betaFeedback

WorksMathematics

Uniformization of topological MV-algebras

Abstract

In this paper, First, we prove that for every topological MV-algebra the topology is induced by a uniformity. As a application, we obtain that every topological MV-algebra is completely regular and every T0T_0 first countable topological MV-algebra is metrizable. Second, we study the completion of the canonical difference uniformity in topological MV-algebra. Finally, we construct some examples show that not all topological MV-algebras are canonically completable and the property of canonically completable topological MV-algebra does not imply local order convexity.

Introduction

MV-algebras were introduced by Chang [3] as an algebraic framework for infinite-valued Łukasiewicz logic. They extend Boolean algebras by allowing truth values to vary beyond the classical two-valued setting. The development of the subject revealed a close connection between many-valued logic and ordered algebra. In particular, Mundici’s representation theorem [10] establishes a categorical equivalence between MV-algebras and abelian lattice-ordered groups with a strong unit. Under this equivalence, an MV-algebra is represented as an interval Γ(G,u)=[0,u]\Gamma(G,u)=[0,u] in such a group, with x⊕y=(x+y)∧ux \oplus y=(x+y)\wedge u and ¬x=u−x\neg x=u-x. This correspondence provides a structural foundation for the subject; a systematic account of MV-algebras and their logical and order-theoretic aspects can be found in [4].

A natural question is how this algebraic structure constrains a topology for which the MV-operations are continuous. More generally, topological MV-algebras bring together algebraic operations, lattice order, and convergence, and make it possible to study the extent to which familiar principles of topological algebra persist in a setting governed by truncated addition. Hoo initiated the systematic study of topological MV-algebras in [6]. His work investigated topologies constructed from ideals, including II-adic topologies and their completions, and established basic interactions between algebraic and topological properties. An important separation result from this early theory is that the axioms T0T_{0}, T1T_{1}, and T2T_{2} are equivalent for topological MV-algebras. Subsequent work developed the relationships among neighborhood systems, order, and uniform structures. Weber [13] proved that the topology of a topological MV-algebra is determined by its neighborhoods of 00, and studied the relation between local convexity and uniform MV-algebras.

The study of filter topologies further clarified the role of algebraic compatibility. Luan, Weber, and Yang [8] characterized when a filter topology turns an MV-algebra into a topological MV-algebra and, under those conditions, related it to a uniform MV-algebra. Their results also show why continuity requirements must be stated carefully: a filter topology need not make the binary MV-operation jointly continuous. These investigations provide substantial information about particular constructions and about topologies with additional order properties. They also lead to a broader question concerning an arbitrary topology already known to make the MV-operations continuous: must the underlying topological space admit a compatible uniformity?

There is a strong precedent for a positive answer in the theory of topological groups. The group operations canonically generate left and right uniformities, and the resulting uniform structures explain the complete regularity of topological groups. Moreover, the Birkhoff–Kakutani metrization theorem asserts that a Hausdorff topological group is metrizable precisely when it is first countable [1]. For MV-algebras, however, these conclusions cannot simply be transferred through the algebraic interval representation. The operation ⊕\oplus is truncated addition, and its translations are generally not invertible. In addition, Mundici’s algebraic equivalence does not by itself supply a compatible topological group structure for a given topology on the interval. Uniformizability and metrization therefore require a topological argument that takes account of the MV-structure.

Recent results have brought the separation problem into sharper focus. Gan, Luan, Deng, and Yang [5] proved that every topological MV-algebra is regular. Together with the earlier separation theorem, this yields the equivalence of T0T_{0}, T1T_{1}, T2T_{2}, and T3T_{3} in this class. Regularity, however, does not in general imply complete regularity, so the existence of enough continuous real-valued functions to separate points from closed sets requires a further argument. Li and Yang [7] subsequently proved that every locally convex topological MV-algebra is uniformizable and hence completely regular. Their theorem establishes the desired conclusion under an additional order-theoretic hypothesis and makes it natural to ask whether that hypothesis is needed for the underlying space to be uniformizable.

The purpose of this paper is to establish uniformizability on every topological MV-algebra and to obtain the corresponding separation and metrization theorems. The paper organize as follows: In Section 2, we establish some basic facts. Section 3 prove that for every topological MV-algebra the topology is induced by a uniformity. As a application, we obtain that every topological MV-algebra is completely regular and every T0T_{0} first countable topological MV-algebra is metrizable. In Section 4, we study the completion of the canonical difference uniformity in topological MV-algebra. We mainly prove that a Hausdorff topological MV-algebra AA is canonically completable if and only if the operation ⊕:(A,UA)2→(A,UA)\oplus:(A,\mathcal{U}_{A})^{2}\to(A,\mathcal{U}_{A}) is Cauchy continuous, where the domain carries the product uniformity; when this holds, the topological MV-algebra structure on A^\widehat{A} extending that of AA is unique; its canonical difference uniformity equals the completion uniformity U^A\widehat{\mathcal{U}}_{A}, so A^\widehat{A} is canonically complete (see Theorem 4.5). § 6 constructs some examples show that not all topological MV-algebras are canonically completable and the property of canonically completable topological MV-algebra does not imply local order convexity.

Preliminaries

Definition 2.1. [3, 4] An MV-algebra is an algebra (A,⊕,∗,0)(A,\oplus,{}^{*},0), where (A,⊕,0)(A,\oplus,0) is an abelian monoid and ∗{}^{*} is a unary operation on AA such that for all x,y∈Ax,y\in A,

(MV1) x∗∗=xx^{{}^{**}}=x;

(MV2) x⊕0∗=0∗x\oplus0^{*}=0^{*};

(MV3) (x∗⊕y)∗⊕y=(y∗⊕x)∗⊕x(x^{*}\oplus y)^{*}\oplus y=(y^{*}\oplus x)^{*}\oplus x.

Given an MV-algebra (A,⊕,∗,0)(A,\oplus,{}^{*},0), for each x,y∈Ax,y\in A we define:

1:=0∗,x⊙y:=(x∗⊕y∗)∗,x⊖y:=x⊙y∗.(1)1:=0^{*},\qquad x\odot y:=(x^{*}\oplus y^{*})^{*},\qquad x\ominus y:=x\odot y^{*}. \tag*{(1)}

For any x,y∈Ax,y\in A, we define x≤yx\leq y if and only if x⊖y=0x\ominus y=0. We call this order a natural order on AA.

Proposition 2.2. [4 Proposition 1.1.5] On each MV-algebra AA the natural order determines a lattice structure. Specifically, the join x∨yx\vee y and the meet x∧yx\wedge y of the elements xx and yy are given by

(a) x∨y=(x⊙y∗)⊕y=(x⊖y)⊕yx\vee y=(x\odot y^{*})\oplus y=(x\ominus y)\oplus y;

(b) x∧y=(x∗∨y∗)∗=x⊙(x∗⊕y)x\wedge y=(x^{*}\vee y^{*})^{*}=x\odot(x^{*}\oplus y).

Example 2.3. Let [0,1]={x∈R∣0≤x≤1}[0,1]=\{x\in\mathbb{R}\mid0\leq x\leq1\} and for each x,y∈[0,1]x,y\in[0,1] define by:

x⊕y:=min⁡{1,x+y},x∗:=1−x.x\oplus y:=\min\{1,x+y\},\qquad x^{*}:=1-x.

Then ([0,1],⊕,∗,0)([0,1],\oplus,{}^{*},0) is an MV-algebra, known as the standard MV-algebra.

Lemma 2.4. Let AA be an MV-algebra. For any x,y,z∈Ax,y,z\in A, the following statements hold:

(a) x⊙y≤zx\odot y\leq z if and only if x≤y∗⊕zx\leq y^{*}\oplus z [4 Lemma 1.1.4];

(b) x⊙(y∨z)=(x⊙y)∨(x⊙z)x\odot(y\vee z)=(x\odot y)\vee(x\odot z) [4 Proposition 1.1.6];

(c) if x≤yx\leq y, then y∗≤x∗y^{*}\leq x^{*} and x⊙z≤y⊙zx\odot z\leq y\odot z [4 Lemma 1.1.4];

(d) (x∗⊕y)∨(y∗⊕x)=1(x^{*}\oplus y)\vee(y^{*}\oplus x)=1 [4 Proposition 1.1.7];

Lemma 2.5. Let AA be an MV-algebra. For any x,y,z∈Ax,y,z\in A, the following statements hold:

(a) x∧y=x⊖(x⊖y)x\wedge y=x\ominus(x\ominus y);

(b) (x⊖y)⊖z=x⊖(y⊕z)=(x⊖z)⊖y(x\ominus y)\ominus z=x\ominus(y\oplus z)=(x\ominus z)\ominus y;

(c) (x∧y)⊖z=(x⊖z)∧(y⊖z)(x\wedge y)\ominus z=(x\ominus z)\wedge(y\ominus z);

(d) (x⊖y)⊕(x∧y)=x(x\ominus y)\oplus(x\wedge y)=x;

(e) (x⊖(x⊖y))⊕(y⊖x)=y(x\ominus(x\ominus y))\oplus(y\ominus x)=y.

Proof. For (a). By (1) and Proposition 2.2, we have

x⊖(x⊖y)=x⊖(x⊙y∗)=x⊙(x⊙y∗)∗=x⊙(x∗⊕y)=x∧y.\begin{aligned} x \mathbin{\ominus} (x \mathbin{\ominus} y) &= x \mathbin{\ominus} (x \mathbin{\odot} y^{*}) \\ &= x \mathbin{\odot} (x \mathbin{\odot} y^{*})^{*} \\ &= x \mathbin{\odot} (x^{*} \mathbin{\oplus} y) \\ &= x \mathbin{\wedge} y. \end{aligned}

For (b). By (1) and associativity and commutativity of ⊕\oplus, we have

(x⊖y)⊖z=(x⊙y∗)⊙z∗=(x∗⊕y)∗⊙z∗=((x∗⊕y)⊕z)∗=(x∗⊕(y⊕z))∗=x⊙(y⊕z)=x⊖(y⊕z)\begin{aligned} (x \mathbin{\ominus} y) \mathbin{\ominus} z &= (x \mathbin{\odot} y^{*}) \mathbin{\odot} z^{*} \\ &= (x^{*} \mathbin{\oplus} y)^{*} \mathbin{\odot} z^{*} \\ &= ((x^{*} \mathbin{\oplus} y) \mathbin{\oplus} z)^{*} \\ &= (x^{*} \mathbin{\oplus} (y \mathbin{\oplus} z))^{*} \\ &= x \mathbin{\odot} (y \mathbin{\oplus} z) \\ &= x \mathbin{\ominus} (y \mathbin{\oplus} z) \end{aligned}

and

(x⊖y)⊖z=(x⊙y∗)⊙z∗=(x∗⊕y)∗⊙z∗=((x∗⊕y)⊕z)∗=((x∗⊕z)⊕y)∗=(x⊖z)⊖y.\begin{aligned} (x \mathbin{\ominus} y) \mathbin{\ominus} z &= (x \mathbin{\odot} y^{*}) \mathbin{\odot} z^{*} \\ &= (x^{*} \mathbin{\oplus} y)^{*} \mathbin{\odot} z^{*} \\ &= ((x^{*} \mathbin{\oplus} y) \mathbin{\oplus} z)^{*} \\ &= ((x^{*} \mathbin{\oplus} z) \mathbin{\oplus} y)^{*} \\ &= (x \mathbin{\ominus} z) \mathbin{\ominus} y. \end{aligned}

For (c). Firstly, we shall prove that

(b∧c)⊙a=(b⊙a)∧(c⊙a)(2)(b \mathbin{\wedge} c) \mathbin{\odot} a = (b \mathbin{\odot} a) \mathbin{\wedge} (c \mathbin{\odot} a) \tag*{(2)}

holds for each a,b,c∈Aa,b,c \in A.

Since b∧c≤bb \mathbin{\wedge} c \le b and b∧c≤cb \mathbin{\wedge} c \le c, by Lemma 2.4(c) we have (b∧c)⊙a≤b⊙a(b \mathbin{\wedge} c) \mathbin{\odot} a \le b \mathbin{\odot} a and (b∧c)⊙a≤c⊙a(b \mathbin{\wedge} c) \mathbin{\odot} a \le c \mathbin{\odot} a. Hence, (b∧c)⊙a≤(b⊙a)∧(c⊙a)(b \mathbin{\wedge} c) \mathbin{\odot} a \le(b \mathbin{\odot} a) \mathbin{\wedge} (c \mathbin{\odot} a).

Put w=(b⊙a)∧(c⊙a)w = (b \mathbin{\odot} a) \mathbin{\wedge} (c \mathbin{\odot} a). Then w≤b⊙aw \le b \mathbin{\odot} a. Hence, by Proposition 2.2(b), Lemma 2.4(c) and associativity and commutativity of ⊙\odot,

w⊙(b∗⊕c)≤(b⊙a)⊙(b∗⊕c)=a⊙(b⊙(b∗⊕c))=a⊙(b∧c)=(b∧c)⊙a\begin{aligned} w \mathbin{\odot} (b^{*} \mathbin{\oplus} c) &\le(b \mathbin{\odot} a) \mathbin{\odot} (b^{*} \mathbin{\oplus} c) \\ &= a \mathbin{\odot} (b \mathbin{\odot} (b^{*} \mathbin{\oplus} c)) \\ &= a \mathbin{\odot} (b \mathbin{\wedge} c) \\ &= (b \mathbin{\wedge} c) \mathbin{\odot} a \end{aligned}

and likewise w⊙(c∗⊕b)≤(b∧c)⊙aw \mathbin{\odot} (c^{*} \mathbin{\oplus} b) \le(b \mathbin{\wedge} c) \mathbin{\odot} a by w≤c⊙aw \le c \mathbin{\odot} a. Therefore, by Proposition 2.2(b)(d),

w=w⊙((c∗⊕b)∨((b∗⊕c)))=(w⊙(c∗⊕b))∨(w⊙(b∗⊕c))≤(b∧c)⊙a.\begin{aligned} w &= w \mathbin{\odot} ((c^{*} \mathbin{\oplus} b) \mathbin{\vee} ((b^{*} \mathbin{\oplus} c))) \\ &= (w \mathbin{\odot} (c^{*} \mathbin{\oplus} b)) \mathbin{\vee} (w \mathbin{\odot} (b^{*} \mathbin{\oplus} c)) \\ &\le(b \mathbin{\wedge} c) \mathbin{\odot} a. \end{aligned}

Hence, we have proved that (2): (b∧c)⊙a=(b⊙a)∧(c⊙a)(b \wedge c) \odot a = (b \odot a) \wedge(c \odot a). By (1) and (2), we have that

(x∧y)⊖z=(x∧y)⊙z∗=(x⊙z∗)∧(y⊙z∗)=(x⊖z)∧(y⊖z).\begin{aligned} (x \wedge y) \ominus z &= (x \wedge y) \odot z^{*} \\ &= (x \odot z^{*}) \wedge(y \odot z^{*}) \\ &= (x \ominus z) \wedge(y \ominus z). \end{aligned}

For (d). From Proposition 2.2(a) and Lemma 2.5(a) it follows that

(x⊖y)⊕(x∧y)=(x⊖y)⊕(x⊖(x⊖y))=x∨(x⊖y)=x,\begin{aligned} (x \ominus y) \oplus(x \wedge y) &= (x \ominus y) \oplus(x \ominus(x \ominus y)) \\ &= x \vee(x \ominus y) \\ &= x, \end{aligned}

where the last equality uses x⊖y≤xx \ominus y \le x.

For (e). This follows from Lemma 2.5(a) and (d).

Let AA be an MV-algebra. For each x,y∈Ax,y \in A, define δ:A×A⟶A×A\delta: A \times A \longrightarrow A \times A by

δ(x,y)=(x⊖y,y⊖x)(3)\delta(x,y) = (x \ominus y,y \ominus x) \tag*{(3)}

and for p=(a,b)p=(a,b), q=(c,d)q=(c,d) in A2A^2, define Φ:A2×A2⟶A2\Phi: A^2 \times A^2 \longrightarrow A^2 by

Φ(p,q)=((a⊖d)⊕(c⊖b),(b⊖c)⊕(d⊖a)).(4)\Phi(p,q) = ((a \ominus d) \oplus(c \ominus b),(b \ominus c) \oplus(d \ominus a)). \tag*{(4)}

Lemma 2.6. Let AA be an MV-algebra. δ\delta and Φ\Phi are defined by (3) and (4), respectively. Then

δ(x,z)=Φ(δ(x,y),δ(y,z))(5)\delta(x,z) = \Phi(\delta(x,y),\delta(y,z)) \tag*{(5)}

hold for each x,z,y∈Ax,z,y \in A. Moreover, Φ((0,0),(0,0))=(0,0)\Phi((0,0),(0,0))=(0,0).

Proof. Put p=x⊖zp=x \ominus z, q=y⊖zq=y \ominus z. By Proposition 2.2(a), Lemma 2.5(b), we have

(x⊖y)⊖(z⊖y)=x⊖(y⊕(z⊖y))=x⊖(y∨z)=x⊖(z⊕(y⊖z))=(x⊖z)⊖(y⊖z)=p⊖q\begin{aligned} (x \ominus y) \ominus(z \ominus y) &= x \ominus(y \oplus(z \ominus y)) \\ &= x \ominus(y \vee z) \\ &= x \ominus(z \oplus(y \ominus z)) \\ &= (x \ominus z) \ominus(y \ominus z) = p \ominus q \end{aligned}

and, by Lemma 2.5(a)(c), we have

(y⊖z)⊖(y⊖x)=(y⊖(y⊖x))⊖z=(y∧x)⊖z=(y⊖z)∧(x⊖z)=p∧q.\begin{aligned} (y \ominus z) \ominus(y \ominus x) &= (y \ominus(y \ominus x)) \ominus z \\ &= (y \wedge x) \ominus z \\ &= (y \ominus z) \wedge(x \ominus z) \\ &= p \wedge q. \end{aligned}

Therefore, by Lemma 2.5(d), we have

((x⊖y)⊖(z⊖y))⊕((y⊖z)⊖(y⊖x))=(p⊖q)⊕(p∧q)=p=x⊖z.((x \ominus y) \ominus(z \ominus y)) \oplus((y \ominus z) \ominus(y \ominus x)) = (p \ominus q) \oplus(p \wedge q) = p = x \ominus z.

This proves the first coordinate of (5). Interchanging xx and zz, and using commutativity of ⊕\oplus, proves the second coordinate. The assertion about Φ((0,0),(0,0))=(0,0)\Phi((0,0),(0,0))=(0,0) follows immediately from (4).

Compatible uniformity on every topological MV-algebra

We begin by introducing some notations. Let XX be a set and let AA and BB be subsets of X×XX \times X, i.e. relations on the set XX. The inverse relation of AA will be denoted by −A-A, i.e., −A={(x,y):(y,x)∈A}-A = \{(x,y) : (y,x) \in A\}. The composition of AA and BB will be denoted by A+BA+B, i.e., A+B={(x,z):there is a y∈X such that (x,y)∈A and (y,z)∈B}A+B = \{(x,z) : \text{there is a } y \in X \text{ such that } (x,y) \in A \text{ and } (y,z) \in B\}. 2A2A means that A+AA+A.

The diagonal of the Cartesian product X×XX \times X is the set ΔX={(x,x):x∈X}\Delta_X = \{(x,x) : x \in X\}. Every set V⊆X×XV \subseteq X \times X that contains ΔX\Delta_X and satisfies the condition V=−VV = -V is called an entourage of diagonal. The family of all entourages of diagonal will be denoted by DX\mathcal{D}_X. Let x0x_0 be a point in XX and let V∈DXV \in\mathcal{D}_X. The set V[x0]={x∈X:(x0,x)∈V}V[x_0] = \{x \in X : (x_0,x) \in V\} is called the ball with center x0x_0 and radius VV.

A uniformity [2] on a set XX is a subfamily U⊆DX\mathcal{U} \subseteq\mathcal{D}_X which satisfies the following conditions:

  • (U1)(U1) If V∈UV \in\mathcal{U} and V⊆W∈DXV \subseteq W \in\mathcal{D}_X, then W∈UW \in\mathcal{U}.

  • (U2)(U2) If V1,V2∈UV_1,V_2 \in\mathcal{U}, then V1∩V2∈UV_1 \cap V_2 \in\mathcal{U}.

  • (U3)(U3) For every V∈UV \in\mathcal{U}, there exists a W∈UW \in\mathcal{U} such that 2W⊆V2W \subseteq V.

A uniformity U\mathcal{U} on a set XX is called a separated uniformity if U\mathcal{U} satisfies the condition: (U4)(U4) ⋂U=ΔX\bigcap\mathcal{U} = \Delta_X.

A subfamily B⊆U\mathcal{B} \subseteq\mathcal{U} is called a base for the uniformity U\mathcal{U} if for every V∈UV \in\mathcal{U} there exists a W∈BW \in\mathcal{B} such that W⊆VW \subseteq V.

The following result is well known.

Theorem 3.1. [2 Theorem 8.1.1] For every uniformity U\mathcal{U} on a set XX the family O={U⊆X:for every x∈U there is a V∈U such that V[x]⊆U}\mathcal{O} = \{U \subseteq X : \text{for every } x \in U \text{ there is a } V \in\mathcal{U} \text{ such that } V[x] \subseteq U\} is a topology on the set.

The topology O\mathcal{O} is called the topology induced by the uniformity U\mathcal{U}. A topological space XX with a topology T\mathcal{T} is (pseudo)metrizable if the topology T\mathcal{T} is induced by a (pseudo)metric dd on the set XX.

Definition 3.2. [6] A topological MVMV-algebra is an MVMV-algebra AA equipped with a topology such that

⊕:A×A⟶A,∗:A⟶A\oplus: A \times A \longrightarrow A, \qquad{}^{*} : A \longrightarrow A

are continuous. Then the term operations ⊙,⊖,∨,∧\odot,\ominus,\vee,\wedge are also continuous.

Let N0\mathcal{N}_0 denote the collection of open neighborhoods of zero in a topological MVMV-algebra AA.

Theorem 3.3. Let AA be a topological MVMV-algebra. For U∈N0U \in\mathcal{N}_0, put

EU={(x,y)∈A2:δ(x,y)∈U×U}, where δ is defined by (3).(6)E_U = \{(x,y) \in A^2 : \delta(x,y) \in U \times U\}, \text{ where } \delta\text{ is defined by (3).} \tag*{(6)}

Then {EU:U∈N0}\{E_U : U \in\mathcal{N}_0\} is a base for the uniformity UA={D∈DA:EU⊆D for some U∈N0 such that EU⊆D}\mathcal{U}_A = \{D \in\mathcal{D}_A : E_U \subseteq D \text{ for some } U \in\mathcal{N}_0 \text{ such that } E_U \subseteq D\} inducing the given topology of AA. If AA is T0T_0, then UA\mathcal{U}_A is a separated uniformity.

Proof. Firstly, we show that UA\mathcal{U}_A is a uniformity on AA.

For (U1)(U1). It is trivial.

For (U2)(U2). This follows from the fact EU∩EV=EU∩VE_U \cap E_V = E_{U\cap V} holds for each U,V∈N0U,V \in\mathcal{N}_0.

For (U3)(U3). Let Φ:A2×A2→A2\Phi:A^2 \times A^2 \to A^2 be defined by (4), that is,

Φ((a,b),(c,d))=((a⊖d)⊕(c⊖b),(b⊖c)⊕(d⊖a))\Phi((a,b),(c,d)) = ((a \mathbin{\ominus} d) \mathbin{\oplus} (c \mathbin{\ominus} b),(b \mathbin{\ominus} c) \mathbin{\oplus} (d \mathbin{\ominus} a))

for each ((a,b),(c,d))∈A2×A2((a,b),(c,d)) \in A^2 \times A^2. Since AA is a topological MV-algebra, Φ\Phi is continuous on A2×A2A^2 \times A^2, which is endowed the the product topology. Take any D∈UAD \in\mathcal{U}_A. Then there is U∈N0U \in\mathcal{N}_0 such that EU⊆DE_U \subseteq D. By Lemma 2.6, there is an open neighborhood VV of 00 such that

Φ((V×V)×(V×V))⊆U×U,\Phi((V \times V) \times(V \times V)) \subseteq U \times U,

because Φ((0,0),(0,0))=(0,0)\Phi((0,0),(0,0))=(0,0). If (x,y),(y,z)∈EV(x,y),(y,z) \in E_V, identity (5) now gives δ(x,z)∈U×U\delta(x,z) \in U \times U. Hence 2EV⊆EU2E_V \subseteq E_U, as required. Clearly, {EU:U∈N0}\{E_U:U \in\mathcal{N}_0\} is a base for the uniformity UA\mathcal{U}_A.

Secondly, we shall prove that UA\mathcal{U}_A induce the given topology of AA. In fact, since AA is a topological MV-algebra, for each fix a∈Aa \in A, the left translation ρa:A→A\rho_a:A \to A defined by ρa(x)=a⊖x\rho_a(x)=a \mathbin{\ominus} x for each x∈Ax \in A is continuous. Similarly, the right translation ρa:A→A\rho_a:A \to A defined by λa(x)=x⊖a\lambda_a(x)=x \mathbin{\ominus} a for each x∈Ax \in A is continuous. Thus,

EU[a]={b∈A:a⊖b∈U, b⊖a∈U}=ρa−1(U)∩λa−1(U).E_U[a]=\{b \in A:a \mathbin{\ominus} b \in U,\ b \mathbin{\ominus} a \in U\}=\rho_a^{-1}(U) \cap\lambda_a^{-1}(U).

is an open neighborhood of aa in the original topology for each U∈N0U \in\mathcal{N}_0. This implies that the original topology on AA is finer than the topology induced by UA\mathcal{U}_A.

Conversely, let OO be an original open neighborhood of aa. Define

ϑa:A2→A\vartheta_a:A^2 \to A

by ϑa(x,y)=(a⊖x)⊕y\vartheta_a(x,y)=(a\mathbin{\ominus}x)\mathbin{\oplus}y for each (x,y)∈A2(x,y) \in A^2. Clearly, ϑa\vartheta_a is continuous and ϑa(0,0)=a\vartheta_a(0,0)=a. Choose U∈N0U \in\mathcal{N}_0 with ϑa(U×U)⊆O\vartheta_a(U \times U) \subseteq O. For x∈EU[a]x \in E_U[a], Lemma 2.5(e) yields

x=(a⊖(a⊖x))⊕(x⊖a)=ϑa(a⊖x,x⊖a)∈O.x=(a \mathbin{\ominus} (a \mathbin{\ominus} x)) \mathbin{\oplus} (x \mathbin{\ominus} a)=\vartheta_a(a \mathbin{\ominus} x,x \mathbin{\ominus} a) \in O.

Thus EU[a]⊆OE_U[a] \subseteq O. The sets EU[a]E_U[a] are consequently a neighborhood base for the original topology at every aa. This implies that the topology induced by UA\mathcal{U}_A is finer than the original topology on AA. Thus we have proved that UA\mathcal{U}_A induces the given topology of AA.

If AA is T0T_0, then AA is Hausdorff by [5 Corollary 3.10]. Hence, ⋂U∈N0U={0}\bigcap_{U\in\mathcal{N}_0} U=\{0\}, so membership in every EUE_U forces both directional differences to be zero and hence x=yx=y. This implies that the condition: (U4) ⋂UA=ΔA(U4)\ \bigcap\mathcal{U}_A=\Delta_A holds. □

Definition 3.4

A topological space XX is called

  • (a) a T0T_0 space if for every pair of distinct points x,y∈Xx,y \in X, there exists an open set U∈τU \in\tau such that exactly one of xx and yy belongs to UU;

  • (b) a T1T_1-space if for any distinct points x,y∈Xx,y \in X the point xx has a neighborhood Ux⊆XU_x \subseteq X such that y∉Uxy \notin U_x;

  • (c) a T2T_2-space if for any distinct points x,y∈Xx,y \in X the point xx has a neighborhood Ux⊆XU_x \subseteq X such that y∉Uxy \notin U_x;

  • (d) regular if for any point x∈Xx \in X and a neighborhood Ox⊆XO_x \subseteq X of xx there is a neighborhood Ux⊆XU_x \subseteq X of xx such that Ux‾⊆Ox\overline{U_x} \subseteq O_x;

  • (e) a T3T_3-space if XX is a regular T1T_1-space;

  • (f) completely regular if for every closed set F⊆XF \subseteq X and every point x∈X\Fx \in X \backslash F, there exists a continuous function f:X→[0,1]f : X \to[0,1] such that f(x)=0f(x)=0 and f(y)=1f(y)=1 for all y∈Fy \in F;

  • (g) a Tychonoff space is a completely regular T1T_1 space.

Corollary 3.5. Every topological MV-algebra is completely regular.

Proof. Since every topological space the topology of which is induced by a uniformity is completely regular [2], the statement follows from Theorem 3.3.

Corollary 3.6. Let AA be a topological MV-algebra. Then the following statements are equivalent:

  • (a) AA is a T0T_0 space;

  • (b) AA is a T1T_1 space;

  • (c) AA is a T2T_2 space;

  • (d) AA is a T3T_3 space;

  • (e) AA is a Tychonoff space.

Proof. The equivalences of (a)-(d) follows from [5 Corollary 3.10]. Since every T3T_3 topological space the topology of which is induced by a uniformity is Tychonoff space[2 Corollary 8.1.13], the statement follows from Theorem 3.3.

Lemma 3.7. [2 Theorem 8.1.21] A uniformity U\mathcal{U} on a set XX is induced by a metric on XX if and only if U\mathcal{U} is separated with a countable base.

Corollary 3.8. Every T0T_0, first countable topological MV-algebra is metrizable.

Proof. This follows from Lemma 3.7 and Theorem 3.3.

The completion of the canonical difference uniformity in topological MV-algebra

In this section, we assume that every topological MV-algebra is Hausdorff. In this section, we study the Hausdorff completion of the compatible uniformity on a topological MV-algebra defined in Theorem 3.3. We call this uniformity the canonical difference uniformity on a topological MV-algebra. Denote the canonical difference uniformity on a topological MV-algebra AA by UA\mathcal{U}_A.

Let (X,U)(X,\mathcal{U}) be a uniform space. A family F\mathcal{F} of subsets in XX is Cauchy if, for every entourage E∈UE \in\mathcal{U}, some F∈FF \in\mathcal{F} satisfies F×F⊆EF \times F \subseteq E. A filter (resp. a filter basis) F\mathcal{F} is called a Cauchy filter (resp. a Cauchy filter basis) if F\mathcal{F} is Cauchy [11 0.34 Definition]. A uniform space (X,U)(X,\mathcal{U}) is called complete if every U\mathcal{U}-Cauchy filter on (X,U)(X,\mathcal{U}) converges to some xx in (X,U)(X,\mathcal{U}). Let both (X,UX)(X,\mathcal{U}_X) and (Y,UY)(Y,\mathcal{U}_Y) be uniform spaces. A map f:(X,UX)→(Y,UY)f : (X,\mathcal{U}_X) \to(Y,\mathcal{U}_Y) is called uniformly continuous if for every M∈UYM \in\mathcal{U}_Y there exists N∈UXN \in\mathcal{U}_X such that (f(x),f(y))∈M(f(x),f(y)) \in M for each (x,y)∈N(x,y) \in N, where ff is called a uniform equivalence if ff is bijective and both f:(X,UX)→(Y,UY)f : (X,\mathcal{U}_X) \to(Y,\mathcal{U}_Y) and its inverse f−1:(Y,UY)→(X,UX)f^{-1} : (Y,\mathcal{U}_Y) \to(X,\mathcal{U}_X) are uniformly continuous [11, 0.14 Definition]. A map f:(X,UX)→(Y,UY)f : (X,\mathcal{U}_X) \to(Y,\mathcal{U}_Y) is called a uniform embedding if the induced map (X,UX)→(f(X),UY∣f(X))(X,\mathcal{U}_X) \to(f(X),\mathcal{U}_Y|_{f(X)}), x↦f(x)x \mapsto f(x) is a uniform equivalence [11 0.43 Definition]. A map f:(X,UX)→(Y,UY)f : (X,\mathcal{U}_X) \to(Y,\mathcal{U}_Y) is Cauchy continuous if it sends Cauchy filters to Cauchy filter basises. A completion of a uniform space (X,UX)(X,\mathcal{U}_X) is a pair ((Y,UY),i)((Y,\mathcal{U}_Y),i) consisting of a complete uniform space (Y,UY)(Y,\mathcal{U}_Y) and a uniform embedding i:(X,UX)→(Y,UY)i : (X,\mathcal{U}_X) \to(Y,\mathcal{U}_Y) such that i(X)i(X) is dense in YY [11 0.44 Definition]. For each Hausdorff uniform space (X,UX)(X,\mathcal{U}_X) there is a unique Hausdorff completion ((X^,U^X),i)((\widehat{X},\widehat{\mathcal{U}}_X),i) of (X,UX)(X,\mathcal{U}_X) in the sense of uniform equivalence [11 0.48 Theorem, 0.51 Proposition]. In this section, the term “completion of (X,UX)(X,\mathcal{U}_X)” will then also be applied to the space (X^,U^X)(\widehat{X},\widehat{\mathcal{U}}_X) and, still more abusively, to the set X^\widehat{X}.

Lemma 4.1. Let XX be dense in a Hausdorff uniform space ZZ, let YY be a complete Hausdorff uniform space, and let f:X→Yf : X \to Y be continuous. Then ff has a unique continuous extension f^:Z→Y\widehat{f}: Z \to Y if ff is Cauchy continuous.

Proof. Assume ff is Cauchy continuous. For p∈Zp \in Z, the traces on XX of neighborhoods of pp generate a Cauchy filter Fp\mathcal{F}_p in XX. The image f(Fp)f(\mathcal{F}_p) is a Cauchy filter basis, and therefore has a unique limit in YY, because YY is a complete Hausdorff uniform space. Define

f^(p)=lim⁡f(Fp).\widehat{f}(p) = \lim f(\mathcal{F}_p).

For p∈Xp \in X, continuity of ff shows that this is f(p)f(p).

To check the continuity of f^\widehat{f}, let WW be an open neighborhood of f^(p)\widehat{f}(p). Choose a closed neighborhood CC of f^(p)\widehat{f}(p) contained in WW, using the regularity of YY. There is an open neighborhood OO of pp in ZZ with f(O∩X)⊆Cf(O \cap X) \subseteq C. For every q∈Oq \in O, the filter basis f(Fq)f(\mathcal{F}_q) is eventually in CC, so its limit belongs to CC. Hence f^(O)⊆C⊆W\widehat{f}(O) \subseteq C \subseteq W. Uniqueness follows because XX is dense and YY is Hausdorff. □

Proposition 4.2. A continuous MV-homomorphism f:A→Bf : A \to B is uniformly continuous from (A,UA)(A,\mathcal{U}_A) to (B,UB)(B,\mathcal{U}_B). If AA is an MV-subalgebra of BB with the subspace topology, then UA=UB∣A×A\mathcal{U}_A = \mathcal{U}_B|_{A \times A}.

Proof. Given a zero-neighborhood UU in BB, choose a zero-neighborhood VV in AA with f(V)⊆Uf(V) \subseteq U. Since ff preserves directional differences, (f×f)(EV)⊆EU(f \times f)(E_V) \subseteq E_U. For a subalgebra, the identities EUB∩A2=EU∩AAE_U^B \cap A^2 = E_{U \cap A}^A and the description of subspace zero-neighborhoods prove the second assertion. □

Lemma 4.3. Suppose two uniformities on a set YY induce the same topology and have the same restriction to a dense subset XX. Then the two uniformities are equal.

Proof. Let EE be a closed entourage for the first uniformity. Equality of the restrictions gives an entourage FF of the second uniformity with F∩X2⊆E∩X2F \cap X^2 \subseteq E \cap X^2. Choose a product-topologically open entourage G⊆FG \subseteq F for the second uniformity. Such entourages form a base, and the product topology is the same for both uniformities. Density of X2X^2 gives

G⊆G∩X2‾⊆E.G \subseteq\overline{G \cap X^2} \subseteq E.

Thus EE is also an entourage of the second uniformity. Closed entourages form a base for every uniformity, so the first uniformity is contained in the second. Interchanging them proves the reverse containment. □

Definition 4.4. A Hausdorff topological MV-algebra AA is canonically complete if its canonical difference uniformity is complete. It is canonically completable if the completion A^\widehat{A} of the uniform space (A,UA)(A,\mathcal{U}_A) is a topological MV-algebra containing AA as an MV-subalgebra.

Here A^\widehat{A} carries its uniform completion topology, and the containment is understood through the canonical uniform embedding jA:A→A^j_A:A\to\widehat{A}: the operations on A^\widehat{A} must restrict to those of AA.

From now on a topological MV-algebra AA is Hausdorff and A^\widehat{A} means the completion of (A,UA)(A,\mathcal{U}_A), with completion uniformity U^A\widehat{\mathcal{U}}_A.

Theorem 4.5. A Hausdorff topological MV-algebra AA is canonically completable if and only if the operation ⊕:(A,UA)2→(A,UA)\oplus:(A,\mathcal{U}_A)^2\to(A,\mathcal{U}_A) is Cauchy continuous, where the domain carries the product uniformity. When this holds, the topological MV-algebra structure on A^\widehat{A} extending that of AA is unique. Its canonical difference uniformity equals the completion uniformity U^A\widehat{\mathcal{U}}_A, so A^\widehat{A} is canonically complete.

Proof. Identify AA with its image under jAj_A, and write U^A\widehat{\mathcal{U}}_A for the uniformity of its Hausdorff completion. Thus AA is a dense uniform subspace of A^\widehat{A}.

Necessity. Suppose that AA is canonically completable, and denote the continuous addition on its completion by ⊕^\widehat{\oplus}. Let F\mathcal{F} be a Cauchy filter on (A,UA)2(A,\mathcal{U}_A)^2. The product (A^,U^A)2(\widehat{A},\widehat{\mathcal{U}}_A)^2 is complete. The image of F\mathcal{F} under the uniform embedding

jA×jA:A×A↪A^×A^j_A\times j_A:A\times A\hookrightarrow\widehat{A}\times\widehat{A}

therefore converges to some (p,q)∈A^×A^(p,q)\in\widehat{A}\times\widehat{A}. Joint continuity gives

⊕^(F)⟶p⊕^qin A^.\widehat{\oplus}(\mathcal{F})\longrightarrow p\widehat{\oplus}q\quad\text{in }\widehat{A}.

Since every convergent family on a uniform space is Cauchy [11 Proposition 0.35], ⊕^(F)\widehat{\oplus}(\mathcal{F}) is a Cauchy filter basis in A^\widehat{A}. Because ⊕^\widehat{\oplus} restricts to ⊕\oplus and U^A∣A×A=UA\widehat{\mathcal{U}}_A|_{A\times A}=\mathcal{U}_A, it follows that ⊕^(F)=⊕(F)\widehat{\oplus}(\mathcal{F})=\oplus(\mathcal{F}) is a Cauchy filter basis in (A,UA)(A,\mathcal{U}_A). This proves Cauchy continuity.

Sufficiency. Assume that ⊕\oplus is Cauchy continuous. The identities

x∗⊕y∗=y⊕x,y∗⊕x∗=x⊕yx^*\oplus y^*=y\oplus x,\qquad y^*\oplus x^*=x\oplus y

give, for every zero-neighborhood UU,

(x,y)∈EU⟺(x∗,y∗)∈EU.(x,y)\in E_U\Longleftrightarrow(x^*,y^*)\in E_U.

Thus the involution is a uniform isomorphism of (A,UA)(A,\mathcal{U}_A) onto itself. The extension theorem for uniformly continuous maps into complete Hausdorff uniform spaces gives a unique uniformly continuous map ∗^:A^→A^\widehat{*}:\widehat{A}\to\widehat{A} with ∗^∣A=∗\widehat{*}|_A= {*}. Both ∗^∘∗^\widehat{*}\circ\widehat{*} and the identity map are continuous and agree on the dense subset AA; hence ∗^∘∗^=id⁡A^\widehat{*}\circ\widehat{*}=\operatorname{id}_{\widehat{A}}.

Assume ⊕:(A,UA)2→(A,UA)\oplus:(A,\mathcal{U}_A)^2\to(A,\mathcal{U}_A) is Cauchy continuous. Then so is ⊕:(A,UA)2→(A^,U^A)\oplus:(A,\mathcal{U}_A)^2\to(\widehat{A},\widehat{\mathcal{U}}_A), because U^A∣A×A=UA\widehat{\mathcal{U}}_A|_{A\times A}=\mathcal{U}_A. Since A×AA\times A is dense in the uniform space A^×A^\widehat{A}\times\widehat{A} and A^\widehat{A} is a complete Hausdorff uniform space, applying Lemma 4.1 to ⊕:A×A→A^\oplus:A\times A\to\widehat{A} there is a unique continuous extension ⊕^:A^×A^→A^\widehat{\oplus}:\widehat{A}\times\widehat{A}\to\widehat{A} of ⊕\oplus.

Put 0^=jA(0)\widehat{0}=j_A(0). Every term formed from ⊕^\widehat{\oplus}, ∗^\widehat{*}, 0^\widehat{0} defines a continuous map on the appropriate finite power of A^\widehat{A}. For an MV-identity in nn variables, its two term maps

agree on AnA^n, because the operations restrict to those of AA. The set on which two continuous maps into a Hausdorff space agree is closed. Since AnA^n is dense in A^n\widehat{A}^n, the identity holds throughout A^n\widehat{A}^n. In particular, for p,q,r∈A^p,q,r \in\widehat{A}, we obtain

p⊕^q=q⊕^p,(p⊕^q)⊕^r=p⊕^(q⊕^r),p⊕^0^=p,p∗^∗^=p,p⊕^0^∗=0^∗,(p∗⊕^q)∗⊕^q=(q∗⊕^p)∗⊕^p.\begin{aligned} p \widehat{\oplus} q &= q \widehat{\oplus} p,\\ (p \widehat{\oplus} q) \widehat{\oplus} r &= p \widehat{\oplus} (q \widehat{\oplus} r),\\ p \widehat{\oplus} \widehat{0} &= p,\qquad p^{\widehat{*}\widehat{*}}=p,\\ p \widehat{\oplus} \widehat{0}^{*} &= \widehat{0}^{*},\\ (p^{*} \widehat{\oplus} q)^{*} \widehat{\oplus} q &= (q^{*} \widehat{\oplus} p)^{*} \widehat{\oplus} p. \end{aligned}

These are all the defining MV-axioms. Consequently (A^,⊕^,∗^,0^)(\widehat{A},\widehat{\oplus},\widehat{*},\widehat{0}) is a Hausdorff topological MV-algebra, and jAj_A is an MV-embedding. Definition 4.4 now shows that AA is canonically completable.

Two continuous extensions of addition agree on the dense set A2A^2, and two continuous extensions of the involution agree on the dense set AA. Since A^\widehat{A} is Hausdorff, the respective extensions agree everywhere; the zero is forced to be jA(0)j_A(0). This proves uniqueness.

Let V\mathcal{V} be the canonical difference uniformity of the resulting topological MV-algebra A^\widehat{A}. By Theorem 3.3, it induces the completion topology. Since AA is a topological MV-subalgebra, Proposition 4.2 gives

V∣A2=UA=U^A∣A2.\mathcal{V}|_{A^2}=\mathcal{U}_A=\widehat{\mathcal{U}}_A|_{A^2}.

The two uniformities therefore induce the same topology and have the same restriction to the dense subset AA. Lemma 4.3 yields V=U^A\mathcal{V}=\widehat{\mathcal{U}}_A. As the latter is complete, A^\widehat{A} is canonically complete.

Corollary 4.6. Let AA be a canonically completable topological MV-algebra. Every continuous MV-homomorphism f:A⟶Bf:A\longrightarrow B into a canonically complete Hausdorff topological MV-algebra extends uniquely to a continuous MV-homomorphism f^:A^⟶B\widehat{f}:\widehat{A}\longrightarrow B. Moreover, AA is canonically completable if and only if it admits a topological MV-embedding into some canonically complete Hausdorff topological MV-algebra.

Proof. By Theorem 4.5, A^\widehat{A} is a canonically complete topological MV-algebra extending AA. By Proposition 4.2, ff is uniformly continuous, so it extends uniquely to a uniformly continuous map f^:A^⟶B\widehat{f}:\widehat{A}\longrightarrow B by [2 Theorem 8.3.10]. Preservation of the operations follows by continuity and density.

One direction of the embedding assertion follows by taking A^\widehat{A}. Conversely, identify AA with a topological MV-subalgebra of a canonically complete algebra BB. Its closure CC in BB is a subalgebra [6 Theorem 3.5]. It is a closed uniform subspace of a complete space [11 Proposition 0.39], hence complete. By Proposition 4.2, its canonical uniformity is the subspace uniformity from BB, and AA is uniformly dense in CC. Thus CC realizes the uniform completion of AA and supplies the required continuous operations.

According to [9], a subset VV of a partially ordered set is order convex if a,b∈Va,b\in V and a≤x≤ba\leq x\leq b imply x∈Vx\in V. A topology is locally order convex if every point has a neighborhood base consisting of order convex sets. Let AA be a MV-algebra and $V is a subset of AA. VV is called a downward closed set if z≤vz \leq v for some v∈Vv \in V, then z∈Vz \in V.

Theorem 4.7. For a Hausdorff topological MV-algebra AA, the following are equivalent:

(a) AA is locally order convex;

(b) zero has a neighborhood base of downward closed sets;

(c) ⊕:(A,UA)2⟶(A,UA)\oplus: (A,\mathcal{U}_A)^2 \longrightarrow(A,\mathcal{U}_A) is uniformly continuous, where the domain carries the product uniformity;

(d) the MV-operations extend to uniformly continuous operations on (A^,U^A)(\widehat{A},\widehat{\mathcal{U}}_A).

When this holds, AA is canonically completable and its completion A^\widehat{A} is again locally order convex.

Proof. An order convex set containing zero is downward closed, so (a) implies (b). Assume (b). For each a∈Aa \in A given a neighborhood OO of aa, choose an open zero-neighborhood UU with EU[a]⊆OE_U[a] \subseteq O, using Theorem 3.3. Choose a downward closed zero-neighborhood V⊆UV \subseteq U. Then EV[a]E_V[a] is a neighborhood of aa and EV[a]⊆EU[a]⊆OE_V[a] \subseteq E_U[a] \subseteq O. If b,c∈EV[a]b,c \in E_V[a] and b≤z≤cb \leq z \leq c, using commutativity of ⊙\odot, the definition of ⊖\ominus by (1) and Lemma 2.4(c) gives

a⊖z≤a⊖b∈V,z⊖a≤c⊖a∈V.a \ominus z \leq a \ominus b \in V,\qquad z \ominus a \leq c \ominus a \in V.

Thus z∈EV[a]z \in E_V[a] and EV[a]E_V[a] is order convex neighborhood, proving (a).

(b) ⇒\Rightarrow (c). Take any a,b,c,d∈Aa,b,c,d \in A. First note the MV-inequality

(a⊕c)⊖(b⊕d)≤(a⊖b)⊕(c⊖d).(7)(a \oplus c) \ominus(b \oplus d) \leq(a \ominus b) \oplus(c \ominus d). \tag*{(7)}

Indeed, a≤b⊕(a⊖b)a \leq b \oplus(a \ominus b) and c≤d⊕(c⊖d)c \leq d \oplus(c \ominus d). Adding and applying Lemma 2.4(a) proves the inequality.

Given a zero-neighborhood UU, choose a downward closed zero-neighborhood W⊆UW \subseteq U. Continuity at (0,0)(0,0) provides an open zero-neighborhood VV with V⊕V⊆WV \oplus V \subseteq W. If (a,b),(c,d)∈EV(a,b),(c,d) \in E_V, then a⊖b,b⊖a∈Va \ominus b,b \ominus a \in V and c⊖d,d⊖c∈Vc \ominus d,d \ominus c \in V. Hence,

(a⊖b)⊕(c⊖d)∈V⊕V⊆W(a \ominus b) \oplus(c \ominus d) \in V \oplus V \subseteq W

Thus, by (7) and WW being downward closed,

(a⊕c)⊖(b⊕d)∈W⊆U(a \oplus c) \ominus(b \oplus d) \in W \subseteq U

Similarly, we have (b⊕d)⊖(a⊕c)∈U(b \oplus d) \ominus(a \oplus c) \in U. This implies that ((a⊕c),(b⊕d))∈EU((a \oplus c),(b \oplus d)) \in E_U. This is uniform continuity.

(c) ⇒\Rightarrow (b). The identities

x∗⊖y∗=y⊖x,y∗⊖x∗=x⊖yx^* \ominus y^* = y \ominus x,\qquad y^* \ominus x^* = x \ominus y

give, for every zero-neighborhood UU,

(x,y)∈EU⟺(x∗,y∗)∈EU.(x,y) \in E_U \quad\Longleftrightarrow\quad(x^*,y^*) \in E_U.

Thus, the involution is always uniformly continuous, so every MV-term, in particular ∧\wedge, is uniformly continuous. Given a zero-neighborhood UU, choose an open zero-neighborhood VV such that

(x,x′),(z,z′)∈EV⟹(x∧z,x′∧z′)∈EU.(x,x'),(z,z')\in E_V \mathrel{\Longrightarrow} (x\wedge z,x'\wedge z')\in E_U.

For x∈Vx\in V, the pair (x,0)(x,0) lies in EVE_V; and (z,z)∈EV(z,z)\in E_V for every zz. Therefore x∧z∈Ux\wedge z\in U, which implies that ↓V⊆U\downarrow V\subseteq U, where ↓V={y∈A:y≤x for some x∈V}\downarrow V=\{y\in A:y\leq x\text{ for some }x\in V\}. Clearly, V⊆↓VV\subseteq\downarrow V. Consequently, the downward closed set ↓V\downarrow V is a zero-neighborhood. This proves (b).

Condition (c), together with uniform continuity of the involution, gives uniformly continuous extensions of the basic operations to the completions by [2 Theorem 8.3.10]. The MV-identities pass to the completion as in Theorem 4.5; hence (d). Conversely, restricting a uniformly continuous extension gives (c). Finally, Theorem 4.5 identifies the completion uniformity with the completed algebra’s canonical uniformity. Applying (c) implies (a) to that algebra proves the last assertion.

The relationship in Theorem 4.7 is consistent with the established theory of lattice and MV-uniformities [7, 12, 13]. It is stronger than the continuous-extension criterion of Theorem 4.5. § 6 demonstrates that the two criteria are genuinely different even for a countable Boolean algebra.

Let (X,U)(X,\mathcal U) be a uniform space. (X,U)(X,\mathcal U) is called totally bounded if for each U∈UU\in\mathcal U, there is a finite set F⊆XF\subseteq X such that X=⋃x∈FU[x]X=\bigcup_{x\in F}U[x].

Corollary 4.8. Let AA be a Hausdorff topological MV-algebra such that the uniform space (A,UA)(A,\mathcal U_A) is totally bounded. Then AA is canonically completable if and only if AA is locally order convex. In this case its canonical completion is a compact Hausdorff topological MV-algebra.

Proof. Assume that AA is locally order convex. Then it follows that AA is canonically completable from Theorem 4.7.

Conversely, Since the completion of a totally bounded Hausdorff uniform space is compact [2 Corollary 8.3.17], A^\widehat{A} is compact topological MV-algebra. Hence, the continuous extension ⊕^:(A^,U^A)×(A^,U^A)⟶(A^,U^A)\widehat{\oplus}:(\widehat{A},\widehat{\mathcal U}_A)\times(\widehat{A},\widehat{\mathcal U}_A)\longrightarrow(\widehat{A},\widehat{\mathcal U}_A) of ⊕\oplus is uniformly continuous by [11 Corollary 0.54], because its domain A^×A^\widehat{A}\times\widehat{A} is compact. Its restriction ⊕\oplus to (A,UA)×(A,UA)(A,\mathcal U_A)\times(A,\mathcal U_A) is therefore uniformly continuous, and Theorem 4.7 gives local order convexity.

:::

Some examples

In this section, we give two examples which show that not all topological MV-algebras are canonically completable and the property of canonically completable topological MV-algebra does not imply local order convexity.

The following construction shows that first countability, second countability, and linearity of the MV-order do not suffice for continuous extendability.

Example 5.1. There exists a countable Hausdorff second-countable topological MV-chain AA with canonical completion

A^=[0,1]×Zdiscrete,\widehat{A}=[0,1]\times\mathbb{Z}_{\mathrm{discrete}},

such that even one fixed truncated translation x↦a⊕xx \mapsto a \oplus x has no continuous extension to A^\widehat{A}. In particular, AA is not canonically completable.

Proof. Fix an irrational real number α\alpha, and put

G=Q+Zα,A=G∩[0,1].G = \mathbb{Q} + \mathbb{Z}\alpha,\qquad A = G \cap[0,1].

Every x∈Gx \in G has a unique expression x=q+nαx = q + n\alpha, where q∈Qq \in\mathbb{Q} and n∈Zn \in\mathbb{Z}. Define the additive map κ:G→Z\kappa: G \to\mathbb{Z} by κ(q+nα)=n\kappa(q + n\alpha) = n. The interval AA is an MV-subalgebra of the standard algebra, with

x⊕y=min⁡{x+y,1},x∗=1−x.x \oplus y = \min\{x+y,1\},\qquad x^* = 1-x.

Its natural order is the usual order, so it is an MV-chain. Define a metric on AA by, for each x,y∈Ax,y \in A

d(x,y)={∣x−y∣,κ(x)=κ(y)∣x−y∣+1,κ(x)≠κ(y).(8)d(x,y) = \begin{cases} |x-y|, & \kappa(x) = \kappa(y)\\ |x-y|+1, & \kappa(x) \ne\kappa(y). \end{cases} \tag*{(8)}

It is a metric because it is the sum of the usual metric and the pullback of the discrete metric on Z\mathbb{Z}. For each n∈Zn \in\mathbb{Z}, the set

An=(nα+Q)∩[0,1]A_n = (n\alpha+\mathbb{Q}) \cap[0,1]

is an open and closed subspace with its usual real subspace topology. Since AA is countable and metrizable, it is Hausdorff and second countable. We verify continuity of the MV-operations. Because 1∈Q1 \in\mathbb{Q}, κ(1−x)=−κ(x)\kappa(1-x) = -\kappa(x). One can easily show that the involution ∗* is an isometry for (8), that is, d(x,y)=d(x∗,y∗)d(x,y) = d(x^*,y^*). Hence, ∗* is continuous. Next, we shall show that ⊕\oplus is continuous. Fix a,b∈Aa,b \in A and ϵ>0\epsilon> 0.

Case 1: a+b<1a+b < 1. Put δ=min⁡{1−(a+b)3,ϵ2}\delta= \min\left\{\frac{1-(a+b)}{3},\frac{\epsilon}{2}\right\}. We shall prove that d(x⊕y,a⊕b)<ϵd(x\oplus y,a\oplus b) < \epsilon whenever d(x,a)<δd(x,a) < \delta and d(y,b)<δd(y,b) < \delta. Since δ<1\delta< 1, we have κ(x)=κ(a)\kappa(x) = \kappa(a) and κ(y)=κ(b)\kappa(y) = \kappa(b). Hence, x+y≤a+b+∣x−a∣+∣y−b∣≤a+b+2δ<1x+y \le a+b+|x-a|+|y-b| \le a+b+2\delta< 1. Furthermore,

κ(x⊕y)=κ(x+y)=κ(x)+κ(y)=κ(a)+κ(b)=κ(a+b)=κ(a⊕b).\begin{aligned} \kappa(x\oplus y) &= \kappa(x+y)\\ &= \kappa(x)+\kappa(y)\\ &= \kappa(a)+\kappa(b)\\ &= \kappa(a+b)\\ &= \kappa(a\oplus b). \end{aligned}

By (8), we have

d(x⊕y,a⊕b)=∣(x⊕y)−(a⊕b)∣=∣(x+y)−(a⊕b)∣≤∣x−a∣+∣y−b∣<2δ≤ϵ\begin{aligned} d(x\oplus y,a\oplus b) &= |(x\oplus y)-(a\oplus b)|\\ &= |(x+y)-(a\oplus b)|\\ &\le|x-a|+|y-b|\\ &< 2\delta\le\epsilon \end{aligned}

Case 2: a+b>1a+b>1. Put δ=min⁡{a+b−13,12}\delta=\min\left\{\frac{a+b-1}{3},\frac{1}{2}\right\}. Then

x+y≥a+b−∣x−a∣−∣y−b∣≥a+b−2δ>1\begin{aligned} x+y &\ge a+b-|x-a|-|y-b|\\ &\ge a+b-2\delta>1 \end{aligned}

whenever ∣x−a∣<δ|x-a|<\delta and ∣y−b∣<δ|y-b|<\delta. Hence,

d(x⊕y,a⊕b)=d(1,1)=0<ϵd(x\mathbin{\oplus} y,a\mathbin{\oplus} b)=d(1,1)=0<\epsilon

whenever d(x,a)<δd(x,a)<\delta and d(y,b)<δd(y,b)<\delta.

Case 3: a+b=1a+b=1. Then κ(a⊕b)=κ(1)=0\kappa(a\mathbin{\oplus} b)=\kappa(1)=0. Put δ=min⁡{13,ϵ2}\delta=\min\left\{\frac{1}{3},\frac{\epsilon}{2}\right\}. Let d(x,a)<δd(x,a)<\delta and d(y,b)<δd(y,b)<\delta. Then κ(x)=κ(a)\kappa(x)=\kappa(a) and κ(y)=κ(b)\kappa(y)=\kappa(b). Hence, if x+y≤1x+y\le1, then

κ(x⊕y)=κ(x+y)=κ(x)+κ(y)=κ(a)+κ(b)=κ(a+b).\begin{aligned} \kappa(x\mathbin{\oplus} y)&=\kappa(x+y)\\ &=\kappa(x)+\kappa(y)\\ &=\kappa(a)+\kappa(b)\\ &=\kappa(a+b). \end{aligned}

Thus, by (8), we have

d(x⊕y,a⊕b)=∣x+y−(a+b)∣≤∣x−a∣+∣y−b∣≤2δ<ϵ.d(x\mathbin{\oplus} y,a\mathbin{\oplus} b)=|x+y-(a+b)|\le|x-a|+|y-b|\le2\delta<\epsilon.

If x+y>1x+y>1, then x⊕y=1x\mathbin{\oplus} y=1. Hence, κ(x⊕y)=κ(1)=κ(a⊕b)\kappa(x\mathbin{\oplus} y)=\kappa(1)=\kappa(a\mathbin{\oplus} b). Thus,

d(x⊕y,a⊕b)=d(1,1)=0<ϵ.d(x\mathbin{\oplus} y,a\mathbin{\oplus} b)=d(1,1)=0<\epsilon.

This proves continuity of ⊕\mathbin{\oplus}. Therefore AA with the topology induced by dd defined in (8) is a topological MV-algebra.

A zero-neighborhood base is {Uϵ:0<ϵ<1}\{U_\epsilon:0<\epsilon<1\}, where

Uϵ={x∈A:d(0,x)<ϵ}={x∈A0:d(0,x)<ϵ}=Q∩[0,ϵ).\begin{aligned} U_\epsilon&=\{x\in A:d(0,x)<\epsilon\}\\ &=\{x\in A_0:d(0,x)<\epsilon\}\\ &=\mathbb{Q}\cap[0,\epsilon). \end{aligned}

Put

EUϵ={(x,y)∈A×A:x⊖y∈Uϵ and y⊖x∈Uϵ}E_{U_\epsilon}=\{(x,y)\in A\times A:x\mathbin{\ominus} y\in U_\epsilon\text{ and }y\mathbin{\ominus} x\in U_\epsilon\}

and

Edϵ={(x,y)∈A×A:d(x,y)<ϵ}.E_d^\epsilon=\{(x,y)\in A\times A:d(x,y)<\epsilon\}.

If (x,y)∈Edϵ(x,y)\in E_d^\epsilon, then d(x,y)<ϵ<1d(x,y)<\epsilon<1. Thus, we have κ(x)=κ(y)\kappa(x)=\kappa(y). This implies that x−y∈Qx-y\in\mathbb{Q} and ∣x−y∣=d(x,y)<ϵ|x-y|=d(x,y)<\epsilon. Since the standard-algebra difference is x⊖y=max⁡{x−y,0}x\mathbin{\ominus} y=\max\{x-y,0\}, we have x⊖y∈Uϵx\mathbin{\ominus} y\in U_\epsilon and y⊖x∈Uϵy\mathbin{\ominus} x\in U_\epsilon, proving Edϵ⊆EUϵE_d^\epsilon\subseteq E_{U_\epsilon}.

If (x,y)∈EUϵ(x,y)\in E_{U_\epsilon}, then x⊖y∈Uϵx\mathbin{\ominus} y\in U_\epsilon and y⊖x∈Uϵy\mathbin{\ominus} x\in U_\epsilon. Hence κ(x⊖y)=κ(y⊖x)=0\kappa(x\mathbin{\ominus} y)=\kappa(y\mathbin{\ominus} x)=0 and ∣x−y∣<ϵ|x-y|<\epsilon. Therefore we have κ(x)=κ(y)\kappa(x)=\kappa(y). Then d(x,y)=∣x−y∣<ϵd(x,y)=|x-y|<\epsilon, which implies that (x,y)∈Edϵ(x,y)\in E_d^\epsilon, proving Edϵ⊇EUϵE_d^\epsilon\supseteq E_{U_\epsilon}. Thus the canonical difference uniformity UAU_A on AA is induced by the metric (8).

Give [0,1]×Z[0,1]\times\mathbb{Z} the complete metric

ρ((s,n),(t,m))={∣s−t∣,n=m∣s−t∣+1,n≠m.(9)\rho((s,n),(t,m))= \begin{cases} |s-t|, & n=m\\ |s-t|+1, & n\ne m. \end{cases} \tag*{(9)}

Completeness follows because a Cauchy sequence eventually is contained in [0,1]×{n}[0,1]\times\{n\} for some n∈Zn\in\mathbb{Z} and [0,1]×{n}[0,1]\times\{n\} is complete. The map j:A→[0,1]×Zj:A\to[0,1]\times\mathbb{Z} defined by

j(x)=(x,κ(x))(10)j(x)=(x,\kappa(x)) \tag*{(10)}

is isometric. Since each nα+Qn\alpha+\mathbb{Q} is dense in R\mathbb{R}, An=(nα+Q)∩[0,1]A_n=(n\alpha+\mathbb{Q})\cap[0,1] is dense in [0,1][0,1] for each n∈Zn\in\mathbb{Z}. Hence j(An)j(A_n) is dense in the branch [0,1]×{n}[0,1]\times\{n\} for each n∈Zn\in\mathbb{Z}. This implies that j(A)j(A) is dense in [0,1]×Z[0,1]\times\mathbb{Z}. This proves that the uniform space [0,1]×Z[0,1]\times\mathbb{Z} with the uniformity induced by the metric ρ\rho defined by (9) is the canonical completion of the uniform space (A,UA)(A,\mathcal{U}_A).

Now take a=1/2∈A0a=1/2\in A_0. We shall show that the left translation λa:A→A\lambda_a:A\to A defined by λa(x)=a⊕x\lambda_a(x)=a\oplus x for each x∈Ax\in A has no continuous extension to A^\widehat{A}.

Since A1=(α+Q)∩[0,1]A_1=(\alpha+\mathbb{Q})\cap[0,1] is dense in [0,1][0,1] with the Euclidean topology, we can choose a sequence {bn:n∈N}⊆A1\{b_n:n\in\mathbb{N}\}\subseteq A_1 such that {bn:n∈N}\{b_n:n\in\mathbb{N}\} converges to 1/21/2 in [0,1][0,1] and

b2n<1/2,b2n+1>1/2b_{2n}<1/2,\qquad b_{2n+1}>1/2

for each n∈Nn\in\mathbb{N}. Hence, the sequence {bn:n∈N}\{b_n:n\in\mathbb{N}\} is dd-Cauchy in AA and {j(bn):n∈N}\{j(b_n):n\in\mathbb{N}\} converges to (1/2,1)(1/2,1) in [0,1]×Z=A^[0,1]\times\mathbb{Z}=\widehat{A}.

If λ^a\widehat{\lambda}_a has a continuous extension to A^\widehat{A}, then λ^a(j(bn))→λ^a((1/2,1))\widehat{\lambda}_a(j(b_n))\to\widehat{\lambda}_a((1/2,1)). This is impossible. Since

a⊕b2n=1/2+b2n,κ(a⊕b2n)=1a\oplus b_{2n}=1/2+b_{2n},\qquad\kappa(a\oplus b_{2n})=1

and

a⊕b2n+1=1,κ(a⊕b2n+1)=0,a\oplus b_{2n+1}=1,\qquad\kappa(a\oplus b_{2n+1})=0,

by j∘λa=λ^a∘jj\circ\lambda_a=\widehat{\lambda}_a\circ j, we have

λ^a(j(b2n))=j(λa(b2n))=j(a⊕b2n)⟶(1,1)\widehat{\lambda}_a(j(b_{2n}))=j(\lambda_a(b_{2n}))=j(a\oplus b_{2n})\longrightarrow(1,1)

and

λ^a(j(b2n+1))=j(λa(b2n+1))j(a⊕b2n+1)⟶(1,0).\widehat{\lambda}_a(j(b_{2n+1}))=j(\lambda_a(b_{2n+1}))j(a\oplus b_{2n+1})\longrightarrow(1,0).

This is contradiction. Thus the fixed translation has no continuous extension, and neither does the binary operation.

Example 5.2. There is a second countable and metrizable topological MV-algebra but not locally order convex, and hence, its MV-operation ⊕\oplus is not uniformly continuous for the canonical difference uniformity.

Proof. This example comes from [12 Example 1.1.11(a)]. For the sake of completeness, we give the construction.

Let AA be the finite and cofinite subsets of N\mathbb{N}, that is,

A={X⊆N:X is finite or cofinite}.A=\{X\subseteq\mathbb{N}:X\text{ is finite or cofinite}\}.

Then AA with the symmetric difference △\triangle is a Boolean group. Indeed, ∅\varnothing is the identity in (A,△)(A,\triangle), the inverse element of each X∈AX\in A is itself and X△Y=(X∪Y)∖(X∩Y)X\triangle Y=(X\cup Y)\setminus(X\cap Y) is in AA for each X,Y∈AX,Y\in A. One can easily show that the operator △\triangle is associative on AA. Put

Bk={2k−1,2k}B_k=\{2k-1,2k\}

for each k∈N∖{0}k \in\mathbb{N} \setminus\{0\} and

Un={⋃k∈FBk:F⊆{n,n+1,…} is finite}.(11)\mathcal{U}_n = \left\{\bigcup_{k \in F} B_k : F \subseteq\{n,n+1,\ldots\}\text{ is finite}\right\}. \tag*{(11)}

for each n∈Nn \in\mathbb{N}. Then each Un\mathcal{U}_n is a subgroup of the Boolean group (A,△)(\mathbb{A},\triangle) such that ⋂n∈NUn={∅}\bigcap_{n\in\mathbb{N}}\mathcal{U}_n=\{\varnothing\} and Un⊇Um\mathcal{U}_n \supseteq\mathcal{U}_m when m≥nm\ge n, because (⋃k∈F1Bk)△(⋃k∈F2Bk)=⋃k∈F1△F2Bk\left(\bigcup_{k\in F_1}B_k\right)\triangle\left(\bigcup_{k\in F_2}B_k\right)=\bigcup_{k\in F_1\triangle F_2}B_k. Put

τ={O⊆A: each X∈O there some n∈N such that X△Un⊆O}.(12)\tau=\{O\subseteq\mathbb{A}: \text{ each }X\in O\text{ there some }n\in\mathbb{N}\text{ such that }X\triangle\mathcal{U}_n\subseteq O\}. \tag*{(12)}

Then the Boolean group (A,△)(\mathbb{A},\triangle) with the topology τ\tau is a Hausdorff first countable topological group. Hence, it is second countable and metrizable, because A\mathbb{A} is countable.

Define ⊕\oplus and ∗* on A\mathbb{A} by

X⊕Y=X∪Y and X∗=N∖XX\oplus Y=X\cup Y\text{ and }X^*=\mathbb{N}\setminus X

for each X,Y∈AX,Y\in\mathbb{A}, respectively. Then (A,⊕,∗,0)(\mathbb{A},\oplus,*,0) is an MV-algebra, where 0=∅0=\varnothing. We claim that the MV-algebra (A,⊕,∗,0)(\mathbb{A},\oplus,*,0) with the topology τ\tau defined by (12) is a topological MV-algebra.

Since

(X△Un)∗={(X△Y)∗:Y∈Un}={N∖(X△Y):Y∈Un}={(N∖X)△Y:Y∈Un}={X∗△Y:Y∈Un}=X∗△Un,\begin{aligned} (X\triangle\mathcal{U}_n)^*&=\{(X\triangle Y)^*:Y\in\mathcal{U}_n\}\\ &=\{\mathbb{N}\setminus(X\triangle Y):Y\in\mathcal{U}_n\}\\ &=\{(\mathbb{N}\setminus X)\triangle Y:Y\in\mathcal{U}_n\}\\ &=\{X^*\triangle Y:Y\in\mathcal{U}_n\}\\ &=X^*\triangle\mathcal{U}_n, \end{aligned}

the operator ∗* is continuous.

Next we shall prove that ⊕:A×A→A\oplus:\mathbb{A}\times\mathbb{A}\to\mathbb{A} is continuous. Take any X,Y∈AX,Y\in\mathbb{A} and any neighbourhood VV of X⊕YX\oplus Y. Then there some n∈Nn\in\mathbb{N} such that (X⊕Y)△Un⊆V(X\oplus Y)\triangle\mathcal{U}_n\subseteq V. Since X,YX,Y are either finite or cofinite sets, there is m≥nm\ge n such that either Bk⊆ZB_k\subseteq Z or Bk∩Z=∅B_k\cap Z=\varnothing for each k≥mk\ge m, where Z=X,YZ=X,Y. We claim that

(X△Um)⊕(Y△Um)⊆(X⊕Y)△Un⊆V,(X\triangle\mathcal{U}_m)\oplus(Y\triangle\mathcal{U}_m)\subseteq(X\oplus Y)\triangle\mathcal{U}_n\subseteq V,

which implies that ⊕\oplus is continuous.

Indeed, take E,F∈UmE,F\in\mathcal{U}_m. Then

((X△E)⊕(Y△F))△(X⊕Y)=((X△E)∪(Y△F))△(X∪Y)⊆((X△E)△X)∪((Y△F)△Y)=E∪F.\begin{aligned} ((X\triangle E)\oplus(Y\triangle F))\triangle(X\oplus Y) &=((X\triangle E)\cup(Y\triangle F))\triangle(X\cup Y)\\ &\subseteq((X\triangle E)\triangle X)\cup((Y\triangle F)\triangle Y)\\ &=E\cup F. \end{aligned}

Since either Bk⊆ZB_k\subseteq Z or Bk∩Z=∅B_k\cap Z=\varnothing for each k≥mk\ge m (where Z∈{X,Y,E,F}Z\in\{X,Y,E,F\}), we have either

(((X△E)∪(Y△F))△(X∪Y))∩Bk=∅(((X\triangle E)\cup(Y\triangle F))\triangle(X\cup Y))\cap B_k=\varnothing

or

Bk⊆(((X△E)∪(Y△F))△(X∪Y))B_k \subseteq(((X \mathbin{\triangle} E) \cup(Y \mathbin{\triangle} F)) \mathbin{\triangle} (X \cup Y))

for each k≥mk \ge m. Note that E∪FE \cup F is finite. Hence,

((X△E)∪(Y△F))△(X∪Y)∈Um⊆Un.((X \mathbin{\triangle} E) \cup(Y \mathbin{\triangle} F)) \mathbin{\triangle} (X \cup Y) \in U_m \subseteq U_n.

This implies that

(X△E)⊕(Y△F)=(X△E)∪(Y△F)∈(X∪Y)△Un=(X⊕Y)△Un.(X \mathbin{\triangle} E) \oplus(Y \mathbin{\triangle} F) = (X \mathbin{\triangle} E) \cup(Y \mathbin{\triangle} F) \in(X \cup Y) \mathbin{\triangle} U_n = (X \oplus Y) \mathbin{\triangle} U_n.

Thus, we have proved

(X△Um)⊕(Y△Um)⊆(X⊕Y)△Un⊆V.(X \mathbin{\triangle} U_m) \oplus(Y \mathbin{\triangle} U_m) \subseteq(X \oplus Y) \mathbin{\triangle} U_n \subseteq V.

Finally, we shall show that the topological MV-algebra A\mathbb{A} is not locally order convex. If an order convex zero-neighborhood W⊆U1W \subseteq U_1 existed, it would contain some UnU_n. For k≥nk \ge n, both ∅\varnothing and BkB_k would be in WW, forcing {2k−1}∈W\{2k-1\} \in W, contrary to {2k−1}∉U1\{2k-1\} \notin U_1. Thus local order convexity fails.

Hence, form Theorem 4.7 it follows that the MV-operation ⊕\oplus is not uniformly continuous for the canonical difference uniformity.

Let A\mathbb{A} be the topological MV-algebra constructed in Example 5.2. We retain the notation BkB_k and UnU_n from the proof of Example 5.2. Take a point ∞∉N\infty\notin\mathbb{N}, and put Ω=N∪{∞}\Omega= \mathbb{N} \cup\{\infty\}. For X⊆ΩX \subseteq\Omega, define the set P(X)P(X) by

P(X)={k≥1:∣X∩Bk∣=1}.P(X) = \{k \ge1 : |X \cap B_k| = 1\}.

Put

B={X⊆Ω:P(X) is finite}.(13)\mathbb{B} = \{X \subseteq\Omega: P(X) \text{ is finite}\}. \tag*{(13)}

Define ⊕\oplus and ∗^* on B\mathbb{B} by

X⊕Y=X∪YandX∗=Ω∖XX \oplus Y = X \cup Y \quad\text{and} \quad X^* = \Omega\setminus X

for each X,Y∈BX,Y \in\mathbb{B}, respectively. Then (B,⊕,∗,0B)(\mathbb{B}, \oplus, ^*, 0_{\mathbb{B}}) is an MV-algebra, where 0B=∅0_{\mathbb{B}} = \varnothing. Indeed, for all X,Y∈BX,Y \in\mathbb{B},

P(Ω∖X)=P(X),P(X∪Y)⊆P(X)∪P(Y).P(\Omega\setminus X) = P(X), \qquad P(X \cup Y) \subseteq P(X) \cup P(Y).

It follows that B\mathbb{B} is closed under complementation and finite unions. Since ∅,Ω∈B\varnothing,\Omega\in\mathbb{B}, it is a Boolean algebra.

Example 5.3. Let topological MVMV-algebra A\mathbb{A} be from Example 5.2 and MVMV-algebra B\mathbb{B} defined by (13). Define a map

j:A⟶B,j(X)={X,X is finite,X∪{∞},X is cofinite.(14)j : \mathbb{A} \longrightarrow\mathbb{B}, \qquad j(X) = \begin{cases} X, & X \text{ is finite},\\ X \cup\{\infty\}, & X \text{ is cofinite}. \end{cases} \tag*{(14)}

Then there is a topology on MVMV-algebra B\mathbb{B} such that B\mathbb{B} is a canonically complete metrizable topological Boolean MVMV-algebra but not locally order convex and jj is a uniform MVMV-embedding with dense image, for the canonical difference uniformities. Consequently, (B,j)(\mathbb{B},j) realizes the canonical uniform completion of A\mathbb{A}.

Proof. For each n≥1n \ge1, set

Vn={⋃k∈SBk:S⊆{n,n+1,…}}.(15)V_n=\left\{\bigcup_{k\in S}B_k:S\subseteq\{n,n+1,\ldots\}\right\}. \tag*{(15)}

Then each VnV_n is a subgroup of the Boolean group (B,△)(\mathbb{B},\triangle). Indeed, if E=⋃k∈SBkE=\bigcup_{k\in S}B_k and F=⋃k∈TBkF=\bigcup_{k\in T}B_k belong to VnV_n, then

E△F=⋃k∈S△TBk∈Vn.E\triangle F=\bigcup_{k\in S\triangle T}B_k\in V_n.

Moreover,

Vn+1⊆Vn,⋂n≥1Vn={∅}.V_{n+1}\subseteq V_n,\qquad\bigcap_{n\ge1}V_n=\{\varnothing\}.

Put

τB={O⊆B: each X∈O there some n∈N such that X△Vn⊆O}.(16)\tau_{\mathbb{B}}=\{O\subseteq\mathbb{B}: \text{ each }X\in O\text{ there some }n\in\mathbb{N}\text{ such that }X\triangle V_n\subseteq O\}. \tag*{(16)}

Then the Boolean group (B,△)(\mathbb{B},\triangle) with the topology τB\tau_{\mathbb{B}} is a Hausdorff first countable topological group. Hence, it is metrizeble. We claim that the MV-algebra (B,⊕,∗,0B)(\mathbb{B},\oplus,{}^*,0_{\mathbb{B}}) with the topology τB\tau_{\mathbb{B}} defined by (16) is a topological MV-algebra.

Complementation is continuous because, for every E∈VnE\in V_n,

Ω∖(X△E)=(Ω∖X)△E.\Omega\setminus(X\triangle E)=(\Omega\setminus X)\triangle E.

Thus it maps X△VnX\triangle V_n onto X∗△VnX^*\triangle V_n.

To prove joint continuity of union at a fixed pair (X,Y)∈B2(X,Y)\in\mathbb{B}^2, let n≥1n\ge1. Choose

m=max⁡{n,1+max⁡(P(X)∪P(Y)∪{0})}.m=\max\{n,1+\max(P(X)\cup P(Y)\cup\{0\})\}.

Then neither ∣X∩Bk∣=1|X\cap B_k|=1 nor ∣Y∩Bk∣=1|Y\cap B_k|=1 with k≥mk\ge m. For arbitrary E,F∈VmE,F\in V_m, put

D=((X△E)∪(Y△F))△(X∪Y).D=((X\triangle E)\cup(Y\triangle F))\triangle(X\cup Y).

If k<mk<m, then E∩Bk=F∩Bk=∅E\cap B_k=F\cap B_k=\varnothing, so D∩Bk=∅D\cap B_k=\varnothing. If k≥mk\ge m, each of X,Y,E,FX,Y,E,F meets BkB_k in either ∅\varnothing or BkB_k. Consequently, D∩BkD\cap B_k is also either ∅\varnothing or BkB_k. Finally, ∞∉D\infty\notin D, because EE and FF do not change membership of ∞\infty. Hence D∈Vm⊆VnD\in V_m\subseteq V_n. This proves

{X′∪Y′:X′∈X△Vm, Y′∈Y△Vm}⊆(X∪Y)△Vn.\{X'\cup Y':X'\in X\triangle V_m,\ Y'\in Y\triangle V_m\}\subseteq(X\cup Y)\triangle V_n.

Thus union is jointly continuous, and B\mathbb{B} is a topological Boolean MV-algebra.

Next, we shall prove that the topological MV-algebra B\mathbb{B} is canonically complete. Define

εX={1,∞∈X,0,∞∉X,σ(X)=(εX,{(k,X∩Bk):k∈P(X)}).\varepsilon_X= \begin{cases} 1,&\infty\in X,\\ 0,&\infty\notin X, \end{cases} \qquad \sigma(X)=\left(\varepsilon_X,\{(k,X\cap B_k):k\in P(X)\}\right).

Put

H0=∅,Hr=⋃k=1rBk(r≥1).H_0=\varnothing,\qquad H_r=\bigcup_{k=1}^{r}B_k\quad(r\ge1).

We claim that

(X,Y)∈EVn⟺{σ(X)=σ(Y),X∩Hn−1=Y∩Hn−1,(17)(X,Y) \in E_{V_n} \Longleftrightarrow \begin{cases} \sigma(X)=\sigma(Y),\\ X \cap H_{n-1}=Y \cap H_{n-1}, \end{cases} \tag*{(17)}

where EVn={(X,Y)∈B×B:X∖Y∈Vn,Y∖X∈Vn}E_{V_n}=\{(X,Y)\in\mathbb{B}\times\mathbb{B}:X\setminus Y\in V_n,Y\setminus X\in V_n\} for each n∈Nn\in\mathbb{N}.

Suppose first that (X,Y)∈EVn(X,Y)\in E_{V_n}. Both directional differences avoid ∞\infty and Hn−1H_{n-1}, so XX and YY agree there. On any block BkB_k with k≥nk\ge n, both directional differences must be either empty or the whole block. If X∩BkX\cap B_k is a singleton, this forces Y∩Bk=X∩BkY\cap B_k=X\cap B_k: any other choice would make at least one directional difference a singleton on that block. The same argument with XX and YY interchanged shows that they have identical patterns.

Conversely, suppose the right-hand side of (17) holds. On each split block, XX and YY agree. On every other block, each is either empty or the whole block. Agreement on Hn−1H_{n-1} and at ∞\infty then shows that both directional differences are unions of whole blocks with indices at least nn. They belong to VnV_n, proving the claim.

Let F\mathcal{F} be any Cauchy filter on (B,UB)(\mathbb{B},\mathcal{U}_{\mathbb{B}}). For every n≥1n\ge1, choose

Qn∈F,Qn×Qn⊆EVn,Q_n\in\mathcal{F},\qquad Q_n\times Q_n\subseteq E_{V_n},

and put

Gn=⋂i=1nQi.G_n=\bigcap_{i=1}^{n}Q_i.

Then

Gn∈F,Gn≠∅,Gn+1⊆Gn,Gn×Gn⊆EVn.(18)G_n\in\mathcal{F},\qquad G_n\ne\varnothing,\qquad G_{n+1}\subseteq G_n,\qquad G_n\times G_n\subseteq E_{V_n}. \tag*{(18)}

By (17) with n=1n=1, all members of G1G_1 have the same pattern. Fix Y0∈G1Y_0\in G_1, and write

ε=εY0,F=P(Y0).\varepsilon=\varepsilon_{Y_0},\qquad F=P(Y_0).

The set FF is finite, and every member of every GnG_n has pattern σ(Y0)\sigma(Y_0).

For each k≥1k\ge1, the inclusion Gk+1×Gk+1⊆EVk+1G_{k+1}\times G_{k+1}\subseteq E_{V_{k+1}} implies that all members of Gk+1G_{k+1} agree on BkB_k. Hence the following definition is independent of the choice of YY:

Tk=Y∩Bk(Y∈Gk+1).T_k=Y\cap B_k\qquad(Y\in G_{k+1}).

If k∈Fk\in F, then TkT_k is the singleton prescribed by σ(Y0)\sigma(Y_0). If k∉Fk\notin F, then TkT_k is either ∅\varnothing or BkB_k. Set

Iε={∅,ε=0,{∞},ε=1,X=(⋃k≥1Tk)∪Iε.I_\varepsilon= \begin{cases} \varnothing,&\varepsilon=0,\\ \{\infty\},&\varepsilon=1, \end{cases} \qquad X=\left(\bigcup_{k\ge1}T_k\right)\cup I_\varepsilon.

Then P(X)=FP(X)=F, so X∈BX\in\mathbb{B}, and σ(X)=σ(Y0)\sigma(X)=\sigma(Y_0).

We show that F\mathcal{F} converges to XX. Fix n≥1n\ge1 and take any Y∈GnY\in G_n. If k<nk<n, then Gn⊆Gk+1G_n\subseteq G_{k+1}, whence Y∩Bk=Tk=X∩BkY\cap B_k=T_k=X\cap B_k. Also, XX and YY have the same pattern. Equation (17) therefore gives (X,Y)∈EVn(X,Y)\in E_{V_n}. Thus

Gn⊆EVn[X].G_n\subseteq E_{V_n}[X].

Since Gn∈FG_n \in\mathcal{F}, it follows that EVnB[X]∈FE^{\mathbb{B}}_{V_n}[X] \in\mathcal{F} for every nn. The sets EVnB[X]E^{\mathbb{B}}_{V_n}[X] form a neighborhood base for the canonical uniform topology, so F→X\mathcal{F} \to X. Every Cauchy filter therefore converges in B\mathbb{B}, proving canonical completeness.

The map jj in (14) is injective because j(X)∩N=Xj(X) \cap\mathbb{N} = X. Complementation interchanges the finite and cofinite cases, so

j(N∖X)=Ω∖j(X).j(\mathbb{N} \setminus X) = \Omega\setminus j(X).

Moreover, the union of two members of A\mathbb{A} is cofinite if and only if at least one of them is cofinite. Consequently,

j(X∪Y)=j(X)∪j(Y),j(∅)=∅.j(X \cup Y) = j(X) \cup j(Y), \qquad j(\varnothing) = \varnothing.

Thus jj is a Boolean MV-homomorphism.

We have the exact identity

Vn∩j(A)=j(Un)(n≥1).(19)V_n \cap j(\mathbb{A}) = j(U_n) \qquad(n \ge1). \tag*{(19)}

Indeed, the image of a cofinite set contains ∞\infty and hence cannot belong to VnV_n. The elements on the left are therefore precisely the finite unions of whole blocks with indices at least nn, which are the elements on the right. Preservation of symmetric difference now gives

(j(X)△Vn)∩j(A)=j(X△Un).(j(X) \mathbin{\triangle} V_n) \cap j(\mathbb{A}) = j(X \mathbin{\triangle} U_n).

Hence jj is a topological embedding.

The canonical basic entourages of A\mathbb{A} are

EUnA={(X,Y)∈A2:X∖Y∈Un, Y∖X∈Un}.E^{\mathbb{A}}_{U_n} = \{(X,Y) \in\mathbb{A}^{2} : X \setminus Y \in U_n,\ Y \setminus X \in U_n\}.

Since jj preserves directional differences, (19) yields

(j(X),j(Y))∈EVnB⟺(X,Y)∈EUnA.(j(X),j(Y)) \in E^{\mathbb{B}}_{V_n} \quad\Longleftrightarrow\quad (X,Y) \in E^{\mathbb{A}}_{U_n}.

The canonical uniformity of A\mathbb{A} is therefore exactly the pullback of the restricted canonical uniformity of B\mathbb{B}. This proves that jj is a uniform embedding.

Finally, we shall show that j(A)j(\mathbb{A}) is dense in B\mathbb{B}. Fix X∈BX \in\mathbb{B}, and put r0=max⁡(P(X)∪{1})r_0 = \max(P(X) \cup\{1\}). For every integer r≥r0r \ge r_0, define

ar={X∩Hr,∞∉X,(X∩Hr)∪(N∖Hr),∞∈X.a_r = \begin{cases} X \cap H_r, & \infty\notin X,\\ (X \cap H_r) \cup(\mathbb{N} \setminus H_r), & \infty\in X. \end{cases}

In the first case ara_r is finite, and in the second it is cofinite; thus ar∈Aa_r \in\mathbb{A}. The sets j(ar)j(a_r) and XX agree on HrH_r and at ∞\infty. All split blocks of XX lie in HrH_r, while neither set has split blocks beyond rr. Thus σ(j(ar))=σ(X)\sigma(j(a_r)) = \sigma(X), and (17) gives

(j(ar),X)∈EVr+1B.(j(a_r),X) \in E^{\mathbb{B}}_{V_{r+1}}.

For any fixed nn, this pair belongs to EVnBE^{\mathbb{B}}_{V_n} whenever r≥max⁡{r0,n−1}r \ge\max\{r_0,n-1\}. Hence j(ar)→Xj(a_r) \to X, proving that j(A)j(\mathbb{A}) is dense in B\mathbb{B}.

We have shown that B\mathbb{B} is a complete separated uniform space and jj is a uniform embedding with dense image.

The natural order in these Boolean MV-algebras is inclusion. Given an order-convex zero-neighborhood W satisfying

∅∈W⊆V1.\varnothing\in W \subseteq V_1.

There would then exist nn with Vn⊆WV_n \subseteq W. For any k≥nk \ge n, both ∅\varnothing and BkB_k belong to WW, and

∅⊆{2k−1}⊆Bk.\varnothing\subseteq\{2k-1\} \subseteq B_k.

Order convexity would force {2k−1}∈W⊆V1\{2k-1\} \in W \subseteq V_1, which is impossible because every member of V1V_1 is a union of whole blocks. Thus B\mathbb{B} is not locally order convex.

References

  1. [1]A.V. Arhangel’skii, M.G. Tkachenko, Topological Groups and Related Structures, Atlantis Studies in Mathematics, vol. 1, Atlantis Press, Paris; World Scientific Publishing Co. Pte. Ltd., Hackensack, NJ, 2008.DOI
  2. [2]R. Engelking, General Topology, Sigma Series in Pure Mathematics, Heldermann, Berlin, 1989.
  3. [3]C.C. Chang, Algebraic analysis of many-valued logic, Trans. Am. Math. Soc. 88 (1958) 467–490.DOI
  4. [4]R. L. O. Cignoli, I. M. L. D’Ottaviano, and D. Mundici, Algebraic Foundations of Many-Valued Reasoning, Trends in Logic, vol. 7, Kluwer Academic Publishers, Dordrecht, 2000.DOI
  5. [5]A. P. Gan, W. Luan, F. A. Deng, and Y. C. Yang, On regularity of topological MV-algebras, Fuzzy Sets and Systems 475 (2024), 108747.DOI
  6. [6]C. S. Hoo, Topological MV-algebras, Topology Appl. 81 (1997), 103–121.DOI
  7. [7]Q. Li and Y. Yang, Locally convex topological MV-algebras and uniformization of MV-algebras, Topology and its Applications 382 (2026), 109771.DOI
  8. [8]W. Luan, H. Weber, and Y. Yang, Filter topologies and topological MV-algebras, Fuzzy Sets and Systems 406 (2021), 11–21.DOI
  9. [9]L. Nachbin, Topology and order, D. Van Nostrand Company, New York, 1965.
  10. [10]D. Mundici, Interpretation of AF C*-algebras in Łukasiewicz sentential calculus, Journal of Functional Analysis 65 (1986), 15–63.DOI
  11. [11]W. Roelcke, S. Dierolf, Uniform structures on topological groups and their quotients, McGraw-Hill International Book Co., New York 1981.
  12. [12]H. Weber, Uniform lattices. I: A generalization of topological Riesz spaces and topological Boolean rings, Annali di Matematica Pura ed Applicata (4) 160 (1991), 347–370.DOI
  13. [13]H. Weber, On topological MV-algebras and topological ℓ-groups, Topology and its Applications 159 (2012), 3392–3395.

Paper details

vlnc:W_dnnsymhw
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 9

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents