 3.1. Let T : X → Y be an order bounded op- erator between Banach lattices X and Y , the norms in X′ and Y be order continuous, and Y be weakly sequentially complete. Then T is a quasi KB-operator.

25 Proof. Since Y has order continuous norm, Y is Dedekind complete, and hence the operator T is regular. Therefore, we may assume T ≥ 0. By Proposition 1.2, it suffices to prove that T is quasi σ-KB. Let xn be a k·kX -bounded increasing sequence in X+. Then T xn is k·kY -bounded and 0 ≤ T xn ↑. It follows from [3, Thm.5.28] that T xn

has a weak Cauchy subsequence T xnk . Then T xn is weak w Cauchy and hence T xn −→ y for some y ∈ Y . Since T xn ↑, k·kY it follows from [3, Thm.3.52] that T xn −−→ y, and hence T xn is k·kY -Cauchy as required. Question 3.1. Can the assumption of weakly sequentially completeness of Y be dropped in Theorem 3.1?

Example 3.1. Let X =(cω(R), k·k∞) be the Banach lat- tice of all bounded R-valued functions on R such that each f ∈ X differs from a constant say af on a countable subset of R. It is easy to see that X is Dedekind σ-complete yet not Dedekind complete. We define a positive operator T : X → X as follows. T f is the constant function af · IR ∈ X, for which the set {d ∈ R : f(d) =6 af } is countable. Clearly, T is quasi KB and hence quasi Lebesgue.

(i) For the sequence fn := I{1,2,...n} ∈ X we have that o fn −→ IN ∈ X, yet kfn − INk∞ = 1 for all n. Thus, the norm in X is not σ-order continuous. In particular, Proposition 3.1 is not applicable to T . (ii) T is rank-one continuous (therefore quasi KB, com- o pact, and oτ-compact) operator in X. Let fn −→ 0. Since for every ε> 0 there exists nε such that the set R ∪n≥nε {d ∈ : |fn(d)| ≥ ε} is countable, kT fnk∞ < ε for all n ≥ nε. Thus, T is σoτ-continuous and hence σ-Lebesgue with respect to the norm topology on X. (iii) Operator T is not Lebesgue. Indeed, for the net fα := IR\α ∈ X indexed by the family ∆ of all finite subsets of R ordered by inclusion, we have fα ↓ 0 yet kT fαk∞ = kIRk∞ = 1 for all α ∈ ∆. The same

26 argument shows that T is not weakly Lebesgue. Since T fk =6 0 for at most one k for each disjoint sequence fn in X, operator T is M-weakly compact. It should be clear that T is not L-weakly compact.

(iv) T is KB. Indeed, let X ∋ fα ↑ and kfαk ≤ M ∈ R. I Then afα ↑≤ M. Take f := a · R ∈ X for a = R supα afα ∈ . Clearly, kT (f − fα)k↓ 0, as desired. (v) Clearly, T is a Dunford-Pettis lattice homomorphism. By Corollary 2.3, the span of [0, T ] is a vector sublat- tice of KB-operators in the ordered space Lr(X) of all regular operators in X. (vi) The vector lattice X is τ-laterally σ-complete yet not τ-laterally complete with respect to the norm topol- ogy on X.

Observe that a KB-operator T : ℓ1 → ℓ∞ defined by T a ∞ a k N L M [ ( )]k := Pn=1 n for all ∈ is neither - nor - weakly compact (cf., e.g., [3, p.322]). In the present paper we do not investigate conditions under that KB-operators are (weakly) Lebesgue. The following lemma generalizes the observation that the identity operator in a Banach lattice X is quasi KB iff X is a KB-space.

Lemma 3.1. If c0 does not embed in a Banach space Y , then every continuous operator T from any Banach lattice X to Y is a quasi KB operator. Proof. It follows from [3, Thm.4.63] that T = S ◦Q, where Q is a lattice homomorphism from X to a KB-space Z (and hence Q is a quasi KB-operator by d) of Remark 1.1) and S is a continuous operator from Z to Y . Now apply the fact that the class of quasi KB-operators is closed under the left composition with continuous operators. Theorem 3.2. Let X be a Banach lattice and Y a Banach space. TFAE. (i) Every continuous operator T : X → Y is quasi KB.

(ii) c0 does not embed in Y .

27 Proof. (ii) ⇒ (i) is Lemma 3.1. (i) ⇒ (ii) It suffices to prove that any embedding J : c0 → Y is not a quasi KB-operator. Assume by way of contradiction that such a J is quasi KB and let xn = n e c x Pk=1 k ∈ 0. The increasing sequence n is norm bounded, yet the sequence Jxn is not k·kY -Cauchy because

n 1 Jx Jx J e > n m . k n − mkY = X k ≥ −1 0 ( =6 ) kJ |J(c0)k k=m

The obtained contradiction completes the proof. Among other things, Theorem 3.2 asserts that, in the Banach space setting, the concept of continuous quasi KB-operator does not depend on the domain X and is completely determined by the property that the range space Y not containing an isomorphic copy of c0. In light of this observation, the last sentence in item d) of Remark 1.1 is reduced to the condition 1 ≤ p < ∞ under which k·kp c00 = ℓp is a KB-space and to the condition 1 ≤ l ≤ p under which the identity operator I from (c00, k·kl) to (c00, k·kp) is continuous. Now we pass to a discussion of KB operators. Theorem 3.3. Let T : X → Y be regular operator between ′ Banach lattices. If X has order continuous norm and c0 does not embed in Y then T is KB. Proof. Without lost of generality, we suppose T ≥ 0. By [3, Thm.4.63], T admits a factorization T = S◦Q, where Z is a KB-space and Q : X → Z is a lattice homomorphism. Let xα be an increasing norm bounded net in X+. Since Q ≥ 0 then Qxα is an increasing norm bounded net in Z+, and since Z is a KB-space, there exist z ∈ Z with kQxα −zk→ 0 and hence kT xα −Szk→ 0 as desired. The following theorem gives conditions under which each positive weakly compact operator from a Banach lat- tice X to an arbitrary KB-space Y is σoτ-continuous.

28 Theorem 3.4. For a Banach lattice X TFAE. (i) The image i(X) under the natural embedding i : X → X′′ is a regular sublattice of X′′. (ii) Each positive weakly compact operator T from X to a KB-space Y is σoτ-continuous. (iii) Each positive compact operator T from X to a KB- space Y is σoτ-continuous. (iv) T 2 is σoτ-continuous for every positive weakly com- pact operator T in X. (v) T 2 is σoτ-continuous for every positive compact op- erator T in X. (vi) T 2 is Lebesgue for every positive compact operator T in X. Proof. (i) ⇒ (ii) Let T be a positive weakly compact operator from X to a KB-space Y . By [3, Thm.5.42], T = R ◦ S with both S : X → Z and R : Z → Y are o positive and Z is a reflexive Banach lattice. Let xα −→ 0 ′′ o ′′ in X. Since i(X) is regular in X then i(xα) −→ 0 in X (e.g., by [2, Thm.1.20]). Since S′′ : X′′ → Z′′ is order ′′ o ′′ continuous (e.g., by [3, Thm.1.73]), S (i(xα)) −→ 0 in Z , ′′ o and hence Sxα = S (i(xα)) −→ 0 in Z. The order continuity of norm in Z implies kSxαkZ → 0 and hence kT xαkY = kR ◦ S(xα)kY → 0, showing that T is σoτ-continuous. (ii) ⇒ (iii) and (iv) ⇒ (v) ⇒ (vi) are trivial. (iii) ⇒ (i) If i(X) is not regular in X′′, then (e.g., by [2, Thm.3.12]) there exists a positive functional f ∈ X′ which is not order continuous. Thus f : X → R is positive compact yet not σoτ-continuous as f(xα) does not converge to 0 in the KB-space Y = R for some net o with xα −→ 0 in X. The obtained contradiction proves that i(X) must be regular in X′′. (i) ⇒ (iv) The proof is similar to the proof of (i) ⇒ (ii) above with the only difference that the factorization of T 2 by [3, Cor.5.46] must be used instead of the factorization of T by [3, Thm.5.42].

29 (vi) ⇒ (i) The proof is again similar to the proof of (iii) ⇒ (i) above. If i(X) is not regular in X′′, then there exists a positive functional f ∈ X′ which is not order continuous. Take any z ∈ X+ with f(z) = 1 and define a positive compact operator T in X by T x := f ⊗ z. As f is not order continuous, there is a net xα with xα ↓ 0 and f(xα) ≥ 1 for all α. Therefore

2 T (xα)= T (f(xα)z)= f(xα)T z = f(xα)z ≥ z > 0 (∀α) violating that T 2 is Lebesgue. The obtained contradiction completes the proof. It is clear that positive operators in Theorem 3.4 can be replaced by regular ones. Proposition 3.3. Let T : X → Y be a oτ-bounded oper- ator from a Banach lattice X to a Banach space Y . If c0 does not embed in Y , then T is σ-KB. Proof. By [28, Thm.3.4.6], T factors over a KB-space Z as T = S◦Q, where Q is a lattice homomorphism Q : X → Z, and S : Z → Y is continuous. Let xn be a norm bounded increasing sequence in X+. Since Q is positive, Qxn is positive norm bounded increasing sequence in the KB- space Z. Thus, Qxn is norm convergent in Z. As S is continuous, T xn = SQxn is norm convergent and hence T is σ-KB. Proposition 3.4. Let T : X → Y be a continuous opera- tor from a Banach lattice X to a Banach space Y . If either T ′′ is order-weakly compact or the norm in X is order con- tinuous and T does not preserve a sublattice isomorphic to c0, then T is σ-KB. Proof. The operator T : X → Y factors over a KB-space Z as T = S ◦ Q, where Q : X → Z is a lattice homomor- phism and S : Z → Y continuous by [28, Thm.3.5.8]. Let xn be a norm bounded increasing sequence in X+, then Qxn is also norm bounded increasing in Z+. As Z is a

30 KB-space, Qxn is norm convergent in Z. Hence (S ◦Q)xn is convergent in Y and T is a σ-KB operator. Remark that if T : X → Y is a continuous opera- tor from a Banach lattice X to a Banach space Y and T preserves no subspace (or no sublattice) isomorphic to c0, then T xn is norm convergent for each increasing norm bounded sequence in X+ by [28, Thm.3.4.11] and therefore T is σ-KB. It follows from [33, Thm.1] that every uoτ-continuous operator T from an atomic Banach lattice X to a Banach space Y is a Dunford-Pettis operator. By Theorem 5.57 of [3], a continuous operator T : X → Y from a Banach lattice X to a Banach space Y is o-weakly compact iff, for each order bounded disjoint sequence xn in X, we have kT xnk→ 0. Proposition 3.5. Each regular weakly compact operator T between Banach lattices X and Y is Levi. Proof. Without lost of generality, we can suppose T ≥ 0. Let xα be a norm bounded increasing net in X+. Weak w compactness of T ensures that T xαβ −→ y for some y ∈ Y w and some subnet xαβ of xα. Since T xαβ ↑ then T xαβ −→ y implies T xαβ ↑ y and hence T xα ↑ y which means that T is a Levi operator. Remark 3.1. a) It follows directly from [28, Cor.3.4.12] that every continuous operator from a Banach lattice X to a Banach space Y that does not contain c0 is quasi KB. b) Let X be a Banach lattice and Y be a Banach space. Then every continuous operator T : X → Y that does not preserve c0 is quasi-KB by [28, Thm.3.4.11]. Un- der the same settings, it was observed in [28, 3.4.E4, ′′ p. 203] that if T (BX ) ⊆ Y , where BX is the band generated by X in X′′, then T does not preserve a subspace isomorphic to c0. Note that b-weakly com- ′′ pact operators satisfy T (BX ) ⊆ Y , whenever X has

31 order continuous norm. It follows that under these conditions, each b-weakly compact operator is quasi KB (see [4, Prop 2.11]).

References

[1] Abasov N., Pliev M. Dominated orthogonally additive operators in lattice-normed spaces. Advances in Operator Theory 4, 251- 264 (2019). doi: 10.15352/aot.1804-1354 [2] Aliprantis C.D., Burkinshaw O. Locally Solid Riesz Spaces with Applications to Economics, 2nd edition. American Mathemati- cal Society, Providence, RI (2003). doi: 10.1090/surv/105 [3] Aliprantis C.D., Burkinshaw O. Positive Operators. Springer, Dordrecht (2006). doi: 10.1007/978-1-4020-5008-4 [4] Alpay S., Altın B., Tonyali C. On Property (b) of Vector Lattices. Positivity 7, 135-139 (2003). doi: 10.1023/A:1025840528211 [5] Alpay S., Altin B., Tonyali C. A note on Riesz spaces with property-b. Czech. Math. J. 56, 765-772 (2006). doi: 10.1007/s10587-006-0054-0 [6] Alpay S., Gorokhova S. b-Property of sublattices in vector lat- tices. Turkish. J. Math. (to appear) (2021) [7] Altın B., Machrafi N. Some characterizations of KB-operators on Banach lattices and ordered Banach spaces. Turkish. J. Math. 44, 1736-1743 (2020) doi: 10.3906/mat-2004-106 [8] Aydın A., Emelyanov E., Erkur¸sun-Ozcan¨ N., Marabeh M.A.A. Compact-like operators in lattice-normed spaces. Indag. Math. 29, 633-656 (2018). doi: 10.1016/j.indag.2017.11.002 [9] Aydın A., Emelyanov E., Erkur¸sun-Ozcan¨ N., Marabeh M.A.A. Unbounded p-Convergence in Lattice-Normed Vec- tor Lattices. Sib. Adv. Math. 29, 164-182 (2019). doi: 10.3103/S1055134419030027 [10] Aydın A., Emelyanov E., Gorokhova S. Full lattice conver- gence on Riesz spaces. Indag. Math. 32, 658-690 (2021). doi: 10.1016/j.indag.2021.01.008 [11] Aydın A., Gorokhova S.G., G¨ul H. Nonstandard hulls of lattice- normed ordered vector spaces. Turkish. J. Math. 42, 155-163 (2018). doi: 10.3906/mat-1612-59

32 [12] Bahramnezhad A., Azar K.H. KB-operators on Banach lattices and their relationships with Dunford-Pettis and order weakly compact operators. University Politehnica of Bucharest Scien- tific Bulletin 80(2), 91-98 (2018). [13] Dabboorasad Y.A., Emelyanov E.Y., Marabeh M.A.A. un- Topology in multi-normed vector lattices. Positivity 22, 653-667 (2018). doi: 10.1007/s11117-017-0533-6 [14] Dabboorasad Y.A., Emelyanov E.Y., Marabeh M.A.A. uτ- Convergence in locally solid vector lattices. Positivity 22, 1065- 1080 (2018). doi: 10.1007/s11117-018-0559-4 [15] Dabboorasad Y.A., Emelyanov E.Y., Marabeh M.A.A.: Order convergence is not topological in infinite-dimensional vector lat- tices. Uzb.Mat.Zh. 1, 159-166 (2020). doi: 10.29229/uzmj.2020- 1-15 [16] Emelyanov E.Y. Archimedean Cones in Vector Spaces. Journal of Convex Analysis 24, 169-183 (2017). [17] Emelyanov E.Y., Erkur¸sun-Ozcan¨ N., Gorokhova S.G. Koml´os properties in Banach lattices. Acta Math. Hungar. 155, 324-331 (2018). doi: 10.1007/s10474-018-0852-5 [18] Emelyanov E.Y., Gorokhova S.G., Kutateladze S.S. Unbounded Order Convergence and the Gordon Theorem. Vladikavkaz Mat. Zh. 21(4), 56-62 (2019). doi: 10.23671/VNC.2019.21.44624 [19] Emelyanov E., G¨ul H. Nonstandard hulls of ordered vector spaces. Positivity 20, 413-433 (2016). doi: 10.1007/s11117-015- 0364-2 [20] Emelyanov E.Y., Marabeh M.A.A. Internal characterization of Brezis-Lieb spaces. Positivity 24, 585-592 (2020). doi: 10.1007/s11117-019-00695-z [21] Emelyanov E.Y., Marabeh M.A.A. (2021) On the Brezis- Lieb Lemma and Its Extensions. In: Kusraev A.G., Totieva Z.D. (eds) Operator Theory and Differential Equations. Trends in Mathematics. Birkh¨auser, Cham. doi: 10.1007/978-3-030- 49763-7−3 [22] Erkur¸sun-Ozcan¨ N., Gezer N.A., Zabeti O. Spaces of uτ- Dunford-Pettis and uτ-compact operators on locally solid vec- tor lattices. Matematicki Vesnik 71(4), 351-358 (2019). [23] Erkur¸sun-Ozcan¨ N., Gezer N.A., Ozdemir¨ S.E., Urgancı I.M. Order compact and unbounded order compact operators. Turk- ish. J. Math. 45, 634-646 (2021). doi:10.3906/mat-2004-68

33 [24] Fremlin D.H. Topological Riesz spaces and measure theory. Cambridge University Press, London-New York, 1974. [25] Gao N., Troitsky V., Xanthos F. Uo-convergence and its appli- cations to Ces´aro means in Banach lattices. Isr. J. Math. 220, 649-689 (2017). doi: 10.1007/s11856-017-1530-y

[26] Gorokhova, S.G. Intrinsic characterization of the space c0(A) in the class of Banach lattices. Math Notes 60, 330-333 (1996). doi: 10.1007/BF02320373 [27] Jalili S.A., Azar K.H., Moghimi M.B.F. Order-to-topology con- tinuous operators. Positivity (2021). doi: 10.1007/s11117-021- 00817-6 [28] Meyer-Nieberg P. Banach Lattices. Universitext, Springer- Verlag, Berlin (1991). doi: 10.1007/978-3-642-76724-1 [29] Taylor M.A. Completeness of Unbounded Convergences. Proc. Amer. Math. Soc. 146, 3413-3423 (2018). doi: 10.1090/proc/14007 [30] Taylor M.A. Unbounded topologies and uo-convergence in lo- cally solid vector lattices. J. Math. Anal. Appl. 472, 981-1000 (2019). doi: 10.1016/j.jmaa.2018.11.060 [31] Turan B., Altın B. The relation between b-weakly compact op- erator and KB-operator Turkish. J. Math. 43, 2818-2820 (2019). doi:10.3906/mat-1908-11 [32] Vulikh B.Z. Introduction to the theory of partially ordered spaces. Wolters-Noordhoff Scientific Publications, Ltd., Gronin- gen, (1967). [33] Wickstead A.W. Weak and unbounded order convergence in Ba- nach lattices. J. Austral. Math. Soc. Ser. A 24, 312-319 (1977). doi: 10.1017/S1446788700020346 [34] Zaanen A.C. Riesz Spaces II. North-Holland Publishing, Ams- terdam, The Netherlands, (1983).

34