GRZEGORCZYK POINTS AND FILTERS IN BOOLEAN CONTACT ALGEBRAS

Abstract The purpose of this paper is to compare the notion of a Grzegorczyk point introduced in [19] (and thoroughly investigated in [3, 14, 16, 18]) to the standard notions of a filter in Boolean algebras and round filter in Boolean contact algebras. In particular, we compare Grzegorczyk points to filters and ultrafilters of atomic and atomless algebras. We also prove how a certain extra axiom influences topological spaces for Grzegorczyk contact algebras. Last but not least, we do not refrain from a philosophical interpretation of the results from the paper.

§1. Introduction. In [16,18] we carried out an extensive analysis of one of the first systems of point-free topology by Grzegorczyk [19], based on the notion of separation (the dual notion of contact). The two papers witnessed the formulation of representation and duality theorems for a subclass of so-called Grzegorczyk structures. In particular, we established an object duality between a class of Grzegorczyk structures that satisfy the countable chain condition and the concentric spaces (a subclass of the class of regular spaces), which satisfy the topological version of the same condition.
A particular idiosyncrasy of Grzegorczyk's approach to point-free topology is his definition of a point, which is a formal reflection of the geometrical intuition of a point as a system of "shrinking" regions of space. We have devoted much attention to elucidating the idea in [14][15][16]18], so we refer the reader to these works if they want to familiarize themselves with an informal (yet precise) analysis of the notion.
One of the problems that occupied us in the aforementioned works was the relation of Grzegorczyk points to other classical point-like constructions, such as ultrafilters and maximal round filters. We have only scratched the surface of the problem, yet we demonstrated that the following two second-order sentences: (a) Every Grzegorczyk point is an ultrafilter, and (b) Every ultrafilter is a Grzegorczyk point correspond to the following first-order and second-order sentences, respectively: (a ) Each region is separated from its complement.
(cf. Theorem 9.1) We proved as well that every Grzegorczyk point is a maximal round filter. For the sake of completeness of the presentation, some of these results are contained in this paper.
Here, we aim to examine how Grzegorczyk points behave and relate to filters and ultrafilters in atomless and atomic Boolean contact algebras, which in a very natural way relate to Grzegorczyk structures from the earlier works. To be more precise, previously we worked with mereological fields, which are structures that always lack the zero element and might lack the unity. In all other respects, they bear a very strong resemblance to Boolean algebras, thanks to the classical theorems to which we refer in Sections 6 and 7 of [16]. Thanks to these theorems, all results obtained in the earlier works are easily translatable into analogous results in the framework of Boolean algebras. We have also replaced the primitive binary relation of separation with its dual, the contact relation. The main reason to do so was the fact that since more or less the beginning of the twenty-first century and the seminal works of Stell [23], Düntsch & Winter [10,11] and Dimov & Vakarelov [8,9], Boolean contact algebras have been the standard framework for doing region-based topology. Thus, with technical results in focus and philosophical considerations aside, contact algebras are the most natural choice for the presentation of advances in the study of Grzegorczyk point-free topology. §2. Motivations. There are various reasons for doing region-based theories of space. Probably the main one is to describe spatial relations in terms of concepts that can be rooted in the sensory experience. Such were motivations for both Region Connection Calculus of Cohn et al. [5] and Boolean contact algebras. From the purely mathematical point of view, algebraic parts of these theories focused on finding suitable topological representations and duality theorems, most notably via the spaces of clusters as, among others, in the works of Düntsch & Winter [11] and Dimov & Vakarelov [8]. In closely related De Vries algebras from [7] maximal round filters play a crucial role in algebraization of the topological notion of compactness.
The approach taken by Grzegorczyk [19] seems to have yet another motivation-to find spatially satisfactory construction of points in terms of regions and separation (equivalently: contact) relation. Grzegorczyk points are, in a way, generalizations of Tarski's points as sets of concentric balls from his geometry of solids [25], and they embody the idea of a point as a system of shrinking regions of space.
From both philosophical and technical points of view, it is interesting (at least for us, as the authors of the paper) how second-order conditions expressing dependencies between various constructions of points relate to either first-or second-order properties of Boolean contact algebras. This line of research was initiated by Biacino & Gerla [3] with finding sufficient conditions for Grzegorczyk points and Whitehead points from [26] to coincide. We believe that further investigation of dependencies between various constructions of points can bring interesting results in region-based theories of space, even more so because it is very little examined fragment of such theories.
As pointed to in the introduction, in our previous works we showed certain correspondences between statements about points of contact algebras and their properties. In particular, we found it interesting how Grzegorczyk points relate to the standard filter constructions of Boolean contact algebras. Among these, the Fréchet filter plays a distinguished role in the class of atomic Boolean algebras, and, interestingly, its relation to Grzegorczyk points is independent from the axioms of Grzegorczyk contact algebras. We venture to say that this may be interpreted as the fact that the notion of Fréchet filter is independent from the spatial intuitions behind the definition of Grzegorczyk points. Investigating consequences of the statement "Fréchet filter is a Grzegorczyk point" may be interpreted philosophically as answering the question what are consequences of forcing our spatial intuitions about points into the notion of Fréchet filter? To us as the authors, logicians with a strong predilection toward philosophical aspects of mathematical theories, it seemed a very stimulating question which is faced and solved in the paper.
It is hard to say now what, if any, fruits bears the theory of correspondences between statements about points of Boolean contact algebras and their familiar properties. However, in our opinion, it is a path that is worth studying as it may lead to new perspectives on various-both mathematical and philosophical-aspects of regionbased theories of space. §3. Boolean quasi-contact algebras. This section is a refresher of the standard notions, axioms and elementary properties of contact algebras that are essential for the key results of the paper.
A Boolean quasi-contact algebra (BQCA for short) is any pair R = B, C , where B = R, , , -, 0, 1 is a non-degenerate Boolean algebra 1 (BA for short) and C is a binary contact relation on R which meets the following four axioms: 2 where C / is the complement of C, R + := R \ {0} and ≤ is the standard partial order defined by: Furthermore, we will use the standard strict partial order defined by: All elements of the domain R will be called regions; the region 0 will be called empty; and all regions from R + will be called non-empty. In the case x C y (resp. x C / y; x ≤ y; x y) we will say that x is in contact with y (resp. x is separated from y; x is part of y; x is a proper part of y). Axioms (C0)-(C3) say, respectively: 0 is separated from any region; C is symmetric and reflexive on R + ; and if one region is part of another, then each region in contact with the first is also in contact with the other. In BAs condition (C3) is equivalent to From (C3) and the reflexivity of ≤ we obtain that in place of (C2) we can assume that C includes ≤ limited to non-empty regions: So every non-empty region is in contact with 1: We will also use two auxiliary binary relations on R: and ⊥. We define the first by: and we let the second be the complement of . In the case x y (resp. x⊥y) we will say that x overlaps y (resp. x is disjoint from y). From (C3), (C1) and (≤ | R + ⊆ C) we have: Yet another auxiliary standard relation of non-tangential inclusion is defined by: In the case x y we say that x is a non-tangential part of y. From (3.3) it follows that: From (C0), (C1), (≤ | R + ⊆ C), (df ) we obtain that every region is a non-tangential part of 1 and each non-tangential part of a given region is its (ordinary) part: From (df ), (df ) and the transitivity of ≤ we have: Furthermore, we have the following property of : It-together with ( ⊆≤) and the reflexivity of ≤-gives the transitivity of : We will use three families of filters in BAs. The first one will be the family Ult of all ultrafilters, i.e., proper maximal filters. The second one will be the family of principal filters of the form ↑ x := {u ∈ R | x ≤ u} for an x ∈ R (the set ↑ x is called a principal filter generated by x). Let PF and PF At be families of all principal filters and of principal filters generated by atoms, respectively. The third family will be the family of free filters which satisfy the following condition: there is no x ∈ R such that for any y ∈ F we have x ≤ y. Every ultrafilter is free or belongs to PF At .
Moreover, two families of filters in BQCAs will play special roles in the sequel. The first one will be the family Rnd of so-called round (or contracting) filters which are proper filters satisfying: • for any x ∈ F there is a y ∈ F such that y x.
From ( ⊆ C), (df ) and ( ⊆≤) we obtain: The second one is the family M.Rnd of all proper maximal filters in the family Rnd. 3 Lemma 3.2 [18, lemma 10.6]. If a round filter F satisfies the condition ∀ x,y∈R (∀ z∈F z C x ∧ x y) =⇒ y ∈ F , then F belongs to M.Rnd.
It is easy to check that the structures B, and B, R + × R + are BQCAs. We will call them trivial BQCAs. The relation is the smallest contact relation on B (cf. [10, proposition 2.2(1)]. From (df ), ( ⊆ C) and (C2) we obtain: Hence all filters in B, are round; and so Ult = M.Rnd. §4. Boolean contact algebras. After [10], we will call a Boolean contact algebra (BCA for short) any BQCA satisfying From (3.1) and (C4) we have: By means of (3.5), (df ), (C3), ( ⊆≤), (≤ | R + ⊆ C), (C1), (3.2) and the antisymmetry of ≤, we can prove: In BQCAs condition (C5) is equivalent to each of the following: 213] analyzed contact algebras as structures of the form B, C satisfying (C0)-(C2) and (4.1). The algebras that additionally satisfy (ext C ) are named by them as extensional contact algebras (p. 215). In Lemma 2.2 of [8] the authors demonstrate that conditions (ext C ), (C5), (∃ ) and (C3 c ) are equivalent in any contact algebra. But, as the above lemma shows, BQCAs are enough for this, i.e., the condition (C4) can be omitted in the proofs.
It is easy to check that trivial BQCAs of the form B, (see p. 21) are BCAs. In the sequel we will use 'BCA ' as a general name for all BCAs with C = . The BQCAs of the form B, Let At be the set of atoms of a given BA. Directly from (∃ ) and ( ⊆≤) we have: Hence, using ( ⊆ C), (df ) and (3.7), we obtain: From this and (df ), (C1), and (4.1), we get a generalization of (4.2): ∀ a 1 ,...,an ∈At a 1 ··· a n a 1 ··· a n .
For any x ∈ R we put: It is clear that the atomicity of a given BA implies that x = sup At x . A BQCA is atomic iff it is based on an atomic BA. The following fact will be used in the proof of Theorem 7.8.

Lemma 4.3. In any atomic BCA we have:
Proof. Assume for a contradiction that x C y, x⊥y and At x is finite. Then for some atoms a 1 , ..., a n we have x = a 1 ··· a n . So, by (C1), (4.1) and (4.3), we have: x C y iff a 1 C y or ... or a n C y iff a 1 ≤ y or ... or a n ≤ y. Therefore x y, a contradiction.
We say that a BA is finite-cofinite iff for any x ∈ R either x ∈ {0, 1} or there are n > 0 and a 1 , ... , a n ∈ At such that either x = a 1 ··· a n or x =-(a 1 ··· a n ). Clearly, all finite BAs are finite-cofinite. Düntsch & Winter [10] showed that the only BCAs that can be obtained from finite-cofinite BAs are the trivial BCAs .
is the only BCA based on B.
Notice that every non-empty region which is not an atom has at least two nontangential non-empty proper parts which are separated from each other: Indeed, assume that x ∈ R + \ At. Then there are y, z ∈ R + such that y⊥z and x = y z. Hence, by (∃ ), there are u, v ∈ R + such that u y x and v z x. So, by (3.6) and ( ⊆≤), we have u C / v. Moreover, by (3.7), we have u x, v x and A BQCA is atomless iff it is based on an atomless BA.
] define another type of filters. Namely, a filter F is an end iff F is a round filter satisfying the following condition: They show (Lemma 3.2(viii)) that all end filters belong to M.Rnd. Moreover, they prove (Proposition 3.1(iii)) that in any BCA satisfying the so-called interpolation axiom every filter from M.Rnd is an end.
(ii) Gruszczyński [14] proves that in any BCA satisfying (IA), a filter F belongs to M.Rnd if and only if F satisfies the following condition: The following lemma will be used to obtain Proposition 4.8.
Lemma 4.6. In any atomless BCA, any filter F satisfying the following condition is free: Proof. Suppose that in an atomless BCA for a filter F there are x, y ∈ R + such that (a) x⊥y and (b) for any z ∈ F : x ≤ z and y ≤ z. Then, by (a), either x or y does not belong to F. Assume the first possibility holds: (c) x / ∈ F . By (4.5), for some u 1 ∈ R + we have u 1 x and u 1 = x. Observe that by (b) and (3.7), u 1 is a non-tangential part of any z ∈ F . By means of (4.5) and the axiom of dependent choices we produce a sequence u n | n < such that u 0 := x, u i+1 u i and u i+1 = u i . The filter ↑ X generated by the set of the terms of the sequence is obviously round, and by (b) and (c) it is the case that F ↑ X . Therefore, F / ∈ Ult and F / ∈ M.Rnd.
Finally, from Lemma 4.6, Proposition 4.7 and ( ⊆ C) we obtain: A pre-point of R is any non-empty set Q of non-empty regions which satisfies the following conditions: The purpose of this definition is to formally grasp the intuition that a point is a system of diminishing regions determining a unique location in space. We call it a pre-point, since if we understand a point as a perfect representation of some location in space, then it may happen that two different sets of regions represent one and the same location. Further, we will identify such pre-points to act as one point. Let Q be the set of all pre-points of R.
Notice that by ( ⊆≤), all pre-points are chains with respect to the part of relation. Furthermore, by (r1), ( ⊆≤) and (df≤), all pre-points have the finite intersection property. So each pre-point Q generates the filter By (4.2) and ( ⊆ C), all singletons composed of atoms are pre-points in all BCAs: Lemma 5.1. In any BCA, for any a ∈ At the singleton {a} is a pre-point. §6. Grzegorczyk points.

Grzegorczyk points of BQCAs.
Let R be a BQCA based on a BA B. By a Grzegorczyk point (G-point for short) of R we will mean any filter generated by a pre-point of R. Thus, for any filter F we have: Let Gpt be the set of all G-points of R. Its elements will be denoted by small gothic letters 'p' and 'q'.
Every G-point is a maximal filter in Gpt (see [16]): Furthermore, in [18] we proved that in all BQCAs: For Grzegorczyk contact algebras, which are presented in Section 8 and which are BCAs, we obtain the converse implications (8.2) and (8.3).
In the light of (r2) and (3.7) we have Gpt ⊆ Rnd. Hence, by (6.2) and Lemmas 3.1 and 3.2, for any BQCA we obtain the following fact: Proposition 6.1. Every G-point in a BQCA belongs to M.Rnd. 5 Hence, in the light of Lemma 3.1, condition (6.1) is equivalent to the following (cf. condition (%)): Now we will consider BQCAs which are complete. Proof. Ad 1. Every ultrafilter is free or belongs to PF At . Every G-point is generated by a chain. But no free ultrafilter in a complete BA is generated by a chain (see, e.g., [21, lemma 43]). Ad 2. Directly from point 1.

G-points of BCAs.
In the general case of BCAs, from Lemma 5.1 and Proposition 6.1 we obtain:   1)). We will now answer the following two questions: Are there any other G-points than principal ultrafilters in infinite atomic algebras? and In which infinite atomic BCAs the Fréchet filter can or cannot be a G-point? The subclass of BCAs which will serve to answer those question corresponds to a certain subclass of GCAs.
Let B be an infinite atomic BA. Then the set cAt of all coatoms has a finite intersection property and generates the so-called Fréchet (or cofinite) filter F c , which is free. We will use the following properties of F c . Lemma 7.1. For any x ∈ F c there are a 1 , ... , a n ∈ At (n > 0) such that x =-(a 1 ··· a n ).

From Lemma 7.1 and Proposition 4.4 we obtain:
Corollary 7.2. If F c ∈ Ult, then B is finite-cofinite, and so B, is the only BCA based on B.
The following two lemmas are analogous to the facts given in [4], where the completeness of a Boolean algebra B is assumed and the fact that B is isomorphic to the algebra of sets over At is used. Let B have countably many atoms: a 1 , a 2 , .... Then F c is generated by the chain C c := {a 1 ···a n | n ∈ ω}.
Proof. If x ∈ F c then for some k > 0 and a i 1 , ..., a i k ∈ At we have - Furthermore, if for some n ∈ ω we havea 1 ···a n ≤ x, thenx ≤ a 1 ··· a n , which means thatx is composed of finitely many atoms, i.e., x ∈ F c . So also ↑ C c ⊆ F c .
The proof of the lemma below is after [4], yet omits the completeness assumption.

Lemma 7.4. If B has uncountably many atoms, then F c is not generated by a chain.
Proof. Assume for a contradiction that F c is generated by a chain C. Then C ⊆ F c . Hence, by Lemma 7.1, for any x ∈ C there are n > 0 and a 1 , ... , a n ∈ At such that x = a 1 ··· a n . Therefore the set At ⊥ x is finite. Furthermore, since C is a chain, for all x, y ∈ C such that x = y either At ⊥ ∈ ω is injective; and so C and x∈C At ⊥ x are countable. Hence for some a 0 ∈ At we have a 0 / ∈ x∈C At ⊥ x , i.e., for any x ∈ C we have xa 0 . However, for all a ∈ At it is the case thata ∈ F c . Therefore, since F c is generated by C, for some x a ∈ C we have x a ≤a; a contradiction.
Furthermore, let us remind that: Lemma 7.5. F c is included in any free filter in B. Now we will show two classes of BCAs in which F c is not a G-point. From them we will also get two classes of GCAs in which F c / ∈ Gpt (see Section 10). Keep in mind that whether F c is generated by the chain depends on whether B has countably many atoms.
Because all G-points are generated by chains, in the light of Lemma 7.4 we obtain: Proposition 7.6. There is no atomic BQCA with uncountably many atoms such that: Moreover, we have the following generalization of model 6.18 from [14]: There is no infinite complete atomic BCA that satisfies (pF c ).
Proof. Let R = B, be based on an infinite complete atomic BA. If R has uncountably many atoms, then we use Proposition 7.6. Otherwise, F c is free and generated by a chain, by Lemma 7.3. Hence F c / ∈ Ult, because no free ultrafilter in a complete BA is generated by a chain (see, e.g., [21, lemma 43]). But Gpt ⊆ Ult, by Lemma 6.8. Thus, we have F c / ∈ Gpt.
In consequence, there are BCAs that do not satisfy (pF c ). In Theorem 7.9 we will construct a certain class of BCAs satisfying (pF c ). Moreover, in Theorem 10.1 we will show that elements of the class are also GCAs. Now we will focus on some properties of infinite atomic BCAs in which F c is a G-point.
Theorem 7.8. Let R be an infinite atomic BCA satisfying (pF c ). Then: 1. R has countably many atoms. 2. For all x, y ∈ R: x C y iff either x y or both At x and At y are infinite. 7 3. F c is the only G-point of R being a free filter.
6. If F c ∈ Ult , then Gpt = Ult and R is incomplete and finite-cofinite.
Proof. Ad 1. Directly from Proposition 7.6. Ad 2. The "⇒"-part we obtain directly from Lemma 4.3. For the "⇐"-part assume that At x and At y are infinite. Then, by Lemma 7.1, for any z ∈ F c we have: z x and z y. Hence, since F c ∈ Gpt, by condition (r3), we have x C y.
Ad 3. By Lemma 7.5, F c is included in any free filter. In consequence, if F c belongs to Gpt, then its maximality in the set of G-points excludes any free filter from the set Gpt . In other words, none of the other G-points is a free filter.  Ad 6. Suppose that F c ∈ Ult. Then Gpt ⊆ Ult, by point 5. For the converse inclusion assume that U ∈ Ult. But each ultrafilter in a given BA either belongs to PF At or is free. In the first case, U ∈ Gpt, by point 3. In the second case, by Lemma 7.5, we have F c ⊆ U . So U = F c ∈ Gpt. Thus, we also obtain that Ult ⊆ Gpt.
Below we describe a construction of BCAs which satisfy (pF c ). The construction is a generalization of model 6.19 from [14]. Theorem 7.9. Let B be an infinite atomic BA with countably many atoms. For all x, y ∈ R we put: x C y :⇐⇒ either x y or both At x and At y are infinite.

Then:
1. For all x, y ∈ R we have:

For the chain C c from Lemma 7.3 and x ∈ R we have:
x overlaps with each member of C c ⇐⇒ At x is infinite.

C c is a pre-point and F c is a G-point of B, C .
Proof. Ad 1. For all x, y ∈ R: x y iff xC /y iff x⊥y and either At x or At -y is finite iff either both x ≤ y and At x is finite, or both x ≤ y and At -y is finite.
Ad 2. It is obvious that B, C satisfies (C0)-(C2) and (C4). Moreover, from point 1 we have (∃ ). For (C3) suppose that x ≤ y and z C x. If z x, then z y. If both At z and At x are infinite, then At y is also infinite; and so z C y. For any x ∈ C c the set At -x is finite, since for some n ∈ ω we havex = a 0 ··· a n . For (r1) suppose that u, v ∈ C c and u = v. Then u ≤ v or v ≤ u. Thus u v or v u, by point 1. For (r2) notice that in the case u ∈ C c for some n ∈ ω we have: v :=a 1 ···a na n+1 ≤a 1 ···a n = u.
So v ∈ C c and v u, by point 1. For (r3) suppose that x and y overlap with each element of C c . Then, by ( †), At x and At y are infinite. So x C y.
To round off the proof we use Lemma 7.3.
Thus, directly from Theorems 7.8 and 7.9 we obtain the following characterization of infinite atomic BCAs having the Fréchet filter as a G-point: x C y ⇐⇒ either x y or both At x and At y are infinite.
Remark 7.11. (i) For any set X, let P(X ) be the power set algebra of X. Any given atomic BA is monomorphic to P(At). Namely, it is isomorphic to the field of sets {At x | x ∈ R}. If a BA is complete, it is isomorphic to P(At). Thus, any atomic BA with countably infinitely many atoms is isomorphic to a dense subalgebra of P(ω). Moreover, if such a BA is complete, then it is isomorphic to P(ω).
(ii) With reference to Theorems 7.8 and 7.9, let us observe that we can distinguish three kinds of infinite atomic BAs with countably many atoms: 1. finite-cofinite-they are incomplete and isomorphic to the finite-cofinite algebra of sets on ω (shortly: FC(ω)); 2. complete-they are isomorphic to P(ω); 3. incomplete, but not finite-cofinite-each is isomorphic to a dense subalgebra of P(ω). This type includes, for example, the Borel algebra of the space of rational numbers with the standard topology (see, e.g., [22, p. 21]).
If in such BAs we introduce the relation C from Theorem 7.9, we obtain BCAs in which F c is a G-point. For the first type of BAs, we get C = , F c ∈ Ult and Gpt = Ult; but for the other two types, C and F c / ∈ Ult (cf. Proposition 4.4 and Theorem 7.8). By Theorems 7.8 and 7.9, the above three types of BAs constitute all infinite atomic BAs satisfying (pF c ).
It is often walking on thin ice when trying to infer philosophical conclusions from mathematical results. Nevertheless, we venture to sketch a spatial perspective on Corollary 7.10. The notion of Fréchet filter is one of the central concepts of the theory of atomic Boolean algebras. As is well-known, there are algebras whose Stone spaces of points contain the filter, and in the case a BA has the Fréchet filter, it is a subset of every point (ultrafilter) of the Stone space. The notion of a Grzegorczyk point is motivated by geometrical intuitions about the pre-theoretical concept of point. Thus, the sentence "the Fréchet filter is a Grzegorczyk point" might be viewed as saying that the Fréchet filter of a BA satisfies the geometrical intuitions about points. Corollary 7.10 shows that the class of algebras whose Fréchet filter meets those intuitions is relatively narrow, and as such, it testifies to a kind of limitation result about the dependency between the spatial, region-based motivations and purely algebraic constructions. Most of the Boolean algebras have Fréchet filters that do not follow the idea of a point embodied in the definition of Grzegorczyk's. §8. Grzegorczyk contact algebras. We extend the axioms for the class of BQCAs with the following second-order postulates: called Grzegorczyk axioms, introduced in [16, Proposition 6.1] and equivalent to the single axiom from [19]. Every BQCA which satisfies (G ) and (G ⊥ ) is called a Grzegorczyk contact algebra (GCA in short). It has been demonstrated in [16, Theorem 6.4] that (C0)-(C3) together with (G ) and (G ⊥ ) entail (C4) and (C3 c ). 8 All other conditions from Lemma 4.1, i.e., (C5), (∃ ), (ext ≤ ) and (ext C ), hold as well. (We also get (∃ ) from the reflexivity of on R + , (G ), (r2) and (3.7).) Therefore, Notice that in virtue of Lemma 5.1, we have that:

Lemma 8.2. All atomic BCAs satisfy (G ).
All GCAs with C = will be called GCAs . In this case we get (G ⊥ ) in a trivial way. Thus, we obtain: Proposition 8.3. Every atomic BCA is a GCA and vice versa.
Thanks to Proposition 6.9 we obtain examples of GCAs in which Gpt M.Rnd (all infinite complete atomic BCAs ). 9 In any GCA, by (G ), the definition of Gpt and the reflexivity of on R + , we have: The following properties of G-points of R play an important role in the analysis of GCAs (see [14,18]): We get that all points of a given GCA are ultrafilters if and only if it is trivial (and so each of its region is separated from its complement): Theorem 8.4. In any GCA: Proof. From (df ) and (8.4) we obtain: So we use (3.8), Lemma 6.8 and a property of ultrafilters. §9. G-points, ultrafilters, principal filters and atoms of GCAs. Due to the seminal duality established by Stone [24] for BAs, ultrafilters are candidates for points of GCAs and therefore a question how they relate to G-points comes to mind in a natural way. A partial answer, formulated in [14,18], boils down to the following theorem which is closely related to Corollary 6.10. (1). The proof of the theorem refers to two nontrivial properties of BAs: any infinite BA has a free ultrafilter and an infinite antichain and no free ultrafilter in a complete BA is generated by a chain (for the latter see, e.g., [21,Lemma 43]).
Theorem 9.1. If R is a complete GCA, the following conditions are equivalent: In Remark 7.11, we showed that FC(ω), is an example of an infinite incomplete atomic BCA (and so also GCA ) in which we have Gpt = Ult.
In the sequel, we will consider the relationship between atoms and principal filters in GCAs. To begin with, we show that we can strengthen Proposition 6.3.
Proof. The "⇒"-part we obtain directly from Propositions 6.3 and 8.1. For the "⇐"-part, assume that x / ∈ At. Then for some y ∈ R + we have y x. Moreover, by (8.1), for some p 0 ∈ Gpt we have y ∈ p 0 . Hence ↑ x p 0 . Therefore, ↑ x is not maximal in the family of all round filters, since p 0 ∈ Rnd. Thus, we have ↑ x / ∈ M.Rnd.
Furthermore, we obtain the following theorem which previously appeared in [14,18] and which is a strengthening of Proposition 6.3 and Corollary 6.4 for GCAs: Theorem 9.3. In any GCA, for any x ∈ R: Proof. Obviously: x ∈ At iff ↑ x ∈ Ult. From Propositions 6.3 and 8.1 for any x ∈ At we have ↑ x ∈ Gpt. So if ↑ x ∈ Ult then ↑ x ∈ Gpt. Let ↑ x ∈ Gpt. Then ↑ x ∈ M.Rnd, by Proposition 6.1. Hence x ∈ At, by Lemma 9.2. So also ↑ x ∈ Ult.
From the above theorem we have the following generalization of Proposition 6.6: Corollary 9.4. In any GCA, PF ∩ Ult = PF At = Gpt ∩ PF. §10. G-points and free filters in infinite atomic GCAs. This section is devoted to GCAs based on an infinite atomic BAs. In this case condition (pF c ) is independent from the axioms of GCAs.
Firstly, there are GCAs that do not satisfy (pF c ). Indeed, in the light of Proposition 8.1 and Propositions 7.6 and 7.7, respectively, we get that: no atomic GCA with uncountably many atoms or complete atomic GCA with infinitely many atoms fulfills (pF c ).
Secondly, we show that there are GCAs satisfying (pF c ). They are given by the construction we used in Theorem 7.9. Finally, notice that in any infinite atomic GCAs in which (pF c ) holds, the filter F c has all properties from Theorem 7.8. §11. Topological spaces of G-points for infinite atomic GCAs. This section shows how the relation between G-points and ultrafilters influences the topological spaces of G-points.
Let R = B, C be a GCA. For any region x of R we put Ult(x) := {U ∈ Ult | x ∈ U }. It is well-known that the family {Ult(x) | x ∈ R} is a topological basis on Ult. By the Stone space of B we mean-in the standard way-the topological space S(B) in which points are ultrafilters of B and the topology is introduced via the basis {Ult(x) | x ∈ R}. 10 Furthermore, for any region x of R we define the set of all its internal G-points: Of course, Irl(0) = ∅ and Irl(1) = Gpt. Moreover, for all x, y ∈ R: if x y, then Irl(x y) = Irl(x) ∩ Irl(y); x ≤ y iff Irl(x) ⊆ Irl(y) (see [18, p. 826]). It is routine to verify that the family B R := {Irl(x) | x ∈ R} is a topological basis on Gpt . Let T R := Gpt, O R be the topological space introduced via B R . One can prove that every set from B R is regular open in T R , and so T R is semi-regular. Moreover, T R is concentric, i.e., it is a T 1 -space and each point p ∈ Gpt has a local basis B p R satisfying the following condition: It is known that every topological space having the same properties as T R is also a regular space (see [14,16,18] for details). 11 In [18, p. 839] it is shown that: the base B R consists of clopen sets if and only if R is a GCA . Thus, in the general case, the base of T R does not have to consist of clopen sets. However, below in Proposition 11.3 we can prove that for any infinite atomic GCA R satisfying (pF c ), T R is a continuous image of a certain Boolean space (i.e., a Hausdorff, compact and zero-dimensional space).
Given a topological space T = S, O , its compactification is a pair K, e such that: K = K, K is a compact space, e : S → K is an embedding (an injective continuous map) of T into K (so e is a homeomorphism between S and e[S]) and the set e[S] is dense in K. Moreover, we say that K is a compactification of T iff e is the identity map.
The following lemma will be used in the proof of the last theorem of the paper.
Lemma 11.1. Let T 1 = S 1 , O 1 , T 2 = S 2 , O 2 and K = K, K be topological spaces such that: • K, e is a compactification of T 1 for some e : S 1 → K , • e : S 1 → S 2 is an embedding of T 1 into T 2 , • f : K → S 2 is a continuous surjection such that for any p ∈ S 1 , f(e(p)) = e (p).
Furthermore, if F c ∈ Ult, then f is the identity map.
11 Lemma A2 in [16] showed that every concentric topological space is regular. Moreover, it is obvious that every concentric space has a linearly ordered base at each its point, i.e., is a lob-spaces (see, e.g., [6, p. 37]). Motivated by one of the reviewers, in Appendix A we prove that all regular lob-spaces are concentric. The result is not directly related to the problems discussed in the paper, yet it is novel and relevant for the theory of GCAs.
Proof. The proof is trivial if F c ∈ Ult, because then we have Gpt = Ult, and so f(U ) = U for any U ∈ Ult. In other cases, to show that f is continuous it is enough to prove that for any x ∈ R, the set f -1 [Irl(x)] is open in S R . We consider three cases.
Let x be neither finite nor cofinite. Then F c / ∈ Ult and Irl(x) = {↑ a | a ∈ At x }, because x / ∈ F c (see Lemma 7.5). Therefore Thanks to Lemma 11.1 and Propositions 11.2 and 11.3 we have the following theorem which is a stronger version of Theorem 6.33 from [14]: a familiar set-theoretical construction to which the initial structure can be reduced. Concerning this, the theorem does a good job, so to say. The representation theorems in the spirit of Stone [24], Düntsch & Winter [11] and Dimov & Vakarelov [8] were presented by us in our earlier works [16,18] where we produced object duality theorems for the class of Grzegorczyk contact algebras that satisfy an algebraic version of the countable chain condition. 13 In the context of this paper, we are only interested in the much narrower class of Grzegorczyk contact algebras. The results obtained, including the representation theorem above, broaden the understanding of Grzegorczyk pointfree topology, so far a largely neglected system of a region-based theory of space. §A. concentric topological spaces and lob-spaces. In this appendix, we prove that all regular lob-spaces are concentric, and in consequence, a topological space is concentric if and only if it is a regular lob-space.
Let T = S, O be a topological space. We call T a lob-space provided that for each p ∈ S there is a linearly ordered base B p at p (see, e.g., [6, definition 1.1]), i.e., B p satisfies the trichotomy condition: Furthermore, T is concentric iff it is a T 1 -space and for any p ∈ S there is a base B p at p such that: Obviously, all concentric spaces are lob-spaces. Now we prove that all regular lobspaces are concentric.

Theorem A.1 A topological space is concentric if and only if it is a regular lob-space.
Proof. For the "⇒"-part see Lemma A2 in [16]. For the "⇐"-part suppose that T = S, O is a regular lob-space and p ∈ S. Then T is a T 1 -space and there is a base B p at p satisfying the trichotomy condition and such that for any B ∈ B p there is a U ∈ O such that p ∈ U ⊆ Cl U ⊆ B.
For some ordinal α, let W | < α be a coinitial subset of B p , ⊆ such that for all < < α we have W W . 14 For every < α there is a such that < α and Cl W ⊆ W . Indeed, for W there is an open set U such that p ∈ U ⊆ Cl U ⊆ W . Yet W | < α is a base at p, so there is a < α such that W ⊆ U , and in consequence Cl W ⊆ W , so . Thus, there is a monotone coinitial subsequence of W | < α that satisfies (C), as required.