Preliminaries
Set-theoretic and order-theoretic conventions
If X X X is a set, we will denote its powerset by P X \mathcal{P}X P X .
A partially ordered set ( L , ≤ ) (L,\leq) ( L , ≤ ) is a lattice if all finite suprema and infima (denoted with the symbols ∨ \vee ∨ and ∧ \wedge ∧ respectively) exist. We will use ⋁ \bigvee ⋁ and ⋀ \bigwedge ⋀ for suprema and infima over subsets of L L L , if they exist.
As usual, if α ∈ L \alpha\in L α ∈ L , ↑ α = { β ∈ L : α ≤ β } \uparrow\alpha=\{\beta\in L:\alpha\leq\beta\} ↑ α = { β ∈ L : α ≤ β } and if A ⊂ L A\subset L A ⊂ L , ↑ A = ⋃ α ∈ A ↑ α \uparrow A=\bigcup_{\alpha\in A}\uparrow\alpha ↑ A = ⋃ α ∈ A ↑ α and ↓ α \downarrow\alpha ↓ α and ↓ A \downarrow A ↓ A are defined dually. A A A is called up-closed (respectively, down-closed) if A = ↑ A A=\uparrow A A =↑ A (respectively A = ↓ A A=\downarrow A A =↓ A ). We use interval notations when appropriate, so that, for instance, given α , β ∈ L \alpha,\beta\in L α , β ∈ L , the notation [ α , β ] [\alpha,\beta] [ α , β ] means ↑ α ∩ ↓ β \uparrow\alpha\cap\downarrow\beta ↑ α ∩ ↓ β . If L L L has a least element it is denoted by ⊥ L \bot_L ⊥ L and if it has a greatest element, it is denoted ⊤ L \top_L ⊤ L .
A non-empty subset F \mathcal{F} F of L L L is a filter if it is up-closed and closed under finite infima, that is, a , b ∈ F a,b\in\mathcal{F} a , b ∈ F implies a ∧ b ∈ F a\wedge b\in\mathcal{F} a ∧ b ∈ F . A filter F \mathcal{F} F is proper if it is a proper subset of L L L . Note that given ℓ ∈ L \ell\in L ℓ ∈ L , ↑ ℓ \uparrow\ell ↑ ℓ is a filter, called principal filter of ℓ \ell ℓ . Let F ‾ L \overline{\mathcal{F}}_L F L denote the set of filters on L L L , while F L \mathcal{F}_L F L denotes the set of proper filters on L L L . Hence the set of traditional set-theoretic filters on X X X is F P X \mathcal{F}_{\mathcal{P}X} F P X , where P X \mathcal{P}X P X is ordered by inclusion. To make it easier to identify set-theoretic filters, if A ⊂ X A\subset X A ⊂ X , we will write A ↑ A^\uparrow A ↑ for ↑ A \uparrow A ↑ A performed in P X \mathcal{P}X P X , that is, { B ⊂ X : A ⊂ B } \{B\subset X:A\subset B\} { B ⊂ X : A ⊂ B } .
We say that two subsets A , B A,B A , B of a lattice L L L mesh, in symbols A # B A\mathbin{\#}B A # B , if a ∧ b ≠ ⊥ L a\wedge b\neq\bot_L a ∧ b = ⊥ L for every a ∈ A a\in A a ∈ A and b ∈ B b\in B b ∈ B . Moreover, we will use the notation
A # = { ℓ ∈ L : { ℓ } # A } . A^\#=\{\ell\in L:\{\ell\}\mathbin{\#}A\}. A # = { ℓ ∈ L : { ℓ } # A } .
2020 Mathematics Subject Classification: 54A20, 18F99.
Key words and phrases. approach space, convergence space, convergence approach space, pointfree, compactness, continuity.
In particular, A , B ⊂ P X \mathcal{A},\mathcal{B} \subset\mathcal{P}X A , B ⊂ P X mesh if A ∩ B ≠ ∅ A \cap B \ne\varnothing A ∩ B = ∅ for every A ∈ A A \in\mathcal{A} A ∈ A and B ∈ B B \in\mathcal{B} B ∈ B , and then A # = { H ⊂ X : ∀ A ∈ A , A ∩ H ≠ ∅ } \mathcal{A}^{\#} = \{H \subset X : \forall A \in\mathcal{A}, A \cap H \ne\varnothing\} A # = { H ⊂ X : ∀ A ∈ A , A ∩ H = ∅ } .
A filter F \mathcal{F} F on a lattice L L L is prime if a ∈ F a \in\mathcal{F} a ∈ F or b ∈ F b \in\mathcal{F} b ∈ F whenever a ∨ b ∈ F a \vee b \in\mathcal{F} a ∨ b ∈ F and completely prime if there is a ∈ A ∩ F a \in\mathcal{A} \cap\mathcal{F} a ∈ A ∩ F whenever ⋁ A ∈ F \bigvee\mathcal{A} \in\mathcal{F} ⋁ A ∈ F . On the other hand, an element ℓ \ell ℓ of a lattice L L L is meet-prime if a ∧ b ≤ ℓ a \wedge b \le\ell a ∧ b ≤ ℓ implies that either a ≤ ℓ a \le\ell a ≤ ℓ or b ≤ ℓ b \le\ell b ≤ ℓ and join-prime if ℓ ≤ a ∨ b \ell\le a \vee b ℓ ≤ a ∨ b implies that either ℓ ≤ a \ell\le a ℓ ≤ a or ℓ ≤ b \ell\le b ℓ ≤ b . Maximal proper filters are also called ultrafilters and we denote by U L \mathbb{U}L U L the set of ultrafilters on L L L . In particular, U P X \mathbb{U}\mathcal{P}X U P X denotes the set of set-theoretic ultrafilters on X X X . If F ∈ F L \mathcal{F} \in\mathbb{F}L F ∈ F L , we write β ( F ) = { U ∈ U L : F ⊂ U } \beta(\mathcal{F}) = \{ \mathcal{U} \in\mathbb{U}L : \mathcal{F} \subset\mathcal{U} \} β ( F ) = { U ∈ U L : F ⊂ U } .
Convergence spaces
A convergence ξ \xi ξ on a set X X X is a relation between points of X X X and (set-theoretic) filters on X X X , denoted x ∈ lim ξ F x \in\lim_{\xi} \mathcal{F} x ∈ lim ξ F if ( x , F ) ∈ ξ (x,\mathcal{F}) \in\xi ( x , F ) ∈ ξ and interpreted as x x x is a limit point for F \mathcal{F} F in ξ \xi ξ , satisfying
(centered) x ∈ lim ξ { x } ↑ x \in\lim_{\xi} \{x\}^{\uparrow} x ∈ lim ξ { x } ↑
for every x ∈ X x \in X x ∈ X and
(monotone) F ≤ G ⟹ lim ξ F ⊂ lim ξ G \mathcal{F} \le\mathcal{G} \Longrightarrow\lim_{\xi} \mathcal{F} \subset\lim_{\xi} \mathcal{G} F ≤ G ⟹ lim ξ F ⊂ lim ξ G ,
for all F , G ∈ F P X \mathcal{F},\mathcal{G} \in\mathbb{F}\mathcal{P}X F , G ∈ F P X . The pair ( X , ξ ) (X,\xi) ( X , ξ ) is then called a convergence space .
A map f : X → Y f : X \to Y f : X → Y between two convergence spaces ( X , ξ ) (X,\xi) ( X , ξ ) and ( Y , τ ) (Y,\tau) ( Y , τ ) is continuous , in symbols f ∈ C ( ξ , τ ) f \in\mathcal{C}(\xi,\tau) f ∈ C ( ξ , τ ) , if f ( x ) ∈ lim τ f [ F ] f(x) \in\lim_{\tau} f[\mathcal{F}] f ( x ) ∈ lim τ f [ F ] whenever x ∈ lim ξ F x \in\lim_{\xi} \mathcal{F} x ∈ lim ξ F , where f [ F ] = { f ( F ) : F ∈ F } ↑ f[\mathcal{F}] = \{f(F) : F \in\mathcal{F}\}^{\uparrow} f [ F ] = { f ( F ) : F ∈ F } ↑ is the image filter. Let Conv denote the category of convergence spaces and continuous maps. If ξ , σ \xi,\sigma ξ , σ are two convergences on X X X , we say that ξ \xi ξ is finer than σ \sigma σ or that σ \sigma σ is coarser than ξ \xi ξ if the identity map id X ∈ C ( ξ , σ ) \operatorname{id}_{X} \in\mathcal{C}(\xi,\sigma) id X ∈ C ( ξ , σ ) , that is, if lim ξ F ⊂ lim σ F \lim_{\xi} \mathcal{F} \subset\lim_{\sigma} \mathcal{F} lim ξ F ⊂ lim σ F for every filter F \mathcal{F} F on X X X .
If a convergence additionally satisfies
(finite depth) lim ξ ( F ∩ G ) = lim ξ F ∩ lim ξ G \lim_{\xi}(\mathcal{F} \cap\mathcal{G}) = \lim_{\xi} \mathcal{F} \cap\lim_{\xi} \mathcal{G} lim ξ ( F ∩ G ) = lim ξ F ∩ lim ξ G
for every F , G ∈ F P X \mathcal{F},\mathcal{G} \in\mathbb{F}\mathcal{P}X F , G ∈ F P X , we say that ξ \xi ξ has finite depth .1
A convergence satisfying the stronger condition that
(pretopology) lim ξ ( ⋂ D ∈ D D ) = ⋂ D ∈ D lim ξ D \lim_{\xi}\left(\bigcap_{\mathcal{D}\in\mathbb{D}} \mathcal{D}\right) = \bigcap_{\mathcal{D}\in\mathbb{D}} \lim_{\xi} \mathcal{D} lim ξ ( ⋂ D ∈ D D ) = ⋂ D ∈ D lim ξ D
for every D ⊂ F P X \mathbb{D} \subset\mathbb{F}\mathcal{P}X D ⊂ F P X is called a pretopology (and ( X , ξ ) (X,\xi) ( X , ξ ) is a pretopological space ).
A subset A A A of a convergence space ( X , ξ ) (X,\xi) ( X , ξ ) is ξ \xi ξ -open if
lim ξ F ∩ A ≠ ∅ ⟹ A ∈ F , \lim_{\xi} \mathcal{F} \cap A \ne\varnothing\Longrightarrow A \in\mathcal{F}, lim ξ F ∩ A = ∅ ⟹ A ∈ F ,
and ξ \xi ξ -closed if it is closed for limits, that is,
A ∈ F ⟹ lim ξ F ⊂ A , A \in\mathcal{F} \Longrightarrow\lim_{\xi} \mathcal{F} \subset A, A ∈ F ⟹ lim ξ F ⊂ A ,
equivalently,
A ∈ F # ⟹ lim ξ F ⊂ A . A \in\mathcal{F}^{\#} \Longrightarrow\lim_{\xi} \mathcal{F} \subset A. A ∈ F # ⟹ lim ξ F ⊂ A .
Let O ξ \mathcal{O}_{\xi} O ξ denote the set of open subsets of ( X , ξ ) (X,\xi) ( X , ξ ) and let O ξ ( x ) = { U ∈ O ξ : x ∈ U } \mathcal{O}_{\xi}(x) = \{U \in\mathcal{O}_{\xi} : x \in U\} O ξ ( x ) = { U ∈ O ξ : x ∈ U } . Similarly, let C ξ \mathcal{C}_{\xi} C ξ denote the set of closed subsets of ( X , ξ ) (X,\xi) ( X , ξ ) . It turns out that O ξ \mathcal{O}_{\xi} O ξ is a topology on X X X . Moreover, a topology τ \tau τ on a set X X X determines a convergence ξ τ \xi_{\tau} ξ τ on X X X by
x ∈ lim ξ τ F ⟺ F ≥ N τ ( x ) , x \in\lim_{\xi_{\tau}} \mathcal{F} \Longleftrightarrow\mathcal{F} \ge\mathcal{N}_{\tau}(x), x ∈ lim ξ τ F ⟺ F ≥ N τ ( x ) ,
where N τ ( x ) \mathcal{N}_{\tau}(x) N τ ( x ) denotes the neighborhood filter of x x x for τ \tau τ . In turn, ξ τ \xi_{\tau} ξ τ completely determines τ \tau τ because τ = O ξ τ \tau= \mathcal{O}_{\xi_{\tau}} τ = O ξ τ , so that we do not distinguish between τ \tau τ and ξ τ \xi_{\tau} ξ τ . and identify topologies with special convergences. Moreover, a convergence ξ \xi ξ on X X X determines the topology O ξ \mathcal{O}_{\xi} O ξ on X X X which turns out to be the finest among the topologies on X X X that are coarser than ξ \xi ξ . We call it the topological modification of ξ \xi ξ and denote it T ξ \mathrm{T}\xi T ξ . Hence the category Top of topological spaces and continuous maps is a concretely reflective subcategory of Conv. Let cl ξ \operatorname{cl}_{\xi} cl ξ denote the closure operator in T ξ \mathrm{T}\xi T ξ .
A convergence ξ \xi ξ determines an adherence operator on filters, namely,
adh F = ⋃ G # F lim ξ G = ⋃ G ∈ β ( F ) lim ξ G = ⋃ U ∈ β ( F ) lim ξ U . (1) \operatorname{adh}\mathcal{F}=\bigcup_{\mathcal{G}\mathbin{\#}\mathcal{F}}\lim_{\xi}\mathcal{G}=\bigcup_{\mathcal{G}\in\beta(\mathcal{F})}\lim_{\xi}\mathcal{G}=\bigcup_{\mathcal{U}\in\beta(\mathcal{F})}\lim_{\xi}\mathcal{U}. \tag*{(1)} adh F = G # F ⋃ ξ lim G = G ∈ β ( F ) ⋃ ξ lim G = U ∈ β ( F ) ⋃ ξ lim U . (1)
A convergence ξ \xi ξ on X X X is a pseudotopology if
lim ξ F = ⋂ U ∈ β ( F ) lim ξ U \lim_{\xi}\mathcal{F}=\bigcap_{\mathcal{U}\in\beta(\mathcal{F})}\lim_{\xi}\mathcal{U} ξ lim F = U ∈ β ( F ) ⋂ ξ lim U
for every filter F ∈ F P X \mathcal{F}\in\mathbb{F}PX F ∈ F P X . The full subcategory of Conv formed by pseudotopological spaces (and continuous maps), denoted PsTop, is concretely reflective and the reflector S can be described on objects by
lim S ξ F = ⋂ U ∈ β ( F ) lim ξ U = ⋂ G # F adh ξ G , (2) \lim_{S\xi}\mathcal{F}=\bigcap_{\mathcal{U}\in\beta(\mathcal{F})}\lim_{\xi}\mathcal{U}=\bigcap_{\mathcal{G}\mathbin{\#}\mathcal{F}}\operatorname{adh}_{\xi}\mathcal{G}, \tag*{(2)} S ξ lim F = U ∈ β ( F ) ⋂ ξ lim U = G # F ⋂ adh ξ G , (2)
so that adherences determine the pseudotopological reflection.
When restricted to principal filters, (1.1) defines a principal adherence operator adh ξ : P X → P X \operatorname{adh}_{\xi}:PX\to PX adh ξ : P X → P X given by
adh ξ A = adh ξ { A } ↑ = ⋃ A ∈ G # lim ξ G = ⋃ A ∈ G lim ξ G = ⋃ U ∈ β A lim ξ U . \operatorname{adh}_{\xi}A=\operatorname{adh}_{\xi}\{A\}^{\uparrow}=\bigcup_{A\in\mathcal{G}^{\#}}\lim_{\xi}\mathcal{G}=\bigcup_{A\in\mathcal{G}}\lim_{\xi}\mathcal{G}=\bigcup_{\mathcal{U}\in\beta A}\lim_{\xi}\mathcal{U}. adh ξ A = adh ξ { A } ↑ = A ∈ G # ⋃ ξ lim G = A ∈ G ⋃ ξ lim G = U ∈ β A ⋃ ξ lim U .
The full subcategory of Conv formed by pretopological spaces (and continuous maps), denoted PrTop, is concretely reflective and the reflector S 0 S_{0} S 0 can be described on objects by
lim S 0 ξ F = ⋂ A ∈ F # adh ξ A , (3) \lim_{S_{0}\xi}\mathcal{F}=\bigcap_{A\in\mathcal{F}^{\#}}\operatorname{adh}_{\xi}A, \tag*{(3)} S 0 ξ lim F = A ∈ F # ⋂ adh ξ A , (3)
so that the principal adherence determines the pretopological reflection. In general,
adh ξ A ⊂ cl ξ A \operatorname{adh}_{\xi}A\subset\operatorname{cl}_{\xi}A adh ξ A ⊂ cl ξ A
but not conversely. In contrast to cl ξ \operatorname{cl}_{\xi} cl ξ , the principal adherence is in general non-idempotent because adh ξ A \operatorname{adh}_{\xi}A adh ξ A need not be closed. A pretopology is a topology if and only if its principal adherence operator is idempotent, in which case adh ξ = cl ξ \operatorname{adh}_{\xi}=\operatorname{cl}_{\xi} adh ξ = cl ξ . Moreover, the reflector T from Conv onto Top is given by
lim T ξ F = ⋂ A ∈ F # cl ξ A . (4) \lim_{\mathrm{T}\xi}\mathcal{F}=\bigcap_{A\in\mathcal{F}^{\#}}\operatorname{cl}_{\xi}A. \tag*{(4)} T ξ lim F = A ∈ F # ⋂ cl ξ A . (4)
1.2.1. Compactness for filters in Conv. As we will discuss compactness in Section § 9 , let us note that a convergence space is compact if every ultrafilter converges, equivalently if the adherence of every filter is non-empty. Generalizations in several directions have proved to be unifying notions in the context of convergence spaces. Given a class D \mathbb{D} D of filters 2 , a filter F \mathcal{F} F on a convergence space ( X , ξ ) (X,\xi) ( X , ξ ) and A ⊂ P X \mathcal{A}\subset PX A ⊂ P X , we say that F \mathcal{F} F is D \mathbb{D} D -compact at A A A if
D # F ⇒ adh ξ D # A , \mathbb{D}\#\mathcal{F} \Rightarrow\operatorname{adh}_{\xi}\mathbb{D}\#\mathcal{A}, D # F ⇒ adh ξ D # A ,
for every D ∈ D X \mathcal{D} \in\mathbb{D}_X D ∈ D X and nearly D \mathbb{D} D -compact at A A A if
D ≥ F ⇒ adh ξ D # A , \mathcal{D} \geq\mathcal{F} \Rightarrow\operatorname{adh}_{\xi}\mathbb{D}\#\mathcal{A}, D ≥ F ⇒ adh ξ D # A ,
for every D ∈ D X \mathcal{D} \in\mathbb{D}_X D ∈ D X . To say that F ⊂ X F \subset X F ⊂ X is compact (resp. countably compact, Lindelöf) is to say that the principal filter { F } ↑ \{F\}^{\uparrow} { F } ↑ is D \mathbb{D} D -compact at itself, that is, at { F } \{F\} { F } (equivalently, nearly D \mathbb{D} D -compact at { F } \{F\} { F } ), where D \mathbb{D} D is the class F \mathbb{F} F of all filters (resp., the class F 1 \mathbb{F}_1 F 1 of countably based filters, the class F ∧ 1 \mathbb{F}_{\wedge1} F ∧ 1 of countably complete filters). It is relatively compact if { F } ↑ \{F\}^{\uparrow} { F } ↑ is compact at { X } \{X\} { X } . Accordingly, we omit the prefix F \mathbb{F} F - when D = F \mathbb{D}=\mathbb{F} D = F . Note that x ∈ lim ξ F x \in\lim_{\xi}\mathcal{F} x ∈ lim ξ F if and only if F \mathcal{F} F is compact at { x } \{x\} { x } and x ∈ lim S 0 F x \in\lim_{\mathbb{S}_0}\mathcal{F} x ∈ lim S 0 F if and only if F \mathcal{F} F is F 0 \mathbb{F}_0 F 0 -compact at { x } \{x\} { x } . See [4 , 12 , 13 ] for an account of the unifying power of these notions.
Let’s say that a class D \mathbb{D} D of filters is transferable if D ∈ D Y \mathbb{D} \in\mathbb{D}_Y D ∈ D Y implies f − [ D ] ∈ D X f^{-}[\mathbb{D}] \in\mathbb{D}_X f − [ D ] ∈ D X for every f : X → Y f:X\to Y f : X → Y . Note that the following is a simple generalization of the fact that the continuous image of a compact set is compact (see, e.g., [5 Proposition IX.8.2, Corollary XV.3.3] ):
Proposition 1. If D \mathbb{D} D is a transferable class of filters, ( X , ξ ) (X,\xi) ( X , ξ ) and ( Y , τ ) (Y,\tau) ( Y , τ ) are convergence spaces, F \mathcal{F} F is a D \mathbb{D} D -compact filter at A ⊂ X A\subset X A ⊂ X and f ∈ C ( ξ , τ ) f\in\mathcal{C}(\xi,\tau) f ∈ C ( ξ , τ ) then f [ F ] f[\mathcal{F}] f [ F ] is D \mathbb{D} D -compact at f ( A ) f(A) f ( A ) .
Remark 2. In particular, applied to principal filters, this yields the classical fact that compactness, countable compactness and the Lindelöf property are preserved under continuous maps. Applied to filters and A = { x } A=\{x\} A = { x } , this yields that S \mathbb{S} S , S 0 \mathbb{S}_0 S 0 and other similarly defined maps preserve continuity, that is, are functors. Though both sets of results are easy to obtain, it is interesting to see them as two sides of the same coin, under the umbrella of Proposition 1.
We refer the reader to [5 ] for a systematic study of convergence spaces.
Pointfree convergence spaces
Note that (monotone) states that lim ξ : F P X → P X \lim_{\xi}:\mathbb{F}P X\to P X lim ξ : F P X → P X is monotone, which allows for a pointfree abstraction of Conv, after observing that given two convergence spaces ( X , ξ ) (X,\xi) ( X , ξ ) and ( Y , τ ) (Y,\tau) ( Y , τ ) , the condition f ∈ C ( ξ , τ ) f\in\mathcal{C}(\xi,\tau) f ∈ C ( ξ , τ ) rephrases as f ( lim ξ F ) ⊂ lim τ f [ F ] f(\lim_{\xi}\mathcal{F})\subset\lim_{\tau}f[\mathcal{F}] f ( lim ξ F ) ⊂ lim τ f [ F ] , equivalently,
lim ξ F ⊂ φ ( lim τ φ − 1 [ F ] ) . \lim_{\xi}\mathcal{F}\subset\varphi(\lim_{\tau}\varphi^{-1}[\mathcal{F}]). ξ lim F ⊂ φ ( τ lim φ − 1 [ F ]) .
where φ = f − 1 : P Y → P X \varphi=f^{-1}:P Y\to P X φ = f − 1 : P Y → P X is a morphism of lattices.
Definition 3. [6 ] Given a category C of lattices, a convergence C-object ( L , lim ) (L,\lim) ( L , lim ) is a C-object L L L together with a monotone map lim : F L → L \lim:\mathbb{F}L\to L lim : F L → L . The objects of the category C C o n v \mathrm{C}^{\mathrm{Conv}} C Conv are the convergence C-objects and the morphisms φ : L → L ′ \varphi:L\to L' φ : L → L ′ are the C-morphisms that are continuous in the sense that for every F ∈ F L ′ \mathcal{F}\in\mathbb{F}L' F ∈ F L ′ ,
(ptfree continuity)
lim L ′ F ≤ φ ( lim L φ − 1 ( F ) ) . \lim_{L'}\mathcal{F}\leq\varphi(\lim_L\varphi^{-1}(\mathcal{F})). L ′ lim F ≤ φ ( L lim φ − 1 ( F )) .
where φ − 1 ( F ) = { ℓ ∈ L : φ ( ℓ ) ∈ F } \varphi^{-1}(\mathcal{F})=\{\ell\in L:\varphi(\ell)\in\mathcal{F}\} φ − 1 ( F ) = { ℓ ∈ L : φ ( ℓ ) ∈ F } .
If C denotes a category C o p \mathrm{C}^{\mathrm{op}} C op denotes its opposite category. The category Conv embeds coreflectively into ( C C o n v ) o p (\mathrm{C}^{\mathrm{Conv}})^{\mathrm{op}} ( C Conv ) op when C is the category of frames or of coframes: the powerset functor P : C o n v → ( C C o n v ) o p \mathbb{P}:\mathrm{Conv}\to(\mathrm{C}^{\mathrm{Conv}})^{\mathrm{op}} P : Conv → ( C Conv ) op sending ( X , ξ ) (X,\xi) ( X , ξ ) to ( P X , lim ξ ) (\mathbb{P}X,\lim_{\xi}) ( P X , lim ξ ) and f : ( X , ξ ) → ( Y , τ ) f:(X,\xi)\to(Y,\tau) f : ( X , ξ ) → ( Y , τ ) to P f = f − 1 : P Y → P X \mathbb{P}f=f^{-1}:\mathbb{P}Y\to\mathbb{P}X P f = f − 1 : P Y → P X (in C C o n v \mathrm{C}^{\mathrm{Conv}} C Conv ) is then right-adjoint to the point-functor p t : ( C C o n v ) o p → C o n v \mathrm{pt}:(\mathrm{C}^{\mathrm{Conv}})^{\mathrm{op}}\to\mathrm{Conv} pt : ( C Conv ) op → Conv (the coreflector) where the underlying set of p t L pt\,L pt L , the set of “points” of L L L , is the set of C C o n v \mathsf{C}^{\mathrm{Conv}} C Conv -morphisms from L L L to P ( { ∗ } ) \mathcal{P}(\{\ast\}) P ({ ∗ }) , hence depends on the choice of C \mathsf{C} C . The convergence structure on p t L pt\,L pt L is given by
( C o n v o n p t ) lim p t L F = ( lim L F ∘ ) ∙ . (\mathrm{Conv\ on\ }pt)\qquad\lim_{pt\,L}\mathcal{F}=\left(\lim_{L}\mathcal{F}^{\circ}\right)^{\bullet}. ( Conv on pt ) pt L lim F = ( L lim F ∘ ) ∙ .
where ℓ ∙ = { φ ∈ p t L : φ ( ℓ ) = { ∗ } } \ell^{\bullet}=\{\varphi\in pt\,L:\varphi(\ell)=\{\ast\}\} ℓ ∙ = { φ ∈ pt L : φ ( ℓ ) = { ∗ }} and F ∘ = { ℓ ∈ L : ℓ ∙ ∈ F } \mathcal{F}^{\circ}=\{\ell\in L:\ell^{\bullet}\in\mathcal{F}\} F ∘ = { ℓ ∈ L : ℓ ∙ ∈ F } . Finally, if φ ∈ C C o n v ( L , L ′ ) \varphi\in\mathsf{C}^{\mathrm{Conv}}(L,L^{\prime}) φ ∈ C Conv ( L , L ′ ) then p t φ : p t L ′ → p t L pt\,\varphi:pt\,L^{\prime}\to pt\,L pt φ : pt L ′ → pt L is defined by p t ( φ ) ( f ) = f ∘ φ pt(\varphi)(f)=f\circ\varphi pt ( φ ) ( f ) = f ∘ φ .
Remark 4. Identifying a point φ ∈ p t L \varphi\in pt\,L φ ∈ pt L with the filter F = φ − 1 ( { ∗ } ) ∈ F L \mathcal{F}=\varphi^{-1}(\{\ast\})\in\mathbb{F}L F = φ − 1 ({ ∗ }) ∈ F L , points can be seen as special filters on L L L . The continuity condition (ptfree continuity) on φ \varphi φ yields lim F ∈ F \lim\mathcal{F}\in\mathcal{F} lim F ∈ F , while the fact that φ \varphi φ is a C \mathsf{C} C -morphism adds another condition. In particular, if C \mathsf{C} C -morphisms preserves arbitrary suprema, for instance in the category of frames, then F \mathcal{F} F is a completely prime filter. If C \mathsf{C} C -morphisms preserves arbitrary infima, for instance in the category of coframes, then F \mathcal{F} F is the principal filter of a join-prime element. Note that for these two cases, the only such filters in P X \mathcal{P}X P X are principal ultrafilters of singletons, so that the underlying set of p t P X pt\,\mathcal{P}X pt P X can be identified with X X X for such base categories C \mathsf{C} C , since x ∈ lim { x } ↑ x\in\lim\{x\}^{\uparrow} x ∈ lim { x } ↑ for every x ∈ X x\in X x ∈ X . See [6 ] and [15 ] for details.
Complete Heyting algebras and grills
Let us focus on the case where C \mathsf{C} C is the category C H \mathsf{CH} CH of complete Heyting algebras. The objects of Heyt are frames, i.e., complete lattices L L L with
a ∧ ⋁ b ∈ B b = ⋁ b ∈ B ( a ∧ b ) (5) a\wedge\bigvee_{b\in B}b=\bigvee_{b\in B}(a\wedge b) \tag*{(5)} a ∧ b ∈ B ⋁ b = b ∈ B ⋁ ( a ∧ b ) (5)
for every a ∈ L a\in L a ∈ L and B ⊂ L B\subset L B ⊂ L . As a ∧ ⋅ a\wedge\cdot a ∧ ⋅ preserves all suprema, it has a right adjoint a → ⋅ a\to\cdot a → ⋅ , so that
a ∧ x ≤ y ⟺ x ≤ a → y (6) a\wedge x\le y\Longleftrightarrow x\le a\to y \tag*{(6)} a ∧ x ≤ y ⟺ x ≤ a → y (6)
for every a , x , y ∈ L a,x,y\in L a , x , y ∈ L . The element a ∗ : = a → 0 a^{*}:=a\to0 a ∗ := a → 0 of L L L is called the pseudocomplement of a a a . The morphisms of C H \mathsf{CH} CH preserve arbitrary joins and meets, and the implication → \to → . Note that a morphism φ ∈ C H ( L , L ′ ) \varphi\in\mathsf{CH}(L,L^{\prime}) φ ∈ CH ( L , L ′ ) preserves all infima (in particular, φ ( 1 L ) = 1 L ′ \varphi(1_{L})=1_{L^{\prime}} φ ( 1 L ) = 1 L ′ ) and therefore admits a left-adjoint φ ℓ : L ′ → L \varphi_{\ell}:L^{\prime}\to L φ ℓ : L ′ → L , and preserves all suprema (in particular φ ( 0 L ) = 0 L ′ \varphi(0_{L})=0_{L^{\prime}} φ ( 0 L ) = 0 L ′ ) and therefore admits a right-adjoint φ r : L ′ → L \varphi_{r}:L^{\prime}\to L φ r : L ′ → L so that
φ ℓ ( m ′ ) ≤ m ⟺ m ′ ≤ φ ( m ) , (7) \varphi_{\ell}(m^{\prime})\le m\Longleftrightarrow m^{\prime}\le\varphi(m), \tag*{(7)} φ ℓ ( m ′ ) ≤ m ⟺ m ′ ≤ φ ( m ) , (7)
and
φ ( m ) ≤ m ′ ⟺ m ≤ φ r ( m ′ ) , (8) \varphi(m)\le m^{\prime}\Longleftrightarrow m\le\varphi_{r}(m^{\prime}), \tag*{(8)} φ ( m ) ≤ m ′ ⟺ m ≤ φ r ( m ′ ) , (8)
whenever m ∈ L m\in L m ∈ L and m ′ ∈ L ′ m^{\prime}\in L^{\prime} m ′ ∈ L ′ .
In this context, we say that two subsets A A A and B B B of a complete lattice L L L mesh, in symbols A # B A\#B A # B , if a ∧ b ≠ 0 L a\wedge b\ne0_{L} a ∧ b = 0 L whenever a ∈ A a\in A a ∈ A and b ∈ B b\in B b ∈ B . We will also consider the grill of A ⊂ L A\subset L A ⊂ L given by A # = { ℓ ∈ L : { ℓ } # A } A^{\#}=\{\ell\in L:\{\ell\}\#A\} A # = { ℓ ∈ L : { ℓ } # A } . As in a frame a ∧ b = 0 a\wedge b=0 a ∧ b = 0 if and only if b ≤ a ∗ b\le a^{*} b ≤ a ∗ , we have that
ℓ ∈ A # ⟺ ℓ ∗ ∉ ↑ A (9) \ell\in A^{\#}\Longleftrightarrow\ell^{*}\notin\uparrow A \tag*{(9)} ℓ ∈ A # ⟺ ℓ ∗ ∈ / ↑ A (9)
for every A ⊂ L A\subset L A ⊂ L if L L L is a frame.
Lemma 5. If L L L is a distributive lattice, H ∈ F L \mathcal{H}\in\mathbb{F}L H ∈ F L and A , B A,B A , B are two upsets in L L L then
H # ( A ∩ B ) ⟺ H # A or H # B . \mathcal{H}\#(A\cap B)\Longleftrightarrow\mathcal{H}\#A\ \text{or}\ \mathcal{H}\#B. H # ( A ∩ B ) ⟺ H # A or H # B .
Proof. If A ⊈ H # A \not\subseteq H^{\#} A ⊆ H # and B ⊈ H # B \not\subseteq H^{\#} B ⊆ H # there is a ∈ A a \in A a ∈ A and b ∈ B b \in B b ∈ B with a ∉ H # a \notin H^{\#} a ∈ / H # and b ∉ H # b \notin H^{\#} b ∈ / H # , that is, there is h 1 , h 2 ∈ H h_{1},h_{2} \in H h 1 , h 2 ∈ H with a ∧ h 1 = b ∧ h 2 = 0 a \wedge h_{1}=b \wedge h_{2}=0 a ∧ h 1 = b ∧ h 2 = 0 . Moreover, a ∨ b ∈ A ∩ B a \vee b \in A \cap B a ∨ b ∈ A ∩ B because A A A and B B B are upsets, and h = h 1 ∧ h 2 ∈ H h=h_{1} \wedge h_{2} \in H h = h 1 ∧ h 2 ∈ H because H H H is a filter. As
h ∧ ( a ∨ b ) = ( h ∧ a ) ∨ ( h ∧ b ) ≤ ( h 1 ∧ a ) ∨ ( h 2 ∧ b ) = 0 , h \wedge(a \vee b)=(h \wedge a) \vee(h \wedge b) \leq(h_{1} \wedge a) \vee(h_{2} \wedge b)=0, h ∧ ( a ∨ b ) = ( h ∧ a ) ∨ ( h ∧ b ) ≤ ( h 1 ∧ a ) ∨ ( h 2 ∧ b ) = 0 ,
we conclude that A ∩ B ⊈ H # A \cap B \not\subseteq H^{\#} A ∩ B ⊆ H # . The reverse implication is clear. □
Moreover, [6 Lemma 6.19] states (in slightly different terms, but equivalently) that if φ : L → L ′ \varphi:L \to L^{\prime} φ : L → L ′ is a morphism of coframes, M ⊆ L M \subseteq L M ⊆ L and N ⊆ L ′ N \subseteq L^{\prime} N ⊆ L ′ then
φ ( M ) # N ⟹ M # φ ℓ ( N ) . (10) \varphi(M)\mathbin{\#}N \Longrightarrow M\mathbin{\#}\varphi_{\ell}(N). \tag*{(10)} φ ( M ) # N ⟹ M # φ ℓ ( N ) . (10)
It turns out that the converse is true for morphisms of complete Heyting algebras. Indeed, in the category C H \mathrm{CH} CH , morphisms preserves implications, so that
φ ∈ C H ( L , L ′ ) ⟹ φ ( ℓ ∗ ) = ( φ ( ℓ ) ) ∗ , (11) \varphi\in\mathrm{CH}(L,L^{\prime}) \Longrightarrow\varphi(\ell^{*})=(\varphi(\ell))^{*}, \tag*{(11)} φ ∈ CH ( L , L ′ ) ⟹ φ ( ℓ ∗ ) = ( φ ( ℓ ) ) ∗ , (11)
for every ℓ ∈ L \ell\in L ℓ ∈ L . As a result:
Lemma 6. Let φ ∈ C H ( L , L ′ ) \varphi\in\mathrm{CH}(L,L^{\prime}) φ ∈ CH ( L , L ′ ) , M ⊆ L M \subseteq L M ⊆ L and N ⊆ L ′ N \subseteq L^{\prime} N ⊆ L ′ . Then
φ ( M ) # N ⟺ M # φ ℓ ( N ) . (12) \varphi(M)\mathbin{\#}N \Longleftrightarrow M\mathbin{\#}\varphi_{\ell}(N). \tag*{(12)} φ ( M ) # N ⟺ M # φ ℓ ( N ) . (12) Proof. Let m ∈ M m \in M m ∈ M and n ∈ N n \in N n ∈ N . Then
φ ( m ) ∧ n = 0 L ′ ⟺ n ≤ ( φ ( m ) ) ∗ = φ ( m ∗ ) ⟺ φ ℓ ( n ) ≤ m ∗ ⟺ φ ℓ ( n ) ∧ m = 0 L . \varphi(m)\wedge n=0_{L^{\prime}} \Longleftrightarrow n \leq(\varphi(m))^{*}=\varphi(m^{*}) \Longleftrightarrow\varphi_{\ell}(n) \leq m^{*} \Longleftrightarrow\varphi_{\ell}(n)\wedge m=0_{L}. φ ( m ) ∧ n = 0 L ′ ⟺ n ≤ ( φ ( m ) ) ∗ = φ ( m ∗ ) ⟺ φ ℓ ( n ) ≤ m ∗ ⟺ φ ℓ ( n ) ∧ m = 0 L .
Consider now C H C o n v \mathrm{CH}^{\mathrm{Conv}} CH Conv . Note that P : C o n v → ( C H C o n v ) o p \mathbb{P}:\mathrm{Conv}\to(\mathrm{CH}^{\mathrm{Conv}})^{\mathrm{op}} P : Conv → ( CH Conv ) op is still a functor as P f = f − 1 : P Y → P X \mathbb{P}f=f^{-1}:\mathbb{P}Y\to\mathbb{P}X P f = f − 1 : P Y → P X preserves arbitrary unions, intersections, and complements, hence the implication A → B = ( X ∖ A ) ∪ B A\to B=(X\setminus A)\cup B A → B = ( X ∖ A ) ∪ B . On the other hand, if L ∈ Ob ( C H C o n v ) L\in\operatorname{Ob}(\mathrm{CH}^{\mathrm{Conv}}) L ∈ Ob ( CH Conv ) then pt L \operatorname{pt}L pt L can be identified with filters on L L L that are principal filters of completely join-prime (or compact) elements such that either a a a or a ∗ a^{*} a ∗ is in the filter, whenever a ∈ L a\in L a ∈ L . In particular, such filters in P X \mathbb{P}X P X are the principal filters of singletons, so that pt P X \operatorname{pt}\mathbb{P}X pt P X can be identified with X X X .
The quantale V V V
1.5. 1.5. 1.5. The quantale V V V . V V V will denote a complete and completely distributive lattice with top element 1 = ⊤ V 1=\top_{V} 1 = ⊤ V and bottom element 0 = ⊥ V 0=\bot_{V} 0 = ⊥ V , equipped with a quantale operation ⊗ : V × V → V \otimes:V\times V\to V ⊗ : V × V → V which is commutative and unital, with unit 1 1 1 (an integral quantale). Hence,
a ⊗ ⋁ b ∈ B b = ⋁ b ∈ B a ⊗ b (13) a\otimes\bigvee_{b\in B}b=\bigvee_{b\in B}a\otimes b \tag*{(13)} a ⊗ b ∈ B ⋁ b = b ∈ B ⋁ a ⊗ b (13)
for every a ∈ V a\in V a ∈ V and B ⊆ V B\subseteq V B ⊆ V , and
∀ a ∈ V , a ⊗ 1 = a . (14) \forall a\in V,\quad a\otimes1=a. \tag*{(14)} ∀ a ∈ V , a ⊗ 1 = a . (14)
We also require that 1 1 1 does not have a unique immediate predecessor, that is,
1 = ⋁ ϵ ∈ V ∖ { 1 } ϵ . (15) 1=\bigvee_{\epsilon\in V\setminus\{1\}}\epsilon. \tag*{(15)} 1 = ϵ ∈ V ∖ { 1 } ⋁ ϵ . (15)
Standard examples include V = [ 0 , 1 ] V=[0,1] V = [ 0 , 1 ] with ⊗ = × \otimes=\times ⊗ = × or ⊗ = ∧ \otimes=\wedge ⊗ = ∧ .
Remark 7. Note that − ln : ( [ 0 , 1 ] , × ) → ( [ 0 , ∞ ] , + ) o p -\ln:([0,1],\times)\to([0,\infty],+)^{\mathrm{op}} − ln : ([ 0 , 1 ] , × ) → ([ 0 , ∞ ] , + ) op is a quantale isomorphism, and thus the classical setting of approach spaces is covered, albeit with the opposite order on [ 0 , ∞ ] [0,\infty] [ 0 , ∞ ] .
We gather below straightforward facts about ( V , ⊗ ) (V,\otimes) ( V , ⊗ ) :
Lemma 8. Let ( V , ⊗ ) (V,\otimes) ( V , ⊗ ) be as above.
(1) a ⊗ v ≤ a ∧ v a\otimes v\le a\wedge v a ⊗ v ≤ a ∧ v for every a , v ∈ V a,v\in V a , v ∈ V ;
(2) a ⊗ 0 = 0 a\otimes0=0 a ⊗ 0 = 0 for every a ∈ V a\in V a ∈ V ;
(3) If a , b , c , d ∈ V a,b,c,d\in V a , b , c , d ∈ V then
a ≤ b and c ≤ d ⟹ a ⊗ c ≤ b ⊗ d . a\le b\text{ and }c\le d\mathrel{\Longrightarrow}a\otimes c\le b\otimes d. a ≤ b and c ≤ d ⟹ a ⊗ c ≤ b ⊗ d . (4) if v ∈ V v\in V v ∈ V then
v = ⋁ ε ∈ V ∖ { 1 } v ⊗ ε . v=\bigvee_{\varepsilon\in V\setminus\{1\}}v\otimes\varepsilon. v = ε ∈ V ∖ { 1 } ⋁ v ⊗ ε .
In ( V , ⊗ ) (V,\otimes) ( V , ⊗ ) , for each v ∈ V v\in V v ∈ V , the map v ⊗ − : V → V v\otimes-:V\to V v ⊗ − : V → V preserves all suprema, hence has a right adjoint − ⊘ v : V → V -\mathbin{\oslash}v:V\to V − ⊘ v : V → V defined by
v ⊗ x ≤ y ⟺ x ≤ y ⊘ v . (16) v\otimes x\le y\iff x\le y\mathbin{\oslash}v. \tag*{(16)} v ⊗ x ≤ y ⟺ x ≤ y ⊘ v . (16)
which preserves all infima, namely,
( ⋀ b ∈ B b ) ⊘ v = ⋀ b ∈ B ( b ⊘ v ) . (17) \left(\bigwedge_{b\in B}b\right)\mathbin{\oslash}v=\bigwedge_{b\in B}(b\mathbin{\oslash}v). \tag*{(17)} ( b ∈ B ⋀ b ) ⊘ v = b ∈ B ⋀ ( b ⊘ v ) . (17)
Note as a result the following straightforward properties of ⊘ \mathbin{\oslash} ⊘ :
Lemma 9. For every v , x , t ∈ V v,x,t\in V v , x , t ∈ V ,
(1) t ⊘ v ≥ t t\mathbin{\oslash}v\ge t t ⊘ v ≥ t ;
(2) v ⊘ t = 1 v\mathbin{\oslash}t=1 v ⊘ t = 1 if and only if v ≥ t v\ge t v ≥ t (in particular, 1 ⊘ v = 1 1\mathbin{\oslash}v=1 1 ⊘ v = 1 , v ⊘ v = 1 v\mathbin{\oslash}v=1 v ⊘ v = 1 , and v ⊘ 0 = 1 v\mathbin{\oslash}0=1 v ⊘ 0 = 1 .)
(3) v ⊘ 1 = v v\mathbin{\oslash}1=v v ⊘ 1 = v ;
(4)
x ≤ t ⟹ v ⊘ x ≥ v ⊘ t ; x\le t\Longrightarrow v\mathbin{\oslash}x\ge v\mathbin{\oslash}t; x ≤ t ⟹ v ⊘ x ≥ v ⊘ t ; (5)
v ⊘ ⋁ a ∈ A a = ⋀ a ∈ A ( v ⊘ a ) ; v\mathbin{\oslash}\bigvee_{a\in A}a=\bigwedge_{a\in A}(v\mathbin{\oslash}a); v ⊘ a ∈ A ⋁ a = a ∈ A ⋀ ( v ⊘ a ) ; (6)
( v ⊘ t ) ⊘ ( v ⊘ x ) ≥ x ⊘ t ; (v\mathbin{\oslash}t)\mathbin{\oslash}(v\mathbin{\oslash}x)\ge x\mathbin{\oslash}t; ( v ⊘ t ) ⊘ ( v ⊘ x ) ≥ x ⊘ t ; (7)
v = ⋀ ε ∈ V ∖ { 1 } v ⊘ ε . v=\bigwedge_{\varepsilon\in V\setminus\{1\}}v\mathbin{\oslash}\varepsilon. v = ε ∈ V ∖ { 1 } ⋀ v ⊘ ε .
The category C C a p \mathrm{C}^{\mathrm{Cap}} C Cap
Reframing Cap
Convergence approach spaces were introduced in [7 ] . Traditionally, a convergence approach space ( X , λ ) (X,\lambda) ( X , λ ) is given by a function λ : F P X → [ 0 , ∞ ] X \lambda:\mathrm{FP}X\to[0,\infty]^X λ : FP X → [ 0 , ∞ ] X satisfying
F ≤ G ⟹ λ ( F ) ≥ λ ( G ) \mathcal{F}\le\mathcal{G}\Longrightarrow\lambda(\mathcal{F})\ge\lambda(\mathcal{G}) F ≤ G ⟹ λ ( F ) ≥ λ ( G )
for the pointwise order of [ 0 , ∞ ] X [0,\infty]^X [ 0 , ∞ ] X induced by the usual order on [ 0 , ∞ ] [0,\infty] [ 0 , ∞ ] , and
λ ( { x } ↑ ) ( x ) = 0 \lambda(\{\!x\!\}^{\uparrow})(x)=0 λ ({ x } ↑ ) ( x ) = 0
and a morphism (then called contraction ) is a function f : ( X , λ X ) → ( Y , λ Y ) f:(X,\lambda_X)\to(Y,\lambda_Y) f : ( X , λ X ) → ( Y , λ Y ) satisfying
λ Y ( f [ F ] ) ∘ f ≤ λ X ( F ) \lambda_Y(f[\mathcal{F}])\circ f\le\lambda_X(\mathcal{F}) λ Y ( f [ F ]) ∘ f ≤ λ X ( F )
for this pointwise order, for every F ∈ F P X \mathcal{F}\in\mathrm{FP}X F ∈ FP X .
In view of Remark 7, we prefer to reframe the structure as a monotone map λ : F P X → [ 0 , 1 ] X \lambda: \mathbb{F}PX \to[0,1]^X λ : F P X → [ 0 , 1 ] X for the pointwise order, with
λ ( { { x } ↑ } ) ( x ) = 1 (19) \lambda(\{\{x\}^{\uparrow}\})(x) = 1 \tag*{(19)} λ ({{ x } ↑ }) ( x ) = 1 (19)
for every x ∈ X x \in X x ∈ X , and morphisms f : ( X , λ X ) → ( Y , λ Y ) f : (X,\lambda_X) \to(Y,\lambda_Y) f : ( X , λ X ) → ( Y , λ Y ) satisfying
λ Y ( f [ F ] ) ∘ f ≥ λ X ( F ) \lambda_Y(f[\mathcal{F}]) \circ f \geq\lambda_X(\mathcal{F}) λ Y ( f [ F ]) ∘ f ≥ λ X ( F )
for every F ∈ F P X . L e t \mathcal{F} \in\mathbb{F}PX. Let F ∈ F P X . L e t \mathcal{C}(X,Y)$ denote the set of all morphisms. The resulting category is denoted Cap.
Note that Cap is a topological category, where initial structures are as follows: If f i : X → ( Y i , λ i ) f_i : X \to(Y_i,\lambda_i) f i : X → ( Y i , λ i ) then the Cap-initial structure on X X X is given by
λ X ( F ) ( x ) = ⋀ i ∈ I λ i ( f i [ F ] ) ( f i ( x ) ) . (20) \lambda_X(\mathcal{F})(x) = \bigwedge_{i\in I} \lambda_i(f_i[\mathcal{F}])(f_i(x)). \tag*{(20)} λ X ( F ) ( x ) = i ∈ I ⋀ λ i ( f i [ F ]) ( f i ( x )) . (20)
Analogously to (pretopology), a convergence approach space ( X , λ ) (X,\lambda) ( X , λ ) is a called a pre-approach space (e.g., [9 ] ) if for any D ⊂ F P X \mathbb{D} \subset\mathbb{F}PX D ⊂ F P X ,
λ ( ⋂ D ) ( ⋅ ) = ⋀ D ∈ D λ D ( ⋅ ) . \lambda\left(\bigcap\mathbb{D}\right)(\cdot) = \bigwedge_{\mathcal{D}\in\mathbb{D}} \lambda_{\mathcal{D}}(\cdot). λ ( ⋂ D ) ( ⋅ ) = D ∈ D ⋀ λ D ( ⋅ ) .
If (PrAp) is only true for subsets D \mathbb{D} D of U P X \mathbb{U}PX U P X of the form β ( F ) \beta(\mathcal{F}) β ( F ) , that is, if
λ ( F ) ( ⋅ ) = ⋀ U ∈ β ( F ) λ U ( ⋅ ) \lambda(\mathcal{F})(\cdot) = \bigwedge_{\mathcal{U}\in\beta(\mathcal{F})} \lambda_{\mathcal{U}}(\cdot) λ ( F ) ( ⋅ ) = U ∈ β ( F ) ⋀ λ U ( ⋅ )
for every F ∈ F X \mathcal{F} \in\mathbb{F}X F ∈ F X , then ( X , λ ) (X,\lambda) ( X , λ ) is a called a pseudo-approach space (e.g.,[8 ] ), generalizing pseudotopologies (compare (pseudotopology)). We denote the corresponding full reflective subcategories of Cap by PrAp and PsAp respectively. Though classical approach theory as in [10 , 11 ] is concerned with the subcategory Ap of PrAp formed by approach spaces , we will only consider them later after we introduce appropriate notions of adherence.
Note that Conv embeds both reflectively and coreflectively in Cap: A Conv-object ( X , lim ) (X,\lim) ( X , lim ) uniquely determines a Cap-object e ( X ) = ( X , λ lim ) e(X)=(X,\lambda_{\lim}) e ( X ) = ( X , λ l i m ) defined by
λ lim F ( x ) = { 1 ; x ∈ lim F 0 ; x ∉ lim F (21) \lambda_{\lim}\mathcal{F}(x) =
\begin{cases}
1; & x \in\lim\mathcal{F} \\
0; & x \notin\lim\mathcal{F}
\end{cases} \tag*{(21)} λ l i m F ( x ) = { 1 ; 0 ; x ∈ lim F x ∈ / lim F (21)
and this embedding e : C o n v → C a p e : \mathrm{Conv} \to\mathrm{Cap} e : Conv → Cap admits a right-adjoint c : C a p → C o n v c : \mathrm{Cap} \to\mathrm{Conv} c : Cap → Conv and a left-adjoint r : C a p → C o n v r : \mathrm{Cap} \to\mathrm{Conv} r : Cap → Conv defined on objects by
x ∈ lim c ( λ ) F ⟺ λ F ( x ) = 1 (22) x \in\lim_{c(\lambda)} \mathcal{F} \Longleftrightarrow\lambda_{\mathcal{F}}(x) = 1 \tag*{(22)} x ∈ c ( λ ) lim F ⟺ λ F ( x ) = 1 (22)
x ∈ lim r ( λ ) F ⟺ λ F ( x ) > 0 (23) x \in\lim_{r(\lambda)} \mathcal{F} \Longleftrightarrow\lambda_{\mathcal{F}}(x) > 0 \tag*{(23)} x ∈ r ( λ ) lim F ⟺ λ F ( x ) > 0 (23)
which extend to a concrete coreflector and a concrete reflector respectively. The functors c c c and r r r restrict to coreflectors and reflectors from PsAp to PsTop, PrAp to PrTop and Ap to Top.
Towards a pointfree abstraction
In order to abstract the structure away to a pointfree setting, we want to abstract P X \mathcal{P}X P X to a lattice L L L and [ 0 , 1 ] [0,1] [ 0 , 1 ] to a quantale V V V as in § 1.5 . For now, let us assume V = [ 0 , 1 ] V=[0,1] V = [ 0 , 1 ] with its usual order. We first need to reinterpret λ : F P X → V X \lambda:\mathbb{F}\mathcal{P}X\to V^{X} λ : F P X → V X via the corresponding monotone map Λ : F P X → ( P X ) V \Lambda:\mathbb{F}\mathcal{P}X\to(\mathcal{P}X)^{V} Λ : F P X → ( P X ) V (giving “convergence by levels”) defined by
Λ ( F ) ( v ) = { x ∈ X : λ ( F ) ( x ) ≥ v } = λ F − 1 [ v , 1 ] . (24) \Lambda(\mathcal{F})(v)=\{x\in X:\lambda(\mathcal{F})(x)\ge v\}=\lambda\mathcal{F}^{-1}[v,1]. \tag*{(24)} Λ ( F ) ( v ) = { x ∈ X : λ ( F ) ( x ) ≥ v } = λ F − 1 [ v , 1 ] . (24)
Identifying Λ ( F ) ( v ) \Lambda(\mathcal{F})(v) Λ ( F ) ( v ) with lim p v F \lim_{p_v}\mathcal{F} lim p v F for a convergence p v p_v p v , the map Λ \Lambda Λ corresponds to the presentation of a Cap-structure in terms of convergence tower [2 ] , that is, a Cap-structure on X X X can equivalently be given by a so-called convergence tower p ˉ = { p ϵ : ϵ ∈ [ 0 , ∞ ] } \bar{p}=\{p_{\epsilon}:\epsilon\in[0,\infty]\} p ˉ = { p ϵ : ϵ ∈ [ 0 , ∞ ]} on X X X where each p ϵ p_{\epsilon} p ϵ is a convergence on X X X , p ∞ p_{\infty} p ∞ is indiscrete,
ϵ ≤ γ ⟹ p γ ≤ p ϵ (25) \epsilon\le\gamma\Longrightarrow p_{\gamma}\le p_{\epsilon} \tag*{(25)} ϵ ≤ γ ⟹ p γ ≤ p ϵ (25)
and
p ϵ = ⋁ γ > ϵ p γ . (26) p_{\epsilon}=\bigvee_{\gamma>\epsilon}p_{\gamma}. \tag*{(26)} p ϵ = γ > ϵ ⋁ p γ . (26)
for each ϵ ∈ [ 0 , ∞ ] \epsilon\in[0,\infty] ϵ ∈ [ 0 , ∞ ] . The morphism condition on a convergence tower is that f : ( X , p ˉ ) → ( Y , q ˉ ) f:(X,\bar{p})\to(Y,\bar{q}) f : ( X , p ˉ ) → ( Y , q ˉ ) be continuous component-wise, that is, f : ( X , p ϵ ) → ( Y , q ϵ ) f:(X,p_{\epsilon})\to(Y,q_{\epsilon}) f : ( X , p ϵ ) → ( Y , q ϵ ) is continuous for each ϵ \epsilon ϵ .
In our setting where V V V is an abstraction of [ 0 , ∞ ] o p [0,\infty]^{\mathrm{op}} [ 0 , ∞ ] op (e.g., [ 0 , 1 ] [0,1] [ 0 , 1 ] ), that p ∞ p_{\infty} p ∞ is antidiscrete becomes that p 0 p_{0} p 0 is antidiscrete, which is built in (2.7). On the other hand, (2.8) becomes
v ≤ t ⟹ p v ≤ p t v\le t\Longrightarrow p_v\le p_t v ≤ t ⟹ p v ≤ p t
and corresponds to the fact that, in view of (2.7), Λ ( F ) \Lambda(\mathcal{F}) Λ ( F ) is a decreasing function of v v v . Similarly, (2.9) becomes
p ϵ = ⋁ γ < ϵ p γ , p_{\epsilon}=\bigvee_{\gamma<\epsilon}p_{\gamma}, p ϵ = γ < ϵ ⋁ p γ ,
which in terms of Λ \Lambda Λ is
Λ ( F ) ( ⋁ v ∈ A v ) = ⋀ v ∈ A Λ ( F ) ( v ) \Lambda(\mathcal{F})\left(\bigvee_{v\in A}v\right)=\bigwedge_{v\in A}\Lambda(\mathcal{F})(v) Λ ( F ) ( v ∈ A ⋁ v ) = v ∈ A ⋀ Λ ( F ) ( v )
for every A ⊂ V A\subset V A ⊂ V . Note that this condition also implies that Λ ( F ) \Lambda(\mathcal{F}) Λ ( F ) be a decreasing function of v v v .
The morphism condition becomes
Λ X ( F ) ( v ) ≤ f − 1 ( Λ Y ( f [ F ] ) ( v ) ) , \Lambda_X(\mathcal{F})(v)\le f^{-1}\left(\Lambda_Y(f[\mathcal{F}])(v)\right), Λ X ( F ) ( v ) ≤ f − 1 ( Λ Y ( f [ F ]) ( v ) ) ,
where ≤ \le ≤ is here the order of P X \mathcal{P}X P X , hence inclusion. Note that in terms of φ = f − 1 : P Y → P X \varphi=f^{-1}:\mathcal{P}Y\to\mathcal{P}X φ = f − 1 : P Y → P X , this can be written
Λ X ( F ) ( v ) ≤ φ ( Λ Y ( φ − 1 ( F ) ( v ) ) ) . \Lambda_X(\mathcal{F})(v)\le\varphi\left(\Lambda_Y\left(\varphi^{-1}(\mathcal{F})(v)\right)\right). Λ X ( F ) ( v ) ≤ φ ( Λ Y ( φ − 1 ( F ) ( v ) ) ) .
Remark 11. Note that φ = f − 1 : P Y → P X \varphi=f^{-1}:\mathcal{P}Y\to\mathcal{P}X φ = f − 1 : P Y → P X preserves arbitrary infima and arbitrary suprema, in particular sends ⊥ P Y \bot_{\mathcal{P}Y} ⊥ P Y to ⊥ P X \bot_{\mathcal{P}X} ⊥ P X and ⊤ P Y \top_{\mathcal{P}Y} ⊤ P Y to ⊤ P X \top_{\mathcal{P}X} ⊤ P X . It also preserves complements.
Pointfree version
We now abstract away powersets to general lattices and V V V is as in § 1.5 . If C is a category of lattices, then a C C a p \mathbf{C}^{\mathrm{Cap}} C Cap -object is a C-object L L L together with a monotone map
Λ : F L → L V , \Lambda:\mathrm{FL}\to L^{V}, Λ : FL → L V ,
such that
Λ ( F ) ( ⋁ v ∈ A v ) = ⋀ v ∈ A Λ ( F ) ( v ) (27) \Lambda(\mathcal{F})\left(\bigvee_{v\in A}v\right)=\bigwedge_{v\in A}\Lambda(\mathcal{F})(v) \tag*{(27)} Λ ( F ) ( v ∈ A ⋁ v ) = v ∈ A ⋀ Λ ( F ) ( v ) (27)
for every A ⊂ V A\subset V A ⊂ V . In particular, taking A = { v 1 , v 2 } A=\{v_{1},v_{2}\} A = { v 1 , v 2 } yields that for each F ∈ F L \mathcal{F}\in\mathrm{FL} F ∈ FL , the map Λ ( F ) \Lambda(\mathcal{F}) Λ ( F ) is a decreasing function of v v v . Taking A = V A=V A = V , together with (1.15), yields that Λ ( F ) ( 1 ) = ⋀ v ∈ V Λ ( F ) ( v ) \Lambda(\mathcal{F})(1)=\bigwedge_{v\in V}\Lambda(\mathcal{F})(v) Λ ( F ) ( 1 ) = ⋀ v ∈ V Λ ( F ) ( v ) and taking A = ∅ A=\varnothing A = ∅ yields that Λ ( F ) ( 0 ) = ⊤ L \Lambda(\mathcal{F})(0)=\top_{L} Λ ( F ) ( 0 ) = ⊤ L .
A C-morphism φ : L → L ′ \varphi:L\to L' φ : L → L ′ between two C C a p \mathbf{C}^{\mathrm{Cap}} C Cap -objects is a C C a p \mathbf{C}^{\mathrm{Cap}} C Cap -morphism if
Λ L ′ ( F ) ( v ) ≤ φ ( Λ L ( φ − 1 ( F ) ) ( v ) ) , (28) \Lambda_{L'}(\mathcal{F})(v)\leq\varphi\left(\Lambda_{L}\left(\varphi^{-1}(\mathcal{F})\right)(v)\right), \tag*{(28)} Λ L ′ ( F ) ( v ) ≤ φ ( Λ L ( φ − 1 ( F ) ) ( v ) ) , (28)
for every v ∈ V v\in V v ∈ V .3
Following [6 ] , we call a category C of lattices admissible if PX is C-object for every set X and there are two classes of index sets I \mathcal{I} I and J \mathcal{J} J such that for every pair L , L ′ L,L' L , L ′ of C-objects, the C-morphisms are exactly the monotone maps that preserve all I \mathcal{I} I -indexed infima that exist in L L L , I ∈ I I\in\mathcal{I} I ∈ I , and all J \mathcal{J} J -indexed suprema that exist in L L L , J ∈ J J\in\mathcal{J} J ∈ J . If C-morphisms additionally preserve all infima, we say that C is fully admissible .
In view of § 2.2 , we have:
Proposition 12. Let C be an admissible category of lattices. The association P : C a p → C C a p \mathbb{P}:\mathrm{Cap}\to\mathbf{C}^{\mathrm{Cap}} P : Cap → C Cap defined on objects by
P ( ( X , λ ) ) = ( P X , Λ P X ) where Λ P X F ( α ) = λ F − 1 [ α , 1 ] , \mathbb{P}((X,\lambda))=(\mathop{\mathrm{P}}X,\Lambda_{\mathop{\mathrm{P}}X})\text{ where }\Lambda_{\mathop{\mathrm{P}}X}\mathcal{F}(\alpha)=\lambda\mathcal{F}^{-1}[\alpha,1], P (( X , λ )) = ( P X , Λ P X ) where Λ P X F ( α ) = λ F − 1 [ α , 1 ] , and on morphisms by
P ( f ) = f − 1 : ( P Y , Λ P Y ) → ( P X , Λ P X ) \mathbb{P}(f)=f^{-1}:(\mathop{\mathrm{P}}Y,\Lambda_{\mathop{\mathrm{P}}Y})\to(\mathop{\mathrm{P}}X,\Lambda_{\mathop{\mathrm{P}}X}) P ( f ) = f − 1 : ( P Y , Λ P Y ) → ( P X , Λ P X ) whenever f : ( X , λ X ) → ( Y , λ Y ) f:(X,\lambda_X)\to(Y,\lambda_Y) f : ( X , λ X ) → ( Y , λ Y ) is a Cap-morphism, defines a contravariant functor, that is, P : C a p → ( C C a p ) o p \mathbb{P}:\mathrm{Cap}\to(\mathbf{C}^{\mathrm{Cap}})^{\mathrm{op}} P : Cap → ( C Cap ) op is a functor.
Remark 13. Note that if ( P X , Λ ) (\mathop{\mathrm{P}}X,\Lambda) ( P X , Λ ) is in the image of the functor P \mathbb{P} P , that is, if Λ = Λ P X \Lambda=\Lambda_{\mathop{\mathrm{P}}X} Λ = Λ P X for a convergence approach space ( X , λ ) (X,\lambda) ( X , λ ) then x ∈ Λ ( ↑ { x } ) ( v ) = ( λ { x } ↑ ) − 1 [ v , 1 ] x\in\Lambda(\uparrow\{x\})(v)=(\lambda\{x\}^{\uparrow})^{-1}[v,1] x ∈ Λ ( ↑ { x }) ( v ) = ( λ { x } ↑ ) − 1 [ v , 1 ] for every v ∈ V v\in V v ∈ V because λ { x } ↑ ( x ) = 1 \lambda\{x\}^{\uparrow}(x)=1 λ { x } ↑ ( x ) = 1 , that is,
∀ x ∈ X , x ∈ Λ ( ↑ { x } ) ( 1 ) . (29) \forall x\in X,\ x\in\Lambda(\uparrow\{x\})(1). \tag*{(29)} ∀ x ∈ X , x ∈ Λ ( ↑ { x }) ( 1 ) . (29)
Space of points
The construction of a functor in the opposite direction mimics that performed in the context of C C o n v \mathbf{C}^{\mathrm{Conv}} C Conv in [6 ] . Several proofs are simple variations of those found in [6 ] and are thus omitted. To start with, we follow [6 ] in appealing to the categorical definition of points, that is, morphisms whose domain is a terminal object. Singletons are terminal objects in topological constructs. In particular, it can be verified that ( { ∗ } , λ ) (\{\ast\},\lambda) ({ ∗ } , λ ) is a Cap-terminal object. Moreover, the corresponding C C a p \mathbf{C}^{\mathrm{Cap}} C Cap -object is initial in C C a p \mathbf{C}^{\mathrm{Cap}} C Cap (this is an analog of [6 Lemma 2.7] , with a similar proof):
Lemma 14. Let C C C be an admissible category of lattices with least and greatest elements. Then 2 : = ( { { ∗ } } , Λ { ∗ } ) 2 := (\{\{\ast\}\},\Lambda_{\{\ast\}}) 2 := ({{ ∗ }} , Λ { ∗ } ) is an initial object in C C a p C^{\mathrm{Cap}} C Cap , hence a terminal object in ( C C a p ) o p (C^{\mathrm{Cap}})^{\mathrm{op}} ( C Cap ) op .
As a result, a point of L L L in ( C C a p ) o p (C^{\mathrm{Cap}})^{\mathrm{op}} ( C Cap ) op is a ( C C a p ) o p (C^{\mathrm{Cap}})^{\mathrm{op}} ( C Cap ) op -morphism φ : 2 → L \varphi:2\to L φ : 2 → L , that is, a C C a p C^{\mathrm{Cap}} C Cap -morphism φ : L → 2 \varphi:L\to2 φ : L → 2 . Following [6 Remark 2.9] , note that letting F = φ − 1 ( { { ∗ } } ) ∈ F L \mathcal{F}=\varphi^{-1}(\{\{\ast\}\})\in\mathrm{FL} F = φ − 1 ({{ ∗ }}) ∈ FL , the morphism condition (2.11) amounts to
∀ v ∈ V , Λ F ( v ) ∈ F , (30) \forall v\in V,\ \Lambda\mathcal{F}(v)\in\mathcal{F}, \tag*{(30)} ∀ v ∈ V , Λ F ( v ) ∈ F , (30)
so that points of L L L can be identified with certain filters on L L L satisfying (3.1). What type of filters need to be considered depends on the nature of C C C -morphisms, so that points of L L L in ( C C a p ) o p (C^{\mathrm{Cap}})^{\mathrm{op}} ( C Cap ) op are filters on L L L satisfying (3.1) that are additionally prime if C C C is the category of lattices, completely prime if C C C is the category of frames, and principal filters of a join-prime element if C C C is the category of coframes. See [6 Remarks 2.10, 2.11, 2.13] . When C = C H C=\mathrm{CH} C = CH then points are principal filters of compact elements ℓ \ell ℓ with ℓ ≤ Λ ( ↑ ℓ ) ( 1 ) \ell\leq\Lambda(\uparrow\ell)(1) ℓ ≤ Λ ( ↑ ℓ ) ( 1 ) .
Let p t C L \mathrm{pt}_{C}L pt C L , or p t L \mathrm{pt}L pt L if C C C is clear, denote the set of all points of L L L (in ( C C a p ) o p (C^{\mathrm{Cap}})^{\mathrm{op}} ( C Cap ) op ).
Following a similar path to that charted in [6 ] , we will now endow p t L \mathrm{pt}L pt L with a C C a p C^{\mathrm{Cap}} C Cap -structure. To this end, given ℓ ∈ L \ell\in L ℓ ∈ L where L L L is a C C a p C^{\mathrm{Cap}} C Cap -object, let
ℓ ∙ = { φ ∈ p t L : φ ( ℓ ) = { ∗ } } = { F ∈ p t L : ℓ ∈ F } \ell^{\bullet}=\{\varphi\in\mathrm{pt}\,L:\varphi(\ell)=\{\ast\}\}=\{\mathcal{F}\in\mathrm{pt}\,L:\ell\in\mathcal{F}\} ℓ ∙ = { φ ∈ pt L : φ ( ℓ ) = { ∗ }} = { F ∈ pt L : ℓ ∈ F }
and, given F ∈ F P ( p t L ) \mathcal{F}\in\mathrm{FP}(\mathrm{pt}\,L) F ∈ FP ( pt L ) , let
F ∘ = { ℓ ∈ L : ℓ ∙ ∈ F } . \mathcal{F}^{\circ}=\{\ell\in L:\ell^{\bullet}\in\mathcal{F}\}. F ∘ = { ℓ ∈ L : ℓ ∙ ∈ F } .
The following is straightforward and can be proved along the lines of the analogous statement [6 Lemma 2.19] :
Lemma 15. Let C C C be an admissible category of lattices with top and bottom, and let L L L be a C C a p C^{\mathrm{Cap}} C Cap -object and let ℓ , ℓ ′ ∈ L \ell,\ell'\in L ℓ , ℓ ′ ∈ L .
if ℓ ≤ ℓ ′ \ell\leq\ell' ℓ ≤ ℓ ′ then ℓ ∙ ⊆ ℓ ′ ∙ \ell^{\bullet}\subseteq\ell'^{\bullet} ℓ ∙ ⊆ ℓ ′ ∙ .
For all F , G ∈ F P ( p t L ) \mathcal{F},\mathcal{G}\in\mathrm{FP}(\mathrm{pt}\,L) F , G ∈ FP ( pt L ) , if F ⊂ G \mathcal{F}\subset\mathcal{G} F ⊂ G then F ∘ ⊂ G ∘ \mathcal{F}^{\circ}\subset\mathcal{G}^{\circ} F ∘ ⊂ G ∘ .
For all ℓ 1 , ℓ 2 , ⋯ , ℓ n ∈ L \ell_1,\ell_2,\cdots,\ell_n\in L ℓ 1 , ℓ 2 , ⋯ , ℓ n ∈ L (n ∈ N n\in\mathbb{N} n ∈ N ), ⋃ i = 1 n ℓ i ∙ = ( ⋁ i = 1 n ℓ i ) ∙ \bigcup_{i=1}^{n}\ell_i^{\bullet}=\left(\bigvee_{i=1}^{n}\ell_i\right)^{\bullet} ⋃ i = 1 n ℓ i ∙ = ( ⋁ i = 1 n ℓ i ) ∙ and if points are completely prime, then
⋃ ℓ ∈ A ℓ ∙ = ( ⋁ A ) ∙ \bigcup_{\ell\in A}\ell^{\bullet}=\left(\bigvee A\right)^{\bullet} ℓ ∈ A ⋃ ℓ ∙ = ( ⋁ A ) ∙ for every A ⊂ L A\subset L A ⊂ L .
For all ℓ 1 , ℓ 2 , ⋯ , ℓ n ∈ L \ell_1,\ell_2,\cdots,\ell_n\in L ℓ 1 , ℓ 2 , ⋯ , ℓ n ∈ L (n ∈ N n\in\mathbb{N} n ∈ N ), ⋂ i = 1 n ℓ i ∙ = ( ⋀ i = 1 n ℓ i ) ∙ \bigcap_{i=1}^{n}\ell_i^{\bullet}=\left(\bigwedge_{i=1}^{n}\ell_i\right)^{\bullet} ⋂ i = 1 n ℓ i ∙ = ( ⋀ i = 1 n ℓ i ) ∙ and if points are principal (e.g., C C C is fully admissible) then
⋂ ℓ ∈ A ℓ ∙ = ( ⋀ A ) ∙ \bigcap_{\ell\in A}\ell^{\bullet}=\left(\bigwedge A\right)^{\bullet} ℓ ∈ A ⋂ ℓ ∙ = ( ⋀ A ) ∙ for every A ⊂ L A\subset L A ⊂ L .
For every F ∈ F P ( p t L ) \mathcal{F}\in\mathrm{FP}(\mathrm{pt}\,L) F ∈ FP ( pt L ) , F ∘ ∈ F L \mathcal{F}^{\circ}\in\mathrm{FL} F ∘ ∈ FL .
For every φ ∈ p t L \varphi\in\mathrm{pt}\,L φ ∈ pt L , ( { φ } ↑ ) ∘ = φ − 1 ( { { ∗ } } ) \left(\{\varphi\}^{\uparrow}\right)^{\circ}=\varphi^{-1}(\{\{\ast\}\}) ( { φ } ↑ ) ∘ = φ − 1 ({{ ∗ }}) is the filter on L L L representing φ \varphi φ .
x ∈ ( Λ L ( { x } ↑ ) ( 1 ) ) ∙ x\in\left(\Lambda_{L}(\{x\}^{\uparrow})(1)\right)^{\bullet} x ∈ ( Λ L ({ x } ↑ ) ( 1 ) ) ∙ for every x ∈ p t L x\in\mathrm{pt}\,L x ∈ pt L .
Similarly to Remark 4, we have:
Definition 18. Given an admissible category C of lattices and a C C a p \mathrm{C}^{\mathrm{Cap}} C Cap -object L L L , pt L \operatorname{pt}L pt L denotes the set of points of L L L (in ( C C a p ) o p (\mathrm{C}^{\mathrm{Cap}})^{\mathrm{op}} ( C Cap ) op ) endowed with the Cap-structure given by
λ pt L F ( x ) = ⋁ { v ∈ V : x ∈ ( Λ L F ∘ ( v ) ) ∙ } . (31) \lambda_{\operatorname{pt}L}\mathcal{F}(x)=\bigvee\{v\in V:x\in(\Lambda_L\mathcal{F}^{\circ}(v))^\bullet\}. \tag*{(31)} λ pt L F ( x ) = ⋁ { v ∈ V : x ∈ ( Λ L F ∘ ( v ) ) ∙ } . (31) That λ pt L \lambda_{\operatorname{pt}L} λ pt L is indeed a Cap-structure is easily seen: it satisfies (2.2) by Lemma 15(7) and λ pt L : F P ( pt L ) → V X \lambda_{\operatorname{pt}L}:\mathbb{F}\mathbb{P}(\operatorname{pt}L)\to V^X λ pt L : FP ( pt L ) → V X is monotone because, in view of Lemma 15(2) and Lemma 15(1) and the fact that Λ : F L → L V \Lambda:\mathbb{F}L\to L^V Λ : F L → L V is monotone, ( Λ L F ∘ ( v ) ) ∙ ⊂ ( Λ L G ∘ ( v ) ) ∙ (\Lambda_L\mathcal{F}^{\circ}(v))^\bullet\subset(\Lambda_L\mathcal{G}^{\circ}(v))^\bullet ( Λ L F ∘ ( v ) ) ∙ ⊂ ( Λ L G ∘ ( v ) ) ∙ whenever F , G ∈ F P ( pt L ) \mathcal{F},\mathcal{G}\in\mathbb{F}\mathbb{P}(\operatorname{pt}L) F , G ∈ FP ( pt L ) with F ⊂ G \mathcal{F}\subset\mathcal{G} F ⊂ G .
To prove that pt : C C a p → C a p \operatorname{pt}:\mathrm{C}^{\mathrm{Cap}}\to\mathrm{Cap} pt : C Cap → Cap defines an adjoint functor, we proceed as in [6 ] showing that ϵ L = ( − ) ∙ : L → P ( pt L ) \epsilon_L=(-)^\bullet:L\to\mathbb{P}(\operatorname{pt}L) ϵ L = ( − ) ∙ : L → P ( pt L ) are universal arrows.
Proposition 19. If C is a fully admissible category of lattices, then for every C C a p \mathrm{C}^{\mathrm{Cap}} C Cap -object L L L , the map ϵ L : L → P ( pt L ) \epsilon_L:L\to\mathbb{P}(\operatorname{pt}L) ϵ L : L → P ( pt L ) defined by ϵ L ( ℓ ) = ℓ ∙ \epsilon_L(\ell)=\ell^\bullet ϵ L ( ℓ ) = ℓ ∙ is a final C C a p \mathrm{C}^{\mathrm{Cap}} C Cap -morphism.
Proof. Proving that ϵ L \epsilon_L ϵ L is a C-morphism can be done exactly as in [6 Lemma 2.23] , so, we focus on proving that ϵ L \epsilon_L ϵ L is continuous, that is,
Λ P pt L F ( α ) ≤ ϵ L Λ ϵ L − 1 F ( α ) , (32) \Lambda_{\mathbb{P}\operatorname{pt}L}\mathcal{F}(\alpha)\leq\epsilon_L\Lambda_{\epsilon_L}^{-1}\mathcal{F}(\alpha), \tag*{(32)} Λ P pt L F ( α ) ≤ ϵ L Λ ϵ L − 1 F ( α ) , (32) for every α ∈ V \alpha\in V α ∈ V and F ∈ F P pt L \mathcal{F}\in\mathbb{F}\mathbb{P}\operatorname{pt}L F ∈ FP pt L , that is,
{ x ∈ pt L : ⋁ { v ∈ V : x ∈ ( Λ L F ∘ ( v ) ) ∙ } ≥ α } ⊂ ( Λ L F ∘ ( α ) ) ∙ . \{x\in\operatorname{pt}L:\bigvee\{v\in V:x\in(\Lambda_L\mathcal{F}^{\circ}(v))^\bullet\}\geq\alpha\}\subset(\Lambda_L\mathcal{F}^{\circ}(\alpha))^\bullet. { x ∈ pt L : ⋁ { v ∈ V : x ∈ ( Λ L F ∘ ( v ) ) ∙ } ≥ α } ⊂ ( Λ L F ∘ ( α ) ) ∙ . Let x ∈ pt L x\in\operatorname{pt}L x ∈ pt L with ⋁ { v ∈ V : x ∈ ( Λ L F ∘ ( v ) ) ∙ } ≥ α \bigvee\{v\in V:x\in(\Lambda_L\mathcal{F}^{\circ}(v))^\bullet\}\geq\alpha ⋁ { v ∈ V : x ∈ ( Λ L F ∘ ( v ) ) ∙ } ≥ α . In view of (2.10),
Λ L F ∘ ( ⋁ x ∈ ( Λ L F ∘ ( v ) ) ∙ v ) = ⋀ x ∈ ( Λ L F ∘ ( v ) ) ∙ Λ L F ∘ ( v ) ≤ Λ L F ∘ ( α ) , \Lambda_L\mathcal{F}^{\circ}\left(\bigvee_{x\in(\Lambda_L\mathcal{F}^{\circ}(v))^\bullet}v\right)=\bigwedge_{x\in(\Lambda_L\mathcal{F}^{\circ}(v))^\bullet}\Lambda_L\mathcal{F}^{\circ}(v)\leq\Lambda_L\mathcal{F}^{\circ}(\alpha), Λ L F ∘ x ∈ ( Λ L F ∘ ( v ) ) ∙ ⋁ v = x ∈ ( Λ L F ∘ ( v ) ) ∙ ⋀ Λ L F ∘ ( v ) ≤ Λ L F ∘ ( α ) , and x ∈ ( Λ L F ∘ ( v ) ) ∙ x\in(\Lambda_L\mathcal{F}^{\circ}(v))^\bullet x ∈ ( Λ L F ∘ ( v ) ) ∙ means Λ L F ∘ ( v ) ∈ x \Lambda_L\mathcal{F}^{\circ}(v)\in x Λ L F ∘ ( v ) ∈ x . As C is fully admissible, x x x is closed under infima so that ⋀ x ∈ ( Λ L F ∘ ( v ) ) ∙ Λ L F ∘ ( v ) \bigwedge_{x\in(\Lambda_L\mathcal{F}^{\circ}(v))^\bullet}\Lambda_L\mathcal{F}^{\circ}(v) ⋀ x ∈ ( Λ L F ∘ ( v ) ) ∙ Λ L F ∘ ( v ) , hence Λ L F ∘ ( α ) \Lambda_L\mathcal{F}^{\circ}(\alpha) Λ L F ∘ ( α ) , belong to x x x , that is, x ∈ ( Λ L F ∘ ( α ) ) ∙ x\in(\Lambda_L\mathcal{F}^{\circ}(\alpha))^\bullet x ∈ ( Λ L F ∘ ( α ) ) ∙ , proving (3.3).
Conversely, if x ∈ ( Λ L F ∘ ( α ) ) ∙ x\in(\Lambda_L\mathcal{F}^{\circ}(\alpha))^\bullet x ∈ ( Λ L F ∘ ( α ) ) ∙ , then ⋁ { v ∈ V : x ∈ ( Λ L F ∘ ( v ) ) ∙ } ≥ α \bigvee\{v\in V:x\in(\Lambda_L\mathcal{F}^{\circ}(v))^\bullet\}\geq\alpha ⋁ { v ∈ V : x ∈ ( Λ L F ∘ ( v ) ) ∙ } ≥ α and thus (3.3) is an equality and ϵ L \epsilon_L ϵ L is final. □ \square □
Corollary 20. If C is a fully admissible category of lattices then pt : ( C C a p ) o p → C a p \operatorname{pt}:(\mathrm{C}^{\mathrm{Cap}})^{\mathrm{op}}\to\mathrm{Cap} pt : ( C Cap ) op → Cap is a right-adjoint to P : C a p → ( C C a p ) o p \mathbb{P}:\mathrm{Cap}\to(\mathrm{C}^{\mathrm{Cap}})^{\mathrm{op}} P : Cap → ( C Cap ) op and thus Cap can be seen as a coreflective subcategory of ( C C a p ) o p (\mathrm{C}^{\mathrm{Cap}})^{\mathrm{op}} ( C Cap ) op .
We call a C C a p \mathrm{C}^{\mathrm{Cap}} C Cap -object ( L , Λ ) (L,\Lambda) ( L , Λ ) spatial if ϵ L : L → P ( pt L ) \epsilon_L:L\to\mathbb{P}(\operatorname{pt}L) ϵ L : L → P ( pt L ) is an isomorphism of C C a p \mathrm{C}^{\mathrm{Cap}} C Cap and weakly spatial if ϵ L \epsilon_L ϵ L has a trivial kernel, that is,
ℓ ∙ = ∅ ⟺ ℓ = ⊥ L . (33) \ell^\bullet=\varnothing\Longleftrightarrow\ell=\bot_L. \tag*{(33)} ℓ ∙ = ∅ ⟺ ℓ = ⊥ L . (33)
Assume pt L ≠ ∅ \operatorname{pt} L \ne\varnothing pt L = ∅ . Note that in view of Lemma 15 (4), if F ∈ FP ( pt L ) \mathcal{F} \in\operatorname{FP}(\operatorname{pt} L) F ∈ FP ( pt L ) then
F ♮ ∙ : = ( F ∩ { ℓ ∙ : ℓ ∈ L } ) ↑ pt L \mathcal{F}^{\natural\bullet}:=\left(\mathcal{F}\cap\{\ell^{\bullet}:\ell\in L\}\right)^{\uparrow\operatorname{pt} L} F ♮ ∙ := ( F ∩ { ℓ ∙ : ℓ ∈ L } ) ↑ pt L
is also a (set-theoretic) filter on pt L \operatorname{pt} L pt L and F ♮ ∙ ≤ F \mathcal{F}^{\natural\bullet}\leq\mathcal{F} F ♮ ∙ ≤ F , but
λ pt L F = λ pt L F ♮ ∙ , (34) \lambda_{\operatorname{pt} L}\mathcal{F}=\lambda_{\operatorname{pt} L}\mathcal{F}^{\natural\bullet}, \tag*{(34)} λ pt L F = λ pt L F ♮ ∙ , (34)
because F ∘ = ( F ♮ ∙ ) ∘ \mathcal{F}^{\circ}=(\mathcal{F}^{\natural\bullet})^{\circ} F ∘ = ( F ♮ ∙ ) ∘ . Note also that if G ∈ FL \mathcal{G}\in\operatorname{FL} G ∈ FL then
G ∙ : = { g ∙ : g ∈ G } ↑ \mathcal{G}_{\bullet}:=\{g^{\bullet}:g\in\mathcal{G}\}^{\uparrow} G ∙ := { g ∙ : g ∈ G } ↑
is a (possibly degenerate, but not in the weakly spatial case) set-theoretic filter on pt L \operatorname{pt} L pt L and G ⊂ ( G ∙ ) ∘ \mathcal{G}\subset(\mathcal{G}_{\bullet})^{\circ} G ⊂ ( G ∙ ) ∘ . On the other hand, if F ∈ FP ( pt L ) \mathcal{F}\in\operatorname{FP}(\operatorname{pt} L) F ∈ FP ( pt L ) , ( F ∘ ) ∙ = F ♮ ∙ ⊂ F (\mathcal{F}^{\circ})_{\bullet}=\mathcal{F}^{\natural\bullet}\subset\mathcal{F} ( F ∘ ) ∙ = F ♮ ∙ ⊂ F .
Lemma 21. If ( L , Λ ) (L,\Lambda) ( L , Λ ) is a C C a p \mathrm{C}^{\mathrm{Cap}} C Cap -object and F , H ∈ FP ( pt L ) \mathcal{F},\mathcal{H}\in\operatorname{FP}(\operatorname{pt} L) F , H ∈ FP ( pt L ) then
F # H ⟹ F ∘ # H ∘ . \mathcal{F}\mathrel{\#}\mathcal{H}\Longrightarrow\mathcal{F}^{\circ}\mathrel{\#}\mathcal{H}^{\circ}. F # H ⟹ F ∘ # H ∘ . Moreover, if L L L is weakly spatial then
F ∘ # H ∘ ⟹ F ♮ ∙ # H ♮ ∙ . \mathcal{F}^{\circ}\mathrel{\#}\mathcal{H}^{\circ}\Longrightarrow\mathcal{F}^{\natural\bullet}\mathrel{\#}\mathcal{H}^{\natural\bullet}. F ∘ # H ∘ ⟹ F ♮ ∙ # H ♮ ∙ . Proof. If F # H \mathcal{F}\mathrel{\#}\mathcal{H} F # H , f ∈ F ∘ f\in\mathcal{F}^{\circ} f ∈ F ∘ and h ∈ H ∘ h\in\mathcal{H}^{\circ} h ∈ H ∘ then f ∙ ∈ F f^{\bullet}\in\mathcal{F} f ∙ ∈ F and h ∙ ∈ H h^{\bullet}\in\mathcal{H} h ∙ ∈ H and thus f ∙ ∩ h ∙ = ( f ∧ h ) ∙ ≠ ∅ f^{\bullet}\cap h^{\bullet}=(f\wedge h)^{\bullet}\ne\varnothing f ∙ ∩ h ∙ = ( f ∧ h ) ∙ = ∅ by Lemma 15 , so that f ∧ h ≠ 0 L f\wedge h\ne0_L f ∧ h = 0 L .
Conversely, if L L L is weakly spatial and F ∘ # H ∘ \mathcal{F}^{\circ}\mathrel{\#}\mathcal{H}^{\circ} F ∘ # H ∘ then f ∧ h ≠ 0 L f\wedge h\ne0_L f ∧ h = 0 L whenever f ∙ ∈ F f^{\bullet}\in\mathcal{F} f ∙ ∈ F and h ∙ ∈ H h^{\bullet}\in\mathcal{H} h ∙ ∈ H . By weak spatiality, ( f ∧ h ) ∙ = f ∙ ∩ h ∙ ≠ ∅ (f\wedge h)^{\bullet}=f^{\bullet}\cap h^{\bullet}\ne\varnothing ( f ∧ h ) ∙ = f ∙ ∩ h ∙ = ∅ and thus F ♮ ∙ # H ♮ ∙ \mathcal{F}^{\natural\bullet}\mathrel{\#}\mathcal{H}^{\natural\bullet} F ♮ ∙ # H ♮ ∙ .
Lemma 22. Let L L L be weakly spatial, F ∈ FP ( pt L ) \mathcal{F}\in\operatorname{FP}(\operatorname{pt} L) F ∈ FP ( pt L ) and G ∈ FL \mathcal{G}\in\operatorname{FL} G ∈ FL . Then
G # F ∘ ⟺ G ∙ # F ♮ ∙ ⟺ ∃ U ∈ β ( F ♮ ∙ ) : U ∘ ⊃ G ∪ F ∘ . \mathcal{G}\mathrel{\#}\mathcal{F}^{\circ}\Longleftrightarrow\mathcal{G}_{\bullet}\mathrel{\#}\mathcal{F}^{\natural\bullet}\Longleftrightarrow\exists U\in\beta(\mathcal{F}^{\natural\bullet}):U^{\circ}\supset\mathcal{G}\cup\mathcal{F}^{\circ}. G # F ∘ ⟺ G ∙ # F ♮ ∙ ⟺ ∃ U ∈ β ( F ♮ ∙ ) : U ∘ ⊃ G ∪ F ∘ . Proof. If G # F ∘ \mathcal{G}\mathrel{\#}\mathcal{F}^{\circ} G # F ∘ then G ∙ # F ♮ ∙ \mathcal{G}_{\bullet}\mathrel{\#}\mathcal{F}^{\natural\bullet} G ∙ # F ♮ ∙ because if g ∈ G g\in\mathcal{G} g ∈ G and f ∈ F ∘ f\in\mathcal{F}^{\circ} f ∈ F ∘ then g ∧ f ≠ ⊥ L g\wedge f\ne\bot_L g ∧ f = ⊥ L and
( g ∧ f ) ∙ = g ∙ ∩ f ∙ ≠ ∅ (g\wedge f)^{\bullet}=g^{\bullet}\cap f^{\bullet}\ne\varnothing ( g ∧ f ) ∙ = g ∙ ∩ f ∙ = ∅ because L L L is weakly spatial. If G ∙ # F ♮ ∙ \mathcal{G}_{\bullet}\mathrel{\#}\mathcal{F}^{\natural\bullet} G ∙ # F ♮ ∙ there is U ∈ β ( G ∙ ∨ F ♮ ∙ ) U\in\beta(\mathcal{G}_{\bullet}\vee\mathcal{F}^{\natural\bullet}) U ∈ β ( G ∙ ∨ F ♮ ∙ ) and
U ∘ ⊃ ( G ∙ ) ∘ ∪ ( F ♮ ∙ ) ∘ ⊃ G ∪ F ∘ . U^{\circ}\supset(\mathcal{G}_{\bullet})^{\circ}\cup(\mathcal{F}^{\natural\bullet})^{\circ}\supset\mathcal{G}\cup\mathcal{F}^{\circ}. U ∘ ⊃ ( G ∙ ) ∘ ∪ ( F ♮ ∙ ) ∘ ⊃ G ∪ F ∘ . Finally, if U ∘ ⊃ G ∪ F ∘ U^{\circ}\supset\mathcal{G}\cup\mathcal{F}^{\circ} U ∘ ⊃ G ∪ F ∘ then G # F ∘ \mathcal{G}\mathrel{\#}\mathcal{F}^{\circ} G # F ∘ because U ∘ ∈ FL U^{\circ}\in\operatorname{FL} U ∘ ∈ FL .
Theorem 23. If C \mathrm{C} C is a fully admissible category of lattices, then C C a p \mathrm{C}^{\mathrm{Cap}} C Cap is a topological category over C \mathrm{C} C . Indeed, the unique final lift ( L , Λ L ) (L,\Lambda_L) ( L , Λ L ) of a sink ( φ i : ( L i , Λ i ) → L ) i ∈ I (\varphi_i:(L_i,\Lambda_i)\to L)_{i\in I} ( φ i : ( L i , Λ i ) → L ) i ∈ I is given by
Λ L F ( v ) = ⋀ i ∈ I φ i ( Λ i ( φ i − 1 ( F ) ) ( v ) ) , \Lambda_L\mathcal{F}(v)=\bigwedge_{i\in I}\varphi_i\left(\Lambda_i\left(\varphi_i^{-1}(\mathcal{F})\right)(v)\right), Λ L F ( v ) = i ∈ I ⋀ φ i ( Λ i ( φ i − 1 ( F ) ) ( v ) ) , with the convention that Λ i ( φ i − 1 ( F ) ) ( v ) = ⊤ L i \Lambda_i(\varphi_i^{-1}(\mathcal{F}))(v)=\top_{L_i} Λ i ( φ i − 1 ( F )) ( v ) = ⊤ L i if φ i − 1 ( F ) \varphi_i^{-1}(\mathcal{F}) φ i − 1 ( F ) is degenerate.
Proof. Let U : C C a p → C U:\mathrm{C}^{\mathrm{Cap}}\to\mathrm{C} U : C Cap → C be the forgetful functor. It is enough to show that every U U U -structure sink ( φ i : ( L i , Λ i ) → L ) i ∈ I (\varphi_i:(L_i,\Lambda_i)\to L)_{i\in I} ( φ i : ( L i , Λ i ) → L ) i ∈ I admits a unique final lift ( L , Λ L ) (L,\Lambda_L) ( L , Λ L ) . Whenever F ∈ FL \mathcal{F}\in\operatorname{FL} F ∈ FL , φ i − 1 ( F ) \varphi_i^{-1}(\mathcal{F}) φ i − 1 ( F ) is a filter on L i L_i L i that is proper if and only if φ i ( ⊥ L i ) ∉ F \varphi_i(\bot_{L_i})\notin\mathcal{F} φ i ( ⊥ L i ) ∈ / F . Define Λ L : FL → L V \Lambda_L:\operatorname{FL}\to L^V Λ L : FL → L V by
Λ L F ( v ) = ⋀ i ∈ I φ i ( Λ i ( φ i − 1 ( F ) ) ( v ) ) = ⋀ i ∈ I , φ i − 1 ( F ) ∈ FL i φ i ( Λ i ( φ i − 1 ( F ) ) ( v ) ) . \Lambda_L\mathcal{F}(v)=\bigwedge_{i\in I}\varphi_i\left(\Lambda_i\left(\varphi_i^{-1}(\mathcal{F})\right)(v)\right)=\bigwedge_{\substack{i\in I,\ \varphi_i^{-1}(\mathcal{F})\in\operatorname{FL}_i}}\varphi_i\left(\Lambda_i\left(\varphi_i^{-1}(\mathcal{F})\right)(v)\right). Λ L F ( v ) = i ∈ I ⋀ φ i ( Λ i ( φ i − 1 ( F ) ) ( v ) ) = i ∈ I , φ i − 1 ( F ) ∈ FL i ⋀ φ i ( Λ i ( φ i − 1 ( F ) ) ( v ) ) .
with the convention that for improper filters, we have Λ i ( L i ) ( v ) = ⊤ L i \Lambda_i(L_i)(v)=\top_{L_i} Λ i ( L i ) ( v ) = ⊤ L i . It is easily verified that Λ L \Lambda_L Λ L is monotone and, for every A ⊂ V A\subset V A ⊂ V ,
Λ L F ( ⋁ A ) = ⋀ i ∈ I φ i ( Λ i ( φ i − 1 ( F ) ) ( ⋁ A ) ) = ⋀ i ∈ I φ i ( ⋀ a ∈ A Λ i ( φ i − 1 ( F ) ) ( a ) ) = ⋀ i ∈ I ⋀ a ∈ A φ i ( Λ i ( φ i − 1 ( F ) ) ( a ) ) . \begin{aligned}
\Lambda_L\mathcal{F}\left(\bigvee A\right)
&=\bigwedge_{i\in I}\varphi_i\left(\Lambda_i\left(\varphi_i^{-1}(\mathcal{F})\right)\left(\bigvee A\right)\right)\\
&=\bigwedge_{i\in I}\varphi_i\left(\bigwedge_{a\in A}\Lambda_i\left(\varphi_i^{-1}(\mathcal{F})\right)(a)\right)\\
&=\bigwedge_{i\in I}\bigwedge_{a\in A}\varphi_i\left(\Lambda_i\left(\varphi_i^{-1}(\mathcal{F})\right)(a)\right).
\end{aligned} Λ L F ( ⋁ A ) = i ∈ I ⋀ φ i ( Λ i ( φ i − 1 ( F ) ) ( ⋁ A ) ) = i ∈ I ⋀ φ i ( a ∈ A ⋀ Λ i ( φ i − 1 ( F ) ) ( a ) ) = i ∈ I ⋀ a ∈ A ⋀ φ i ( Λ i ( φ i − 1 ( F ) ) ( a ) ) .
because C is fully admissible, so that Λ L F ( ⋁ A ) = ⋀ a ∈ A Λ L F ( a ) \Lambda_L\mathcal{F}\left(\bigvee A\right)=\bigwedge_{a\in A}\Lambda_L\mathcal{F}(a) Λ L F ( ⋁ A ) = ⋀ a ∈ A Λ L F ( a ) . Hence ( L , Λ L ) (L,\Lambda_L) ( L , Λ L ) is a C C a p \mathrm{C}^{\mathrm{Cap}} C Cap -object. Moreover, each φ j : ( L j , Λ j ) → ( L , Λ L ) \varphi_j:(L_j,\Lambda_j)\to(L,\Lambda_L) φ j : ( L j , Λ j ) → ( L , Λ L ) is a C C a p \mathrm{C}^{\mathrm{Cap}} C Cap -morphism because
Λ L F ( v ) ≤ ⋀ i ∈ I φ i ( Λ i ( φ i − 1 ( F ) ) ( v ) ) ≤ φ j ( Λ j ( φ j − 1 ( F ) ) ( v ) ) . \Lambda_L\mathcal{F}(v)\leq\bigwedge_{i\in I}\varphi_i\left(\Lambda_i\left(\varphi_i^{-1}(\mathcal{F})\right)(v)\right)\leq\varphi_j\left(\Lambda_j\left(\varphi_j^{-1}(\mathcal{F})\right)(v)\right). Λ L F ( v ) ≤ i ∈ I ⋀ φ i ( Λ i ( φ i − 1 ( F ) ) ( v ) ) ≤ φ j ( Λ j ( φ j − 1 ( F ) ) ( v ) ) .
To see the final lift property, let ψ ∈ C ( L , M ) \psi\in\mathrm{C}(L,M) ψ ∈ C ( L , M ) where ( M , Λ M ) (M,\Lambda_M) ( M , Λ M ) is a C C a p \mathrm{C}^{\mathrm{Cap}} C Cap -object and suppose that each ψ ∘ φ i : ( L i , Λ i ) → ( M , Λ M ) \psi\circ\varphi_i:(L_i,\Lambda_i)\to(M,\Lambda_M) ψ ∘ φ i : ( L i , Λ i ) → ( M , Λ M ) is a C C a p \mathrm{C}^{\mathrm{Cap}} C Cap -morphism, that is,
Λ M ( G ) ( v ) ≤ ( ψ ∘ φ i ) ( Λ i ( ( ψ ∘ φ i ) − 1 ( G ) ( v ) ) ) = ψ ( φ i ( Λ i ( φ i − 1 ( ψ − 1 ( G ) ) ( v ) ) ) ) \Lambda_M(\mathcal{G})(v)\leq(\psi\circ\varphi_i)\left(\Lambda_i\left((\psi\circ\varphi_i)^{-1}(\mathcal{G})(v)\right)\right)=\psi\left(\varphi_i\left(\Lambda_i\left(\varphi_i^{-1}\left(\psi^{-1}(\mathcal{G})\right)(v)\right)\right)\right) Λ M ( G ) ( v ) ≤ ( ψ ∘ φ i ) ( Λ i ( ( ψ ∘ φ i ) − 1 ( G ) ( v ) ) ) = ψ ( φ i ( Λ i ( φ i − 1 ( ψ − 1 ( G ) ) ( v ) ) ) )
for every G ∈ F M \mathcal{G}\in F_M G ∈ F M . Thus
Λ M ( G ) ( v ) ≤ ⋀ i ∈ I ψ ( φ i ( Λ i ( φ i − 1 ( ψ − 1 ( G ) ) ( v ) ) ) ) ≤ ψ ( ⋀ i ∈ I φ i ( Λ i ( φ i − 1 ( ψ − 1 ( G ) ) ( v ) ) ) ) ≤ ψ ( Λ L ( ψ − 1 ( G ) ) ( v ) ) , \begin{aligned}
\Lambda_M(\mathcal{G})(v)
&\leq\bigwedge_{i\in I}\psi\left(\varphi_i\left(\Lambda_i\left(\varphi_i^{-1}\left(\psi^{-1}(\mathcal{G})\right)(v)\right)\right)\right)\\
&\leq\psi\left(\bigwedge_{i\in I}\varphi_i\left(\Lambda_i\left(\varphi_i^{-1}\left(\psi^{-1}(\mathcal{G})\right)(v)\right)\right)\right)\\
&\leq\psi\left(\Lambda_L\left(\psi^{-1}(\mathcal{G})\right)(v)\right),
\end{aligned} Λ M ( G ) ( v ) ≤ i ∈ I ⋀ ψ ( φ i ( Λ i ( φ i − 1 ( ψ − 1 ( G ) ) ( v ) ) ) ) ≤ ψ ( i ∈ I ⋀ φ i ( Λ i ( φ i − 1 ( ψ − 1 ( G ) ) ( v ) ) ) ) ≤ ψ ( Λ L ( ψ − 1 ( G ) ) ( v ) ) ,
so that ψ ∈ C C a p ( L , M ) \psi\in\mathrm{C}^{\mathrm{Cap}}(L,M) ψ ∈ C Cap ( L , M ) . □
CC a p ^{\mathrm{Cap}} Cap and CC o n v ^{\mathrm{Conv}} Conv
Recall that the embedding (21) of Conv into Cap is both reflective and coreflective, with reflector r r r and c c c given by (23) and (22) respectively. Analogously, ( C C o n v ) o p (\mathrm{C}^{\mathrm{Conv}})^{\mathrm{op}} ( C Conv ) op embeds in ( C C a p ) o p (\mathrm{C}^{\mathrm{Cap}})^{\mathrm{op}} ( C Cap ) op both reflectively and coreflectively. Namely, E : ( C C o n v ) o p → ( C C a p ) o p E:(\mathrm{C}^{\mathrm{Conv}})^{\mathrm{op}}\to(\mathrm{C}^{\mathrm{Cap}})^{\mathrm{op}} E : ( C Conv ) op → ( C Cap ) op sends a CC o n v ^{\mathrm{Conv}} Conv -object ( L , lim ) (L,\lim) ( L , lim ) to the CC a p ^{\mathrm{Cap}} Cap -object E ( L ) = ( L , Λ lim ) E(L)=(L,\Lambda_{\lim}) E ( L ) = ( L , Λ l i m ) where
Λ lim F ( v ) = { lim F v > 0 , ⊤ L v = 0. (35) \Lambda_{\lim}\mathcal{F}(v)=
\begin{cases}
\lim\mathcal{F} & v>0,\\
\top_L & v=0.
\end{cases} \tag*{(35)} Λ l i m F ( v ) = { lim F ⊤ L v > 0 , v = 0. (35)
and a morphism φ ∈ ( C C o n v ) o p ( L ′ , L ) = ( C C o n v ) ( L , L ′ ) \varphi\in(\mathrm{C}^{\mathrm{Conv}})^{\mathrm{op}}(L',L)=(\mathrm{C}^{\mathrm{Conv}})(L,L') φ ∈ ( C Conv ) op ( L ′ , L ) = ( C Conv ) ( L , L ′ ) to E ( φ ) = φ E(\varphi)=\varphi E ( φ ) = φ which is a morphism of ( C C a p ) ( L , L ′ ) (\mathrm{C}^{\mathrm{Cap}})(L,L') ( C Cap ) ( L , L ′ ) because
Λ lim L ′ F ( v ) = lim L ′ F ≤ φ ( lim L φ − 1 ( F ) ) = φ ( Λ lim L φ − 1 ( F ) ( v ) ) \Lambda_{\lim_{L'}}\mathcal{F}(v)=\lim_{L'}\mathcal{F}\leq\varphi\left(\lim_L\varphi^{-1}(\mathcal{F})\right)=\varphi\left(\Lambda_{\lim_L}\varphi^{-1}(\mathcal{F})(v)\right) Λ l i m L ′ F ( v ) = L ′ lim F ≤ φ ( L lim φ − 1 ( F ) ) = φ ( Λ l i m L φ − 1 ( F ) ( v ) )
for every v > 0 v>0 v > 0 .
This embedding has a right-adjoint C : ( C C a p ) o p → ( C C o n v ) o p C:(\mathrm{C}^{\mathrm{Cap}})^{\mathrm{op}}\to(\mathrm{C}^{\mathrm{Conv}})^{\mathrm{op}} C : ( C Cap ) op → ( C Conv ) op defined on objects by C ( L , Λ ) = ( L , lim C Λ ) C(L,\Lambda)=(L,\lim_{C\Lambda}) C ( L , Λ ) = ( L , lim C Λ ) where
lim C Λ F = Λ F ( 1 ) (36) \lim_{C\Lambda}\mathcal{F}=\Lambda\mathcal{F}(1) \tag*{(36)} C Λ lim F = Λ F ( 1 ) (36)
and a left-adjoint R : ( C C a p ) o p → ( C C o n v ) o p R : (\mathcal{C}^{\mathrm{Cap}})^{\mathrm{op}} \to(\mathcal{C}^{\mathrm{Conv}})^{\mathrm{op}} R : ( C Cap ) op → ( C Conv ) op defined on objects by R ( L , Λ ) = ( L , lim R Λ ) R(L,\Lambda) = (L,\lim_{R\Lambda}) R ( L , Λ ) = ( L , lim R Λ ) where
lim R Λ F = ⋁ v > 0 Λ F ( v ) , (37) \lim_{R\Lambda}\mathcal{F} = \bigvee_{v>0}\Lambda\mathcal{F}(v), \tag*{(37)} R Λ lim F = v > 0 ⋁ Λ F ( v ) , (37)
and both R R R and C C C are concrete functors. Moreover, the following diagram commutes (we leave the easy verifications to the reader):
C o n v → P ( C C o n v ) o p r ↓ R ↑ C a p → P ( C C a p ) o p c ↑ C ↓ \begin{CD}
\mathrm{Conv} @>{P}>> (\mathcal{C}^{\mathrm{Conv}})^{\mathrm{op}} \\
@V{r}VV @A{R}AA \\
\mathrm{Cap} @>{P}>> (\mathcal{C}^{\mathrm{Cap}})^{\mathrm{op}} \\
@A{c}AA @V{C}VV
\end{CD} Conv r ↓ ⏐ Cap c ⏐ ↑ P P ( C Conv ) op R ⏐ ↑ ( C Cap ) op C ↓ ⏐
On the other hand, it is important to note that the diagram no longer commutes when the point-functors are added.
C o n v → P ( C C o n v ) o p r ↓ R ↑ C a p → P ( C C a p ) o p c ↑ C ↓ C o n v → p t C o n v p t ↓ p t ↑ C a p → p t C a p \begin{CD}
\mathrm{Conv} @>{P}>> (\mathcal{C}^{\mathrm{Conv}})^{\mathrm{op}} \\
@V{r}VV @A{R}AA \\
\mathrm{Cap} @>{P}>> (\mathcal{C}^{\mathrm{Cap}})^{\mathrm{op}} \\
@A{c}AA @V{C}VV
\end{CD}
\qquad
\begin{CD}
\mathrm{Conv} @>{\mathrm{pt}}>> \mathrm{Conv} \\
@V{\mathrm{pt}}VV @A{\mathrm{pt}}AA \\
\mathrm{Cap} @>{\mathrm{pt}}>> \mathrm{Cap}
\end{CD} Conv r ↓ ⏐ Cap c ⏐ ↑ P P ( C Conv ) op R ⏐ ↑ ( C Cap ) op C ↓ ⏐ Conv pt ↓ ⏐ Cap pt pt Conv pt ⏐ ↑ Cap
For instance p t ∘ R ≠ r ∘ p t \mathrm{pt}\circ R \ne r\circ\mathrm{pt} pt ∘ R = r ∘ pt : starting with a C C a p \mathcal{C}^{\mathrm{Cap}} C Cap -object ( L , Λ ) (L,\Lambda) ( L , Λ ) , p t ∘ R ( L , Λ ) = p t ( L , lim R Λ ) \mathrm{pt}\circ R(L,\Lambda) = \mathrm{pt}(L,\lim_{R\Lambda}) pt ∘ R ( L , Λ ) = pt ( L , lim R Λ ) has { φ ∈ C ( L , 2 ) : lim R Λ φ − 1 ( { ∗ } ) = ⋁ v > 0 Λ φ − 1 ( { ∗ } ) ( v ) ∈ φ − 1 ( { ∗ } ) } \{\varphi\in C(L,2) : \lim_{R\Lambda}\varphi^{-1}(\{\ast\}) = \bigvee_{v>0}\Lambda\varphi^{-1}(\{\ast\})(v) \in\varphi^{-1}(\{\ast\})\} { φ ∈ C ( L , 2 ) : lim R Λ φ − 1 ({ ∗ }) = ⋁ v > 0 Λ φ − 1 ({ ∗ }) ( v ) ∈ φ − 1 ({ ∗ })} as underlying set, while r ∘ p t ( L , Λ ) r\circ\mathrm{pt}(L,\Lambda) r ∘ pt ( L , Λ ) has underlying set { φ ∈ C ( L , 2 ) : ∀ v ∈ V , Λ φ − 1 ( { ∗ } ) ( v ) ∈ φ − 1 ( { ∗ } ) } \{\varphi\in C(L,2) : \forall v \in V,\Lambda\varphi^{-1}(\{\ast\})(v) \in\varphi^{-1}(\{\ast\})\} { φ ∈ C ( L , 2 ) : ∀ v ∈ V , Λ φ − 1 ({ ∗ }) ( v ) ∈ φ − 1 ({ ∗ })} which is in general a strictly smaller set. However,
Proposition 24. Let C C C be an admissible category of frames or of coframes. Then
p t ∘ R ∘ P = r , \mathrm{pt}\circ R\circ P = r, pt ∘ R ∘ P = r , p t ∘ C ∘ P = c , \mathrm{pt}\circ C\circ P = c, pt ∘ C ∘ P = c , and
c ∘ p t ∘ E ∘ P = r ∘ p t ∘ E ∘ P = I C o n v . c\circ\mathrm{pt}\circ E\circ P = r\circ\mathrm{pt}\circ E\circ P = I_{\mathrm{Conv}}. c ∘ pt ∘ E ∘ P = r ∘ pt ∘ E ∘ P = I Conv . Proof. In view of Remark 4, starting with a Cap-object ( X , λ ) (X,\lambda) ( X , λ ) , the underlying set of p t ∘ R ∘ P ( X , λ ) \mathrm{pt}\circ R\circ P(X,\lambda) pt ∘ R ∘ P ( X , λ ) and of p t ∘ C ∘ P ( X , λ ) \mathrm{pt}\circ C\circ P(X,\lambda) pt ∘ C ∘ P ( X , λ ) is X X X . Moreover,
lim p t ∘ R ∘ P X F = lim R Λ λ F = { x ∈ X : λ F ( x ) > 0 } = lim r λ F . \lim_{\mathrm{pt}\circ R\circ P_X}\mathcal{F} = \lim_{R\Lambda_\lambda}\mathcal{F} = \{x\in X:\lambda_{\mathcal{F}}(x)>0\} = \lim_{r\lambda}\mathcal{F}. pt ∘ R ∘ P X lim F = R Λ λ lim F = { x ∈ X : λ F ( x ) > 0 } = r λ lim F . Similarly,
lim p t ∘ C ∘ P X F = lim C Λ λ F = { x ∈ X : λ F ( x ) = 1 } = lim c λ F . \lim_{\mathrm{pt}\circ C\circ P_X}\mathcal{F} = \lim_{C\Lambda_\lambda}\mathcal{F} = \{x\in X:\lambda_{\mathcal{F}}(x)=1\} = \lim_{c\lambda}\mathcal{F}. pt ∘ C ∘ P X lim F = C Λ λ lim F = { x ∈ X : λ F ( x ) = 1 } = c λ lim F . In view of Remark 16, the underlying set of p t ∘ E ∘ P \mathrm{pt}\circ E\circ P pt ∘ E ∘ P is X X X . Moreover, starting with a Conv-object ( X , ξ ) (X,\xi) ( X , ξ )
λ p t ∘ E ∘ P F ( x ) = ⋁ { v ∈ V : x ∈ Λ F ( v ) = lim ξ F } = { 1 x ∈ lim ξ F , 0 x ∉ lim ξ F , \lambda_{\mathrm{pt}\circ E\circ P}\mathcal{F}(x) = \bigvee\{v\in V:x\in\Lambda\mathcal{F}(v)=\lim_{\xi}\mathcal{F}\} =
\begin{cases}
1 & x\in\lim_{\xi}\mathcal{F},\\
0 & x\notin\lim_{\xi}\mathcal{F},
\end{cases} λ pt ∘ E ∘ P F ( x ) = ⋁ { v ∈ V : x ∈ Λ F ( v ) = ξ lim F } = { 1 0 x ∈ lim ξ F , x ∈ / lim ξ F , so that c ∘ p t ∘ E ∘ P ( X , ξ ) = r ∘ p t ∘ E ∘ P ( X , ξ ) = ( X , ξ ) c\circ\mathrm{pt}\circ E\circ P(X,\xi)=r\circ\mathrm{pt}\circ E\circ P(X,\xi)=(X,\xi) c ∘ pt ∘ E ∘ P ( X , ξ ) = r ∘ pt ∘ E ∘ P ( X , ξ ) = ( X , ξ ) . □
Adherence
Let us start by inspecting the adherence function
adh λ F ( x ) = ⋁ H # F λ H ( x ) = ⋁ U ∈ U ( F ) λ U ( x ) (38) \operatorname{adh}_{\lambda}\mathcal{F}(x)=\bigvee_{\mathcal{H}\#\mathcal{F}}\lambda\mathcal{H}(x)=\bigvee_{U\in\mathcal{U}(\mathcal{F})}\lambda U(x) \tag*{(38)} adh λ F ( x ) = H # F ⋁ λ H ( x ) = U ∈ U ( F ) ⋁ λ U ( x ) (38)
of a filter F \mathcal{F} F on a Cap-object ( X , λ ) (X,\lambda) ( X , λ ) , as defined in, e.g., [16 ] , which applies also to subsets of X X X via
adh λ A ( ⋅ ) = adh λ { A } ↑ ( ⋅ ) . \operatorname{adh}_{\lambda}A(\mathord{\cdot})=\operatorname{adh}_{\lambda}\{A\}^{\uparrow}(\mathord{\cdot}). adh λ A ( ⋅ ) = adh λ { A } ↑ ( ⋅ ) .
through the lens of P ( X , λ ) ∈ Ob ( C C a p ) \mathbb{P}(X,\lambda)\in\operatorname{Ob}(\mathsf{C}^{\mathrm{Cap}}) P ( X , λ ) ∈ Ob ( C Cap ) , hence seeking a function adh P X F : V → P X \operatorname{adh}_{\mathbb{P}X}\mathcal{F}:V\to\mathbb{P}X adh P X F : V → P X given by
adh P X F ( v ) = { x ∈ X : adh λ F ( x ) ≥ v } . \operatorname{adh}_{\mathbb{P}X}\mathcal{F}(v)=\{x\in X:\operatorname{adh}_{\lambda}\mathcal{F}(x)\geq v\}. adh P X F ( v ) = { x ∈ X : adh λ F ( x ) ≥ v } .
Since ⋁ H # F λ H ( x ) ≥ v \bigvee_{\mathcal{H}\#\mathcal{F}}\lambda\mathcal{H}(x)\geq v ⋁ H # F λ H ( x ) ≥ v if and only if for every ϵ < 1 \epsilon<1 ϵ < 1 there is H ϵ # F \mathcal{H}_{\epsilon}\#\mathcal{F} H ϵ # F with λ H ϵ ( x ) ≥ v ⊗ ϵ \lambda\mathcal{H}_{\epsilon}(x)\geq v\otimes\epsilon λ H ϵ ( x ) ≥ v ⊗ ϵ , equivalently, x ∈ Λ P X H ϵ ( v ⊗ ϵ ) x\in\Lambda_{\mathbb{P}X}\mathcal{H}_{\epsilon}(v\otimes\epsilon) x ∈ Λ P X H ϵ ( v ⊗ ϵ ) , we obtain
adh P X F ( v ) = ⋂ ϵ < 1 ⋃ H # F Λ P X H ( v ⊗ ϵ ) . \operatorname{adh}_{\mathbb{P}X}\mathcal{F}(v)=\bigcap_{\epsilon<1}\bigcup_{\mathcal{H}\#\mathcal{F}}\Lambda_{\mathbb{P}X}\mathcal{H}(v\otimes\epsilon). adh P X F ( v ) = ϵ < 1 ⋂ H # F ⋃ Λ P X H ( v ⊗ ϵ ) .
These considerations lead to the following:
Definition 25. If L L L is a C C a p \mathsf{C}^{\mathrm{Cap}} C Cap -object and F ∈ F L \mathcal{F}\in\mathbb{F}L F ∈ F L then the adherence adh L F : V → L \operatorname{adh}_{L}\mathcal{F}:V\to L adh L F : V → L of F \mathcal{F} F is given by
adh L F ( v ) = ⋀ ϵ < 1 ⋁ H ∈ F L , H # F Λ H ( v ⊗ ϵ ) . \operatorname{adh}_{L}\mathcal{F}(v)=\bigwedge_{\epsilon<1}\bigvee_{\mathcal{H}\in\mathbb{F}L,\mathcal{H}\#\mathcal{F}}\Lambda\mathcal{H}(v\otimes\epsilon). adh L F ( v ) = ϵ < 1 ⋀ H ∈ F L , H # F ⋁ Λ H ( v ⊗ ϵ ) .
Note that in particular adh L F ( 0 ) = ⊤ L \operatorname{adh}_{L}\mathcal{F}(0)=\top_{L} adh L F ( 0 ) = ⊤ L because Λ H ( 0 ⊗ ϵ ) = Λ H ( 0 ) = ⊤ L \Lambda\mathcal{H}(0\otimes\epsilon)=\Lambda\mathcal{H}(0)=\top_{L} Λ H ( 0 ⊗ ϵ ) = Λ H ( 0 ) = ⊤ L for all H \mathcal{H} H and ϵ \epsilon ϵ .
Proposition 26. Let C C C denote a subcategory of that of coframes. If F \mathcal{F} F and G \mathcal{G} G are filters on L ∈ Ob ( C C a p ) L\in\operatorname{Ob}(\mathsf{C}^{\mathrm{Cap}}) L ∈ Ob ( C Cap ) then
adh L ( F ∩ G ) ( v ) = adh L F ( v ) ∨ adh L G ( v ) . \operatorname{adh}_{L}(\mathcal{F}\cap\mathcal{G})(v)=\operatorname{adh}_{L}\mathcal{F}(v)\vee\operatorname{adh}_{L}\mathcal{G}(v). adh L ( F ∩ G ) ( v ) = adh L F ( v ) ∨ adh L G ( v ) . In particular , F ⊂ G \mathcal{F}\subset\mathcal{G} F ⊂ G implies adh L G ≤ adh L F \operatorname{adh}_{L}\mathcal{G}\leq\operatorname{adh}_{L}\mathcal{F} adh L G ≤ adh L F .
Proof. In view of Lemma 5, H # ( F ∩ G ) \mathcal{H}\#(\mathcal{F}\cap\mathcal{G}) H # ( F ∩ G ) if and only if H # F \mathcal{H}\#\mathcal{F} H # F or H # G \mathcal{H}\#\mathcal{G} H # G so that
adh L ( F ∩ G ) ( v ) = ⋀ ϵ < 1 ( ⋁ H ∈ F L , H # F Λ H ( v ⊗ ϵ ) ∨ ⋁ H ∈ F L , H # G Λ H ( v ⊗ ϵ ) ) \operatorname{adh}_{L}(\mathcal{F}\cap\mathcal{G})(v)=\bigwedge_{\epsilon<1}\left(\bigvee_{\mathcal{H}\in\mathbb{F}L,\mathcal{H}\#\mathcal{F}}\Lambda\mathcal{H}(v\otimes\epsilon)\vee\bigvee_{\mathcal{H}\in\mathbb{F}L,\mathcal{H}\#\mathcal{G}}\Lambda\mathcal{H}(v\otimes\epsilon)\right) adh L ( F ∩ G ) ( v ) = ϵ < 1 ⋀ H ∈ F L , H # F ⋁ Λ H ( v ⊗ ϵ ) ∨ H ∈ F L , H # G ⋁ Λ H ( v ⊗ ϵ )
which is adh L F ( v ) ∨ adh L G ( v ) \operatorname{adh}_{L}\mathcal{F}(v)\vee\operatorname{adh}_{L}\mathcal{G}(v) adh L F ( v ) ∨ adh L G ( v ) because of the coframe distributivity law. □
Let us note that (5.1) allows for alternative characterizations of objects of the full subcategories PsAp and PrAp of Cap (See, e.g., [14 ] ). Namely, ( X , λ ) (X,\lambda) ( X , λ ) is a pseudo-approach space if and only if
λ F = ⋀ H # F adh λ H , (39) \lambda\mathcal{F}=\bigwedge_{\mathcal{H}\#\mathcal{F}}\operatorname{adh}_{\lambda}\mathcal{H}, \tag*{(39)} λ F = H # F ⋀ adh λ H , (39)
for every filter F \mathcal{F} F on X X X , and a pre-approach space if and only if
λ F = ⋀ H ∈ F # adh λ H . (40) \lambda\mathcal{F}=\bigwedge_{\mathcal{H}\in\mathcal{F}^{\#}}\operatorname{adh}_{\lambda}\mathcal{H}. \tag*{(40)} λ F = H ∈ F # ⋀ adh λ H . (40)
Letting A ( ϵ ) = ( adh λ A ) − 1 [ ϵ , 1 ] A^{(\epsilon)}=(\operatorname{adh}_{\lambda}A)^{-1}[\epsilon,1] A ( ϵ ) = ( adh λ A ) − 1 [ ϵ , 1 ] for every A ⊂ X A\subset X A ⊂ X , a pre-approach space ( X , λ ) (X,\lambda) ( X , λ ) is an approach space if
adh λ A ( x ) ≥ adh λ A ( ϵ ) ( x ) ⊗ ϵ , (41) \operatorname{adh}_{\lambda}A(x)\geq\operatorname{adh}_{\lambda}A^{(\epsilon)}(x)\otimes\epsilon, \tag*{(41)} adh λ A ( x ) ≥ adh λ A ( ϵ ) ( x ) ⊗ ϵ , (41)
for every A ⊂ X A\subset X A ⊂ X , every x ∈ X x\in X x ∈ X and every ϵ ∈ V \epsilon\in V ϵ ∈ V .
Adherence of image filters
We seek to understand the effect of maps that may or may not be morphisms in the sense of (2.11). Translated into our version of Cap, the measure of continuity of f : X → Y f:X\to Y f : X → Y where ( X , λ X ) (X,\lambda_X) ( X , λ X ) and ( Y , λ Y ) (Y,\lambda_Y) ( Y , λ Y ) are two convergence approach spaces introduced in [3 ] under the name default of contraction is given by
m ( f ) = sup { v ∈ V : ∀ F ∈ F P X , λ Y ( f [ F ] ) ∘ f ( ⋅ ) ≥ v ⊗ λ X F ( ⋅ ) } . (42) m(f)=\sup\{v\in V:\ \forall\mathcal{F}\in\mathbb{F}PX,\ \lambda_Y(f[\mathcal{F}])\circ f(\cdot)\geq v\otimes\lambda_X\mathcal{F}(\cdot)\}. \tag*{(42)} m ( f ) = sup { v ∈ V : ∀ F ∈ F P X , λ Y ( f [ F ]) ∘ f ( ⋅ ) ≥ v ⊗ λ X F ( ⋅ )} . (42)
In other words, λ X F ( x ) ≥ w \lambda_X\mathcal{F}(x)\geq w λ X F ( x ) ≥ w implies λ Y ( f [ F ] ) ( f ( x ) ) ≥ v ⊗ w \lambda_Y(f[\mathcal{F}])(f(x))\geq v\otimes w λ Y ( f [ F ]) ( f ( x )) ≥ v ⊗ w for every v v v in the defining set of m ( f ) m(f) m ( f ) , that is,
Λ X F ( w ) ⊂ f − 1 ( Λ Y ( f [ F ] ) ( w ⊗ v ) ) \Lambda_X\mathcal{F}(w)\subset f^{-1}\bigl(\Lambda_Y(f[\mathcal{F}])(w\otimes v)\bigr) Λ X F ( w ) ⊂ f − 1 ( Λ Y ( f [ F ]) ( w ⊗ v ) )
for every such v v v . These considerations lead to:
Definition 28. The measure of continuity M ( φ ) M(\varphi) M ( φ ) of a map φ : L → L ′ \varphi:L\to L' φ : L → L ′ between two C a p \mathrm{C}^{\mathrm{ap}} C ap -objects is given by
M ( φ ) = ⋁ { v ∈ V : ∀ F ∈ F L ′ , Λ L ′ ( F ) ( ⋅ ) ≤ φ ( Λ L ( φ − 1 ( F ) ) ( ⋅ ⊗ v ) ) } . (43) M(\varphi)=\bigvee\{v\in V:\ \forall\mathcal{F}\in\mathbb{F}L',\ \Lambda_{L'}(\mathcal{F})(\cdot)\leq\varphi(\Lambda_L(\varphi^{-1}(\mathcal{F}))(\cdot\otimes v))\}. \tag*{(43)} M ( φ ) = ⋁ { v ∈ V : ∀ F ∈ F L ′ , Λ L ′ ( F ) ( ⋅ ) ≤ φ ( Λ L ( φ − 1 ( F )) ( ⋅ ⊗ v ))} . (43)
Remark 29. Note that in view of (1.13) and (2.10), letting F = { v ∈ V : ∀ F ∈ F L ′ : Λ L ′ ( F ) ( ⋅ ) ≤ φ ( Λ L ( φ − 1 ( F ) ) ( ⋅ ⊗ v ) ) } F=\{v\in V:\ \forall\mathcal{F}\in\mathbb{F}L':\Lambda_{L'}(\mathcal{F})(\cdot)\leq\varphi(\Lambda_L(\varphi^{-1}(\mathcal{F}))(\cdot\otimes v))\} F = { v ∈ V : ∀ F ∈ F L ′ : Λ L ′ ( F ) ( ⋅ ) ≤ φ ( Λ L ( φ − 1 ( F )) ( ⋅ ⊗ v ))} ,
Λ L ( φ − 1 ( F ) ) ( ⋅ ⊗ M ( φ ) ) = Λ L ( φ − 1 ( F ) ) ( ⋅ ⊗ ⋁ F ) = ⋀ v ∈ F Λ L ( φ − 1 ( F ) ) ( ⋅ ⊗ v ) ≥ Λ L ′ ( F ) ( ⋅ ) \Lambda_L(\varphi^{-1}(\mathcal{F}))(\cdot\otimes M(\varphi))=\Lambda_L(\varphi^{-1}(\mathcal{F}))(\cdot\otimes\bigvee F)=\bigwedge_{v\in F}\Lambda_L(\varphi^{-1}(\mathcal{F}))(\cdot\otimes v)\geq\Lambda_{L'}(\mathcal{F})(\cdot) Λ L ( φ − 1 ( F )) ( ⋅ ⊗ M ( φ )) = Λ L ( φ − 1 ( F )) ( ⋅ ⊗ ⋁ F ) = v ∈ F ⋀ Λ L ( φ − 1 ( F )) ( ⋅ ⊗ v ) ≥ Λ L ′ ( F ) ( ⋅ )
for every F ∈ F L ′ \mathcal{F}\in\mathbb{F}L' F ∈ F L ′ so that M ( φ ) ∈ F M(\varphi)\in F M ( φ ) ∈ F is the largest v ∈ V v\in V v ∈ V satisfying the defining condition of F F F .
It is plain that the morphism condition (2.11) is satisfied if and only if M ( φ ) = 1 M(\varphi)=1 M ( φ ) = 1 .
Theorem 30. If C C C is a category of coframes, φ ∈ C ( L , L ′ ) \varphi\in C(L,L') φ ∈ C ( L , L ′ ) , and F ∈ F L ′ \mathcal{F}\in\mathbb{F}L' F ∈ F L ′ , then
adh L ′ F ( ⋅ ) ≤ φ ( adh L φ ℓ ( F ) ( ⋅ ⊗ M ( φ ) ) ) ≤ φ ( adh L φ − 1 ( F ) ( ⋅ ⊗ M ( φ ) ) ) . \operatorname{adh}_{L'}\mathcal{F}(\cdot)\leq\varphi(\operatorname{adh}_L\varphi_{\ell}(\mathcal{F})(\cdot\otimes M(\varphi)))\leq\varphi(\operatorname{adh}_L\varphi^{-1}(\mathcal{F})(\cdot\otimes M(\varphi))). adh L ′ F ( ⋅ ) ≤ φ ( adh L φ ℓ ( F ) ( ⋅ ⊗ M ( φ ))) ≤ φ ( adh L φ − 1 ( F ) ( ⋅ ⊗ M ( φ ))) . Proof. For each H ∈ F L ′ \mathcal{H}\in\mathbb{F}L' H ∈ F L ′ with H # F \mathcal{H}\#\mathcal{F} H # F , φ ℓ ( F ) # φ − 1 ( H ) \varphi_{\ell}(\mathcal{F})\#\varphi^{-1}(\mathcal{H}) φ ℓ ( F ) # φ − 1 ( H ) by (1.10). Hence,
Λ L ′ H ( w ⊗ ϵ ) ≤ φ ( Λ L φ − 1 ( H ) ( w ⊗ M ( φ ) ⊗ ϵ ) ) ≤ φ ( ⋁ G ∈ F L φ ℓ ( F ) # G Λ L G ( w ⊗ M ( φ ) ⊗ ϵ ) ) . \Lambda_{L'}\mathcal{H}(w\otimes\epsilon)\leq\varphi(\Lambda_L\varphi^{-1}(\mathcal{H})(w\otimes M(\varphi)\otimes\epsilon))\leq\varphi\left(\bigvee_{\substack{\mathcal{G}\in\mathbb{F}L\\\varphi_{\ell}(\mathcal{F})\#\mathcal{G}}}\Lambda_L\mathcal{G}(w\otimes M(\varphi)\otimes\epsilon)\right). Λ L ′ H ( w ⊗ ϵ ) ≤ φ ( Λ L φ − 1 ( H ) ( w ⊗ M ( φ ) ⊗ ϵ )) ≤ φ G ∈ F L φ ℓ ( F ) # G ⋁ Λ L G ( w ⊗ M ( φ ) ⊗ ϵ ) .
Hence, taking the supremum over all filters H \mathcal{H} H meshing with F \mathcal{F} F and using (1.13) yields
⋁ H ∈ F L , H # F Λ L ′ H ( w ⊗ ϵ ) ≤ φ ( ⋁ G ∈ F L φ ℓ ( F ) # G Λ L G ( w ⊗ M ( φ ) ⊗ ϵ ) ) . \bigvee_{\substack{\mathcal{H}\in\mathbb{F}L,\mathcal{H}\#\mathcal{F}}}\Lambda_{L'}\mathcal{H}(w\otimes\epsilon)\leq\varphi\left(\bigvee_{\substack{\mathcal{G}\in\mathbb{F}L\\\varphi_{\ell}(\mathcal{F})\#\mathcal{G}}}\Lambda_L\mathcal{G}(w\otimes M(\varphi)\otimes\epsilon)\right). H ∈ F L , H # F ⋁ Λ L ′ H ( w ⊗ ϵ ) ≤ φ G ∈ F L φ ℓ ( F ) # G ⋁ Λ L G ( w ⊗ M ( φ ) ⊗ ϵ ) .
Finally, taking the infimum over all ϵ < 1 \epsilon<1 ϵ < 1 yields
⋀ ϵ < 1 ⋁ H ∈ F L , H # F Λ L ′ H ( w ⊗ ϵ ) ≤ φ ( ⋀ ϵ < 1 ⋁ G ∈ F L φ ℓ ( F ) # G Λ L G ( w ⊗ M ( φ ) ⊗ ϵ ) ) . \bigwedge_{\epsilon<1}\bigvee_{\substack{\mathcal{H}\in\mathbb{F}L,\mathcal{H}\#\mathcal{F}}}\Lambda_{L'}\mathcal{H}(w\otimes\epsilon)\leq\varphi\left(\bigwedge_{\epsilon<1}\bigvee_{\substack{\mathcal{G}\in\mathbb{F}L\\\varphi_{\ell}(\mathcal{F})\#\mathcal{G}}}\Lambda_L\mathcal{G}(w\otimes M(\varphi)\otimes\epsilon)\right). ϵ < 1 ⋀ H ∈ F L , H # F ⋁ Λ L ′ H ( w ⊗ ϵ ) ≤ φ ϵ < 1 ⋀ G ∈ F L φ ℓ ( F ) # G ⋁ Λ L G ( w ⊗ M ( φ ) ⊗ ϵ ) .
because φ \varphi φ preserves infima, and the conclusion follows. The rightmost inequality follows from φ − 1 ( F ) ⊂ φ ℓ ( F ) \varphi^{-1}(\mathcal{F})\subset\varphi_{\ell}(\mathcal{F}) φ − 1 ( F ) ⊂ φ ℓ ( F ) because φ ℓ ( φ ( m ) ) ≤ m \varphi_{\ell}(\varphi(m))\leq m φ ℓ ( φ ( m )) ≤ m .
Corollary 31. If C C C is a category of coframes, φ ∈ C C a p ( L , L ′ ) \varphi\in\mathrm{C}^{\mathrm{Cap}}(L,L') φ ∈ C Cap ( L , L ′ ) , and F ∈ F L ′ \mathcal{F}\in\mathbb{F}L' F ∈ F L ′ , then
adh L ′ F ( ⋅ ) ≤ φ ( adh L φ ℓ ( F ) ( ⋅ ) ) ≤ φ ( adh L φ − 1 ( F ) ( ⋅ ) ) . \operatorname{adh}_{L'}\mathcal{F}(\cdot)\leq\varphi(\operatorname{adh}_{L}\varphi_{\ell}(\mathcal{F})(\cdot))\leq\varphi(\operatorname{adh}_{L}\varphi^{-1}(\mathcal{F})(\cdot)). adh L ′ F ( ⋅ ) ≤ φ ( adh L φ ℓ ( F ) ( ⋅ )) ≤ φ ( adh L φ − 1 ( F ) ( ⋅ )) .
Note that Theorem 30 translated in Cap yields a more general result than [16 (1.8)] , which applies to functions that may not be morphisms. Namely:
Corollary 32. If ( X , λ X ) (X,\lambda_X) ( X , λ X ) and ( Y , λ Y ) (Y,\lambda_Y) ( Y , λ Y ) are two convergence approach spaces, f : X → Y f:X\to Y f : X → Y and G ∈ F P Y \mathcal{G}\in\mathbb{F}PY G ∈ F P Y then
m ( f ) ⊗ adh λ X f − [ G ] ( x ) ≤ adh λ Y G ( f ( x ) ) m(f)\otimes\operatorname{adh}_{\lambda_X}f^{-}[\mathcal{G}](x)\leq\operatorname{adh}_{\lambda_Y}\mathcal{G}(f(x)) m ( f ) ⊗ adh λ X f − [ G ] ( x ) ≤ adh λ Y G ( f ( x )) for every x ∈ X x\in X x ∈ X .
Maximal filters
Assuming the axiom of choice, every filter on a bounded lattice L L L is contained in a maximal filter. Let us denote by U L UL U L the set of maximal filters on L L L and if F ∈ F L \mathcal{F}\in\mathbb{F}L F ∈ F L , let β F = { u ∈ U L : F ⊂ u } \beta\mathcal{F}=\{u\in UL:\mathcal{F}\subset u\} β F = { u ∈ U L : F ⊂ u } . If G , F ∈ F L \mathcal{G},\mathcal{F}\in\mathbb{F}L G , F ∈ F L and G # F \mathcal{G}\#\mathcal{F} G # F then { g ∧ f : g ∈ G , f ∈ F } \{g\wedge f:g\in\mathcal{G},f\in\mathcal{F}\} { g ∧ f : g ∈ G , f ∈ F } is a filter-base and there is the least filter F ∨ G \mathcal{F}\vee\mathcal{G} F ∨ G containing both F \mathcal{F} F and G \mathcal{G} G and β ( F ∨ G ) = β F ∩ β G \beta(\mathcal{F}\vee\mathcal{G})=\beta\mathcal{F}\cap\beta\mathcal{G} β ( F ∨ G ) = β F ∩ β G .
Lemma 33. If ( L , Λ ) (L,\Lambda) ( L , Λ ) is a C C a p \mathrm{C}^{\mathrm{Cap}} C Cap -object and U ∈ U L \mathcal{U}\in UL U ∈ U L then
adh Λ U = Λ U . \operatorname{adh}_{\Lambda}\mathcal{U}=\Lambda\mathcal{U}. adh Λ U = Λ U . If F ∈ F L \mathcal{F}\in\mathbb{F}L F ∈ F L then
adh Λ F = ⋀ ϵ < 1 ⋁ U ∈ β F Λ U ( v ⊗ ϵ ) . \operatorname{adh}_{\Lambda}\mathcal{F}=\bigwedge_{\epsilon<1}\bigvee_{\mathcal{U}\in\beta\mathcal{F}}\Lambda\mathcal{U}(v\otimes\epsilon). adh Λ F = ϵ < 1 ⋀ U ∈ β F ⋁ Λ U ( v ⊗ ϵ ) .
Proof. Because G # U \mathcal{G}\#\mathcal{U} G # U if and only if G ⊂ U \mathcal{G}\subset\mathcal{U} G ⊂ U by maximality,
Λ U ( v ) ≤ adh Λ U ( v ) = ⋀ ϵ < 1 ⋁ G # U Λ G ( v ⊗ ϵ ) ≤ ⋀ ϵ < 1 Λ U ( v ⊗ ϵ ) = Λ U ( ⋁ ϵ < 1 v ⊗ ϵ ) = Λ U ( v ) . \Lambda\mathcal{U}(v)\leq\operatorname{adh}_{\Lambda}\mathcal{U}(v)=\bigwedge_{\epsilon<1}\bigvee_{\mathcal{G}\#\mathcal{U}}\Lambda\mathcal{G}(v\otimes\epsilon)\leq\bigwedge_{\epsilon<1}\Lambda\mathcal{U}(v\otimes\epsilon)=\Lambda\mathcal{U}\left(\bigvee_{\epsilon<1}v\otimes\epsilon\right)=\Lambda\mathcal{U}(v). Λ U ( v ) ≤ adh Λ U ( v ) = ϵ < 1 ⋀ G # U ⋁ Λ G ( v ⊗ ϵ ) ≤ ϵ < 1 ⋀ Λ U ( v ⊗ ϵ ) = Λ U ( ϵ < 1 ⋁ v ⊗ ϵ ) = Λ U ( v ) .
The second formula follows from the observation that ⋁ G # F Λ G ( v ⊗ ϵ ) = ⋁ U ∈ β F Λ U ( v ⊗ ϵ ) \bigvee_{\mathcal{G}\#\mathcal{F}}\Lambda\mathcal{G}(v\otimes\epsilon)=\bigvee_{\mathcal{U}\in\beta\mathcal{F}}\Lambda\mathcal{U}(v\otimes\epsilon) ⋁ G # F Λ G ( v ⊗ ϵ ) = ⋁ U ∈ β F Λ U ( v ⊗ ϵ ) because for every G # F \mathcal{G}\#\mathcal{F} G # F there is U ∈ β F \mathcal{U}\in\beta\mathcal{F} U ∈ β F with U ⊃ G \mathcal{U}\supset\mathcal{G} U ⊃ G . □
Pseudo-approach, Pre-approach and Approach Spaces within C^{Cap}
Recalling Remark 13, we set out to characterize the images of PsAp, PrAp and Ap under P \mathbb{P} P .
Depth and adherence-determinedness
The introduction of these notions is motivated by Theorem 37 .
Let us fix a category C of lattices. If D \mathbb{D} D is a class of collections of filters, determined by its instantiations D ( L ) ⊂ P ( F L ) \mathbb{D}(L) \subset\mathcal{P}(\mathbb{F}L) D ( L ) ⊂ P ( F L ) for every C-object L L L , we say that a C C a p \mathrm{C}^{\mathrm{Cap}} C Cap -object ( L , Λ ) (L,\Lambda) ( L , Λ ) is D \mathbb{D} D -deep if
Λ ( ⋂ D ∈ D D ) = ⋀ D ∈ D Λ D (44) \Lambda\left(\bigcap_{\mathbb{D}\in\mathbb{D}}\mathbb{D}\right)=\bigwedge_{\mathbb{D}\in\mathbb{D}}\Lambda\mathbb{D} \tag*{(44)} Λ ( D ∈ D ⋂ D ) = D ∈ D ⋀ Λ D (44)
for every D ∈ D ( L ) \mathbb{D}\in\mathbb{D}(L) D ∈ D ( L ) . A set D ⊂ F L \mathbb{D}\subset\mathbb{F}L D ⊂ F L is finitely covering if for every choice ( ℓ D ) D ∈ D ∈ ∏ D ∈ D D (\ell_{\mathbb{D}})_{\mathbb{D}\in\mathbb{D}}\in\prod_{\mathbb{D}\in\mathbb{D}}\mathbb{D} ( ℓ D ) D ∈ D ∈ ∏ D ∈ D D , there is a finite subset D 0 \mathbb{D}_0 D 0 of D \mathbb{D} D with ⋁ D ∈ D 0 ℓ D ∈ ⋂ D ∈ D D \bigvee_{\mathbb{D}\in\mathbb{D}_0}\ell_{\mathbb{D}}\in\bigcap_{\mathbb{D}\in\mathbb{D}}\mathbb{D} ⋁ D ∈ D 0 ℓ D ∈ ⋂ D ∈ D D . Note in particular that when L = P X L=\mathcal{P}X L = P X and F ∈ F P X \mathcal{F}\in\mathbb{F}\mathcal{P}X F ∈ F P X , then β ( F ) \beta(\mathcal{F}) β ( F ) is finitely covering. More generally,
Lemma 34. If a family D \mathbb{D} D of filters on a bounded distributive lattice L L L is finitely covering then
β ( ⋂ D ∈ D D ) = ⋃ D ∈ D β D , \beta\left(\bigcap_{\mathbb{D}\in\mathbb{D}}\mathbb{D}\right)=\bigcup_{\mathbb{D}\in\mathbb{D}}\beta\mathbb{D}, β ( D ∈ D ⋂ D ) = D ∈ D ⋃ β D , and conversely if L L L is a Boolean Algebra.
Proof. The inclusion ⋃ D ∈ D β D ⊂ β ( ⋂ D ∈ D D ) \bigcup_{\mathbb{D}\in\mathbb{D}}\beta\mathbb{D}\subset\beta\left(\bigcap_{\mathbb{D}\in\mathbb{D}}\mathbb{D}\right) ⋃ D ∈ D β D ⊂ β ( ⋂ D ∈ D D ) is always true. Suppose D \mathbb{D} D is finitely covering and let F = ⋂ D ∈ D D \mathcal{F}=\bigcap_{\mathbb{D}\in\mathbb{D}}\mathbb{D} F = ⋂ D ∈ D D . If U ∈ β F ∖ ⋃ D ∈ D β D \mathcal{U}\in\beta\mathcal{F}\setminus\bigcup_{\mathbb{D}\in\mathbb{D}}\beta\mathbb{D} U ∈ β F ∖ ⋃ D ∈ D β D then for each D ∈ D \mathbb{D}\in\mathbb{D} D ∈ D there is ℓ D ∈ D ∖ U \ell_{\mathbb{D}}\in\mathbb{D}\setminus\mathcal{U} ℓ D ∈ D ∖ U . There is a finite subset D 0 \mathbb{D}_0 D 0 of D \mathbb{D} D with ⋁ D ∈ D 0 ℓ D ∈ F ⊂ U \bigvee_{\mathbb{D}\in\mathbb{D}_0}\ell_{\mathbb{D}}\in\mathcal{F}\subset\mathcal{U} ⋁ D ∈ D 0 ℓ D ∈ F ⊂ U . Since L L L is a bounded distributive lattice, maximal filters, in particular U \mathcal{U} U , are prime. Hence, there is ℓ D ∈ U \ell_{\mathbb{D}}\in\mathcal{U} ℓ D ∈ U , which is a contradiction.
Conversely, assume β ( ⋂ D ∈ D D ) = ⋃ D ∈ D β D \beta\left(\bigcap_{\mathbb{D}\in\mathbb{D}}\mathbb{D}\right)=\bigcup_{\mathbb{D}\in\mathbb{D}}\beta\mathbb{D} β ( ⋂ D ∈ D D ) = ⋃ D ∈ D β D and that L L L is a Boolean algebra. Pick ℓ D ∈ D \ell_{\mathbb{D}}\in\mathbb{D} ℓ D ∈ D for each D ∈ D \mathbb{D}\in\mathbb{D} D ∈ D . Let F = ⋂ D ∈ D D \mathcal{F}=\bigcap_{\mathbb{D}\in\mathbb{D}}\mathbb{D} F = ⋂ D ∈ D D . If ⋁ D ∈ D 0 ℓ D ∉ F \bigvee_{\mathbb{D}\in\mathbb{D}_0}\ell_{\mathbb{D}}\notin\mathcal{F} ⋁ D ∈ D 0 ℓ D ∈ / F whenever D 0 \mathbb{D}_0 D 0 is a finite subset of D \mathbb{D} D , then ( ⋁ D ∈ D 0 ℓ D ) ∗ ∈ F # \left(\bigvee_{\mathbb{D}\in\mathbb{D}_0}\ell_{\mathbb{D}}\right)^*\in\mathcal{F}^{\#} ( ⋁ D ∈ D 0 ℓ D ) ∗ ∈ F # and { ( ⋁ D ∈ D 0 ℓ D ) ∗ ∧ f : f ∈ F , D 0 ∈ [ D ] < ∞ } \left\{\left(\bigvee_{\mathbb{D}\in\mathbb{D}_0}\ell_{\mathbb{D}}\right)^*\wedge f:f\in\mathcal{F},\mathbb{D}_0\in[\mathbb{D}]^{<\infty}\right\} { ( ⋁ D ∈ D 0 ℓ D ) ∗ ∧ f : f ∈ F , D 0 ∈ [ D ] < ∞ } is a filter-base and there is U ∈ β F \mathcal{U}\in\beta\mathcal{F} U ∈ β F with ( ⋁ D ∈ D 0 ℓ D ) ∗ = ⋀ D ∈ D 0 ℓ D ∗ ∈ U \left(\bigvee_{\mathbb{D}\in\mathbb{D}_0}\ell_{\mathbb{D}}\right)^*=\bigwedge_{\mathbb{D}\in\mathbb{D}_0}\ell_{\mathbb{D}}^*\in\mathcal{U} ( ⋁ D ∈ D 0 ℓ D ) ∗ = ⋀ D ∈ D 0 ℓ D ∗ ∈ U for every D 0 ∈ [ D ] < ∞ \mathbb{D}_0\in[\mathbb{D}]^{<\infty} D 0 ∈ [ D ] < ∞ , so that ℓ D ∗ ∈ U \ell_{\mathbb{D}}^*\in\mathcal{U} ℓ D ∗ ∈ U for all D ∈ D \mathbb{D}\in\mathbb{D} D ∈ D and thus U ∉ ⋃ D ∈ D β D = β F \mathcal{U}\notin\bigcup_{\mathbb{D}\in\mathbb{D}}\beta\mathbb{D}=\beta\mathcal{F} U ∈ / ⋃ D ∈ D β D = β F , which is a contradiction. □
We say that ( L , Λ ) (L,\Lambda) ( L , Λ ) is deep if it is P ( F L ) \mathcal{P}(\mathbb{F}L) P ( F L ) -deep, and fc-deep if it is deep with respect to the class D \mathbb{D} D of finitely covering sets of filters.
On the other hand, if D \mathbb{D} D is a class of filters determined by its instantiations D ( L ) ∈ P ( F L ) \mathbb{D}(L)\in\mathcal{P}(\mathbb{F}L) D ( L ) ∈ P ( F L ) , we say that a C C a p \mathrm{C}^{\mathrm{Cap}} C Cap -object ( L , Λ ) (L,\Lambda) ( L , Λ ) is D \mathbb{D} D -adherence determined if
Λ ( F ) = ⋀ D ( L ) ∋ D # F adh Λ D (45) \Lambda(\mathcal{F})=\bigwedge_{\mathbb{D}(L)\ni\mathbb{D}\#\mathcal{F}}\operatorname{adh}_{\Lambda}\mathbb{D} \tag*{(45)} Λ ( F ) = D ( L ) ∋ D # F ⋀ adh Λ D (45)
for every F ∈ F L \mathcal{F}\in\mathbb{F}L F ∈ F L . Given C and a class D \mathbb{D} D of filters and a class D \mathbb{D} D of collections of filters as above, let us denote by Adh D \operatorname{Adh}_{\mathbb{D}} Adh D and by Deep D \operatorname{Deep}_{\mathbb{D}} Deep D the full subcategories of C C a p \mathrm{C}^{\mathrm{Cap}} C Cap whose objects are D \mathbb{D} D -adherence determined and D \mathbb{D} D -deep respectively.
Let F \mathbb{F} F denote the class of all filters and F 0 \mathbb{F}_0 F 0 the class of principal filters, so that F 0 L \mathbb{F}_0L F 0 L is the set of principal filters on L L L .
Proposition 35. Let ( L , Λ ) (L,\Lambda) ( L , Λ ) be a C C a p \mathrm{C}^{\mathrm{Cap}} C Cap -object.
If L L L is frame that is F 0 \mathbb{F}_0 F 0 -adherence determined, then L L L is deep.
If a bounded distributive lattice is F \mathbb{F} F -adherence determined then it is fc-deep, and conversely in a Boolean algebra.
Proof. (1). In a frame, ( ⋂ D ∈ D D ) # = ⋃ D ∈ D D # (\bigcap_{\mathcal{D}\in\mathbb{D}} \mathcal{D})^\#=\bigcup_{\mathcal{D}\in\mathbb{D}}\mathcal{D}^\# ( ⋂ D ∈ D D ) # = ⋃ D ∈ D D # , so that when L L L is F 0 \mathbb{F}_0 F 0 -determined, then
Λ ( ⋂ D ∈ D D ) = ⋀ ℓ ∈ ( ⋂ D ∈ D D ) # adh Λ ℓ = ⋀ ℓ ∈ ⋃ D ∈ D D # adh Λ ℓ = ⋀ D ∈ D ⋀ ℓ ∈ D # adh Λ ℓ = ⋀ D ∈ D Λ D , \Lambda\left(\bigcap_{\mathcal{D}\in\mathbb{D}}\mathcal{D}\right)=\bigwedge_{\ell\in(\bigcap_{\mathcal{D}\in\mathbb{D}}\mathcal{D})^\#}\operatorname{adh}_{\Lambda}\ell=\bigwedge_{\ell\in\bigcup_{\mathcal{D}\in\mathbb{D}}\mathcal{D}^\#}\operatorname{adh}_{\Lambda}\ell=\bigwedge_{\mathcal{D}\in\mathbb{D}}\bigwedge_{\ell\in\mathcal{D}^\#}\operatorname{adh}_{\Lambda}\ell=\bigwedge_{\mathcal{D}\in\mathbb{D}}\Lambda\mathcal{D}, Λ ( D ∈ D ⋂ D ) = ℓ ∈ ( ⋂ D ∈ D D ) # ⋀ adh Λ ℓ = ℓ ∈ ⋃ D ∈ D D # ⋀ adh Λ ℓ = D ∈ D ⋀ ℓ ∈ D # ⋀ adh Λ ℓ = D ∈ D ⋀ Λ D ,
and L L L is deep.
(2). In view of Lemma 33, ( L , Λ ) (L,\Lambda) ( L , Λ ) is F \mathbb{F} F -adherence determined if and only if
Λ ( F ) = ⋀ H # F adh Λ H = ⋀ U ∈ β F Λ ( U ) . (46) \Lambda(\mathcal{F})=\bigwedge_{\mathcal{H}\#\mathcal{F}}\operatorname{adh}_{\Lambda}\mathcal{H}=\bigwedge_{U\in\beta\mathcal{F}}\Lambda(U). \tag*{(46)} Λ ( F ) = H # F ⋀ adh Λ H = U ∈ β F ⋀ Λ ( U ) . (46)
If D \mathbb{D} D is finitely covering and ( L , Λ ) (L,\Lambda) ( L , Λ ) is F \mathbb{F} F -adherence determined then, letting F = ⋂ D ∈ D D \mathcal{F}=\bigcap_{\mathcal{D}\in\mathbb{D}}\mathcal{D} F = ⋂ D ∈ D D , Lemma 34 applies to the effect that
Λ F = ⋀ U ∈ β F = ⋃ D ∈ D β D Λ U = ⋀ D ∈ D ⋀ U ∈ β D Λ U = ⋀ D ∈ D Λ D , \Lambda\mathcal{F}=\bigwedge_{U\in\beta\mathcal{F}=\bigcup_{\mathcal{D}\in\mathbb{D}}\beta\mathcal{D}}\Lambda U=\bigwedge_{\mathcal{D}\in\mathbb{D}}\bigwedge_{U\in\beta\mathcal{D}}\Lambda U=\bigwedge_{\mathcal{D}\in\mathbb{D}}\Lambda\mathcal{D}, Λ F = U ∈ β F = ⋃ D ∈ D β D ⋀ Λ U = D ∈ D ⋀ U ∈ β D ⋀ Λ U = D ∈ D ⋀ Λ D ,
and ( L , Λ ) (L,\Lambda) ( L , Λ ) is fc-deep.
Conversely, if ( L , Λ ) (L,\Lambda) ( L , Λ ) is fc-deep and L L L is a Boolean algebra, then each set β F \beta\mathcal{F} β F is finitely covering and thus (6.3). □
Remark 36. When L = P X L=\mathcal{P}X L = P X , the converse to Proposition 35 (1) is true. Indeed, if ( P X , Λ ) (\mathcal{P}X,\Lambda) ( P X , Λ ) is deep, x ∈ X x\in X x ∈ X and w ∈ V w\in V w ∈ V , let N w ( x ) = ⋂ { D ∈ F P X : x ∈ Λ D ( w ) } \mathcal{N}_w(x)=\bigcap\{\mathcal{D}\in\mathbb{F}\mathcal{P}X:x\in\Lambda\mathcal{D}(w)\} N w ( x ) = ⋂ { D ∈ F P X : x ∈ Λ D ( w )} . It is easily checked that
x ∈ Λ F ( w ) ⟺ N w ( x ) ⊂ F ⟺ F # ⊂ N w ( x ) # . x\in\Lambda\mathcal{F}(w)\Longleftrightarrow\mathcal{N}_w(x)\subset\mathcal{F}\Longleftrightarrow\mathcal{F}^\#\subset\mathcal{N}_w(x)^\#. x ∈ Λ F ( w ) ⟺ N w ( x ) ⊂ F ⟺ F # ⊂ N w ( x ) # .
Hence, if
x ∈ ⋂ A ∈ F # adh Λ A ( v ) = ⋂ A ∈ F # ⋃ ϵ < 1 ⋃ H # A Λ H ( v ⊗ ϵ ) x\in\bigcap_{A\in\mathcal{F}^\#}\operatorname{adh}_{\Lambda}A(v)=\bigcap_{A\in\mathcal{F}^\#}\bigcup_{\epsilon<1}\bigcup_{\mathcal{H}\# A}\Lambda\mathcal{H}(v\otimes\epsilon) x ∈ A ∈ F # ⋂ adh Λ A ( v ) = A ∈ F # ⋂ ϵ < 1 ⋃ H # A ⋃ Λ H ( v ⊗ ϵ )
then for every A ∈ F # A\in\mathcal{F}^\# A ∈ F # and ϵ < 1 \epsilon<1 ϵ < 1 , there is H # A \mathcal{H}\# A H # A with x ∈ Λ H ( v ⊗ ϵ ) x\in\Lambda\mathcal{H}(v\otimes\epsilon) x ∈ Λ H ( v ⊗ ϵ ) , so that A ∈ H # ⊂ ( N v ⊗ ϵ ( x ) ) # A\in\mathcal{H}^\#\subset(\mathcal{N}_{v\otimes\epsilon}(x))^\# A ∈ H # ⊂ ( N v ⊗ ϵ ( x ) ) # . Thus F # ⊂ ( N v ⊗ ϵ ( x ) ) # \mathcal{F}^\#\subset(\mathcal{N}_{v\otimes\epsilon}(x))^\# F # ⊂ ( N v ⊗ ϵ ( x ) ) # for every ϵ < 1 \epsilon<1 ϵ < 1 , that is, x ∈ ⋂ ϵ < 1 Λ F ( v ⊗ ϵ ) = Λ F ( v ) x\in\bigcap_{\epsilon<1}\Lambda\mathcal{F}(v\otimes\epsilon)=\Lambda\mathcal{F}(v) x ∈ ⋂ ϵ < 1 Λ F ( v ⊗ ϵ ) = Λ F ( v ) .
In contrast a deep C H C a p \mathrm{CH}^{\mathrm{Cap}} CH Cap -object need not be F 0 \mathbb{F}_0 F 0 -adherence determined, in fact not even F \mathbb{F} F -adherence determined, even if weakly spatial. Consider for instance, the Heyting algebra L = { 0 , d , 1 } L=\{0,d,1\} L = { 0 , d , 1 } with 0 < d < 1 0<d<1 0 < d < 1 and d ∗ = 0 d^*=0 d ∗ = 0 . There are only two proper filters, ↑ 1 \uparrow1 ↑ 1 and ↑ d \uparrow d ↑ d and ↑ 1 ⊊ ↑ d \uparrow1\subsetneq\uparrow d ↑ 1 ⊊ ↑ d . Let Λ \Lambda Λ be given by Λ ( ↑ 1 ) ( v ) = d \Lambda(\uparrow1)(v)=d Λ ( ↑ 1 ) ( v ) = d and Λ ( ↑ d ) ( v ) = 1 \Lambda(\uparrow d)(v)=1 Λ ( ↑ d ) ( v ) = 1 for every v > 0 v>0 v > 0 , which defines a weakly spatial C H C a p \mathrm{CH}^{\mathrm{Cap}} CH Cap -object. The only point is ↑ d \uparrow d ↑ d so that pt L \operatorname{pt}L pt L is a singleton and is thus trivially an approach space. However, L L L fails to be even F \mathbb{F} F -adherence determined because Λ ( ↑ 1 ) ( 1 ) = d < 1 = ⋁ H # ↑ 1 adh Λ H ( 1 ) \Lambda(\uparrow1)(1)=d<1=\bigvee_{\mathcal{H}\#\uparrow1}\operatorname{adh}_{\Lambda}\mathcal{H}(1) Λ ( ↑ 1 ) ( 1 ) = d < 1 = ⋁ H # ↑ 1 adh Λ H ( 1 ) .
Images of Cap, PsAp, PrAp, and Ap under P \mathbb{P} P
Theorem 37. A C C a p \mathrm{C}^{\mathrm{Cap}} C Cap -object ( P X , Λ ) (\mathcal{P}X,\Lambda) ( P X , Λ ) is in the image under P \mathbb{P} P of
Cap if and only if (2.12);
PsAp if and only if (2.12) and it is fc-deep, if and only if (2.12) and it is F \mathbb{F} F -adherence determined;
PrAp if and only if (2.12) and it is deep, if and only if (2.12) and it is F 0 \mathbb{F}_0 F 0 -adherence determined.
Proof. (1). It is clear that the image of a Cap-object under P \mathbb{P} P satisfies (29) . Assume conversely that L = ( P X , Λ ) L=(\mathbb{P}X,\Lambda) L = ( P X , Λ ) satisfies (29) . Then P ( pt L , λ pt L ) = ( P X , Λ ) \mathbb{P}(\operatorname{pt}L,\lambda_{\operatorname{pt}L})=(\mathbb{P}X,\Lambda) P ( pt L , λ pt L ) = ( P X , Λ ) . Indeed, pt L = X \operatorname{pt}L=X pt L = X and F ∘ = F \mathcal{F}^{\circ}=\mathcal{F} F ∘ = F so that
Λ P pt L F ( v ) = { x ∈ X : λ pt L F ( x ) ≥ v } = { x ∈ X : ⋁ { t ∈ V : Λ F ∘ ( t ) ∈ ↑ x } ≥ v } = { x ∈ X : x ∈ Λ F ( v ) } = Λ F ( v ) . \Lambda_{\mathbb{P}\operatorname{pt}L}\mathcal{F}(v)
=
\{x\in X:\lambda_{\operatorname{pt}L}\mathcal{F}(x)\ge v\}
=
\left\{x\in X:\bigvee\{t\in V:\Lambda\mathcal{F}^{\circ}(t)\in\uparrow x\}\ge v\right\}
=
\{x\in X:x\in\Lambda\mathcal{F}(v)\}
=
\Lambda\mathcal{F}(v). Λ P pt L F ( v ) = { x ∈ X : λ pt L F ( x ) ≥ v } = { x ∈ X : ⋁ { t ∈ V : Λ F ∘ ( t ) ∈↑ x } ≥ v } = { x ∈ X : x ∈ Λ F ( v )} = Λ F ( v ) .
Regarding (2) and (3), note that the equivalences between fc-deep and F \mathbb{F} F -adherence determined, and between deep and F 0 \mathbb{F}_{0} F 0 -adherence determined, follow from Proposition 35.
(2). If ( X , λ ) (X,\lambda) ( X , λ ) is a pseudo-approach space then ( P X , Λ ) (\mathbb{P}X,\Lambda) ( P X , Λ ) is F \mathbb{F} F -adherence determined. Indeed,
Λ F ( v ) = { x ∈ X : λ F ( x ) = ⋀ H # F adh λ H ( x ) ≥ v } = ⋂ H # F { adh λ H ≥ v } = ⋀ H # F adh Λ H ( v ) , \Lambda\mathcal{F}(v)
=
\left\{x\in X:\lambda\mathcal{F}(x)=\bigwedge_{\mathcal{H}\#\mathcal{F}}\operatorname{adh}_{\lambda}\mathcal{H}(x)\ge v\right\}
=
\bigcap_{\mathcal{H}\#\mathcal{F}}\{\operatorname{adh}_{\lambda}\mathcal{H}\ge v\}
=
\bigwedge_{\mathcal{H}\#\mathcal{F}}\operatorname{adh}_{\Lambda}\mathcal{H}(v), Λ F ( v ) = ⎩ ⎨ ⎧ x ∈ X : λ F ( x ) = H # F ⋀ adh λ H ( x ) ≥ v ⎭ ⎬ ⎫ = H # F ⋂ { adh λ H ≥ v } = H # F ⋀ adh Λ H ( v ) ,
because adh Λ H ( v ) = { adh λ H ≥ v } \operatorname{adh}_{\Lambda}\mathcal{H}(v)=\{\operatorname{adh}_{\lambda}\mathcal{H}\ge v\} adh Λ H ( v ) = { adh λ H ≥ v } as noted in Section 5. On the other hand, if ( P X , Λ ) (\mathbb{P}X,\Lambda) ( P X , Λ ) is F \mathbb{F} F -adherence determined and (29) , then ( P X , Λ ) = P ( X , λ ) (\mathbb{P}X,\Lambda)=\mathbb{P}(X,\lambda) ( P X , Λ ) = P ( X , λ ) where λ F ( x ) = ⋁ { v ∈ V : x ∈ Λ F ( v ) } \lambda\mathcal{F}(x)=\bigvee\{v\in V:x\in\Lambda\mathcal{F}(v)\} λ F ( x ) = ⋁ { v ∈ V : x ∈ Λ F ( v )} . To see that λ \lambda λ is a pseudo-approach structure, first note that it satisfies (19) because of (29) . Moreover, adh λ H ( x ) = ⋁ G # H ⋁ x ∈ Λ G ( v ) v \operatorname{adh}_{\lambda}\mathcal{H}(x)=\bigvee_{\mathcal{G}\#\mathcal{H}}\bigvee_{x\in\Lambda\mathcal{G}(v)}v adh λ H ( x ) = ⋁ G # H ⋁ x ∈ Λ G ( v ) v , so that adh λ H ( t ) ≥ v \operatorname{adh}_{\lambda}\mathcal{H}(t)\ge v adh λ H ( t ) ≥ v if for every ϵ < 1 \epsilon<1 ϵ < 1 , there is G ϵ # H \mathcal{G}_{\epsilon}\#\mathcal{H} G ϵ # H and there is w ϵ ∈ V w_{\epsilon}\in V w ϵ ∈ V with t ∈ Λ G ϵ ( w ϵ ) t\in\Lambda\mathcal{G}_{\epsilon}(w_{\epsilon}) t ∈ Λ G ϵ ( w ϵ ) with w ϵ ≥ v ⊗ ϵ w_{\epsilon}\ge v\otimes\epsilon w ϵ ≥ v ⊗ ϵ , so that t ∈ Λ G ϵ ( v ⊗ ϵ ) t\in\Lambda\mathcal{G}_{\epsilon}(v\otimes\epsilon) t ∈ Λ G ϵ ( v ⊗ ϵ ) . In other words,
{ adh λ H ≥ v } = adh Λ H ( v ) . (47) \{\operatorname{adh}_{\lambda}\mathcal{H}\ge v\}=\operatorname{adh}_{\Lambda}\mathcal{H}(v). \tag*{(47)} { adh λ H ≥ v } = adh Λ H ( v ) . (47)
Hence,
λ F ( x ) = ⋁ { v ∈ V : x ∈ Λ F ( v ) = ⋀ H # F adh Λ H ( v ) } = ⋁ { v ∈ V : H # F ⇒ adh λ H ( x ) ≥ v } = ⋀ H # F adh λ H ( x ) . \lambda\mathcal{F}(x)
=
\bigvee\left\{v\in V:x\in\Lambda\mathcal{F}(v)=\bigwedge_{\mathcal{H}\#\mathcal{F}}\operatorname{adh}_{\Lambda}\mathcal{H}(v)\right\}
=
\bigvee\{v\in V:\mathcal{H}\#\mathcal{F}\Rightarrow\operatorname{adh}_{\lambda}\mathcal{H}(x)\ge v\}
=
\bigwedge_{\mathcal{H}\#\mathcal{F}}\operatorname{adh}_{\lambda}\mathcal{H}(x). λ F ( x ) = ⋁ ⎩ ⎨ ⎧ v ∈ V : x ∈ Λ F ( v ) = H # F ⋀ adh Λ H ( v ) ⎭ ⎬ ⎫ = ⋁ { v ∈ V : H # F ⇒ adh λ H ( x ) ≥ v } = H # F ⋀ adh λ H ( x ) .
(3). Similarly, if ( X , λ ) (X,\lambda) ( X , λ ) is a pre-approach space then ( P X , Λ ) (\mathbb{P}X,\Lambda) ( P X , Λ ) is deep because
Λ ( ⋂ D ∈ D D ) ( v ) = { x ∈ X : λ ( ⋂ D ∈ D D ) ( x ) = ⋀ D ∈ D λ D ( x ) ≥ v } = ⋂ D ∈ D { λ D ≥ v } = ⋀ D ∈ D Λ D ( v ) . \Lambda\left(\bigcap_{\mathcal{D}\in\mathbb{D}}\mathcal{D}\right)(v)
=
\left\{x\in X:\lambda\left(\bigcap_{\mathcal{D}\in\mathbb{D}}\mathcal{D}\right)(x)=\bigwedge_{\mathcal{D}\in\mathbb{D}}\lambda\mathcal{D}(x)\ge v\right\}
=
\bigcap_{\mathcal{D}\in\mathbb{D}}\{\lambda\mathcal{D}\ge v\}
=
\bigwedge_{\mathcal{D}\in\mathbb{D}}\Lambda\mathcal{D}(v). Λ ( D ∈ D ⋂ D ) ( v ) = { x ∈ X : λ ( D ∈ D ⋂ D ) ( x ) = D ∈ D ⋀ λ D ( x ) ≥ v } = D ∈ D ⋂ { λ D ≥ v } = D ∈ D ⋀ Λ D ( v ) .
On the other hand, if ( P X , Λ ) (\mathbb{P}X,\Lambda) ( P X , Λ ) is deep and (29) , then ( P X , Λ ) = P ( X , λ ) (\mathbb{P}X,\Lambda)=\mathbb{P}(X,\lambda) ( P X , Λ ) = P ( X , λ ) where λ F ( x ) = ⋁ { v ∈ V : x ∈ Λ F ( v ) } \lambda\mathcal{F}(x)=\bigvee\{v\in V:x\in\Lambda\mathcal{F}(v)\} λ F ( x ) = ⋁ { v ∈ V : x ∈ Λ F ( v )} satisfies (19) because of (29) and
λ ( ⋂ D ∈ D D ) ( x ) = ⋁ { v ∈ V : x ∈ Λ ( ⋂ D ∈ D D ) ( v ) = ⋂ D ∈ D Λ D ( v ) } = ⋀ D ∈ D λ D ( x ) . \lambda\left(\bigcap_{\mathcal{D}\in\mathbb{D}}\mathcal{D}\right)(x)
=
\bigvee\left\{v\in V:x\in\Lambda\left(\bigcap_{\mathcal{D}\in\mathbb{D}}\mathcal{D}\right)(v)=\bigcap_{\mathcal{D}\in\mathbb{D}}\Lambda\mathcal{D}(v)\right\}
=
\bigwedge_{\mathcal{D}\in\mathbb{D}}\lambda\mathcal{D}(x). λ ( D ∈ D ⋂ D ) ( x ) = ⋁ { v ∈ V : x ∈ Λ ( D ∈ D ⋂ D ) ( v ) = D ∈ D ⋂ Λ D ( v ) } = D ∈ D ⋀ λ D ( x ) .
On the other hand, recall that an approach space is a pre-approach space satisfying (5.4).
Proposition 38. A C C a p \mathsf{C}^{\mathrm{Cap}} C Cap -object ( P X , Λ ) (\mathbb{P}X,\Lambda) ( P X , Λ ) is in the image of A p \mathsf{Ap} Ap under P \mathbb{P} P if and only if it is in the image of P r A p \mathsf{PrAp} PrAp and
adh Λ ℓ ( v ⊗ ϵ ) ≥ adh Λ ( ( adh Λ ℓ ) ( ϵ ) ) ( v ) (48) \operatorname{adh}_{\Lambda}\ell(v\otimes\epsilon)\geq\operatorname{adh}_{\Lambda}\left((\operatorname{adh}_{\Lambda}\ell)(\epsilon)\right)(v) \tag*{(48)} adh Λ ℓ ( v ⊗ ϵ ) ≥ adh Λ ( ( adh Λ ℓ ) ( ϵ ) ) ( v ) (48) for every ℓ ∈ P X \ell\in\mathbb{P}X ℓ ∈ P X and v , ϵ ∈ V v,\epsilon\in V v , ϵ ∈ V .
Proof. If ( X , λ ) (X,\lambda) ( X , λ ) is an approach space, it is in particular a pre-approach space, hence ( P X , Λ P X ) (\mathbb{P}X,\Lambda_{\mathbb{P}X}) ( P X , Λ P X ) is deep and centered. Moreover, as adh λ A ( x ) ≥ adh λ A ( ϵ ) ( x ) ⊗ ϵ \operatorname{adh}_{\lambda}A(x)\geq\operatorname{adh}_{\lambda}A^{(\epsilon)}(x)\otimes\epsilon adh λ A ( x ) ≥ adh λ A ( ϵ ) ( x ) ⊗ ϵ for every A ∈ P X A\in\mathbb{P}X A ∈ P X and A ( ϵ ) = { adh A ≥ ϵ } = adh Λ A ( ϵ ) A^{(\epsilon)}=\{\operatorname{adh}A\geq\epsilon\}=\operatorname{adh}_{\Lambda}A(\epsilon) A ( ϵ ) = { adh A ≥ ϵ } = adh Λ A ( ϵ ) , we obtain that if x ∈ adh Λ ( ( adh Λ A ) ( ϵ ) ) ( v ) x\in\operatorname{adh}_{\Lambda}\left((\operatorname{adh}_{\Lambda}A)(\epsilon)\right)(v) x ∈ adh Λ ( ( adh Λ A ) ( ϵ ) ) ( v ) then adh λ ( A ( ϵ ) ) ≥ v \operatorname{adh}_{\lambda}(A^{(\epsilon)})\geq v adh λ ( A ( ϵ ) ) ≥ v thus
adh λ A ( x ) ≥ adh λ A ( ϵ ) ( x ) ⊗ ϵ ≥ v ⊗ ϵ , \operatorname{adh}_{\lambda}A(x)\geq\operatorname{adh}_{\lambda}A^{(\epsilon)}(x)\otimes\epsilon\geq v\otimes\epsilon, adh λ A ( x ) ≥ adh λ A ( ϵ ) ( x ) ⊗ ϵ ≥ v ⊗ ϵ , so that x ∈ adh Λ A ( v ⊗ ϵ ) x\in\operatorname{adh}_{\Lambda}A(v\otimes\epsilon) x ∈ adh Λ A ( v ⊗ ϵ ) and (6.5) is satisfied. Moreover, if ( P X , Λ ) (\mathbb{P}X,\Lambda) ( P X , Λ ) is centered and deep, we have Λ = Λ P X \Lambda=\Lambda_{\mathbb{P}X} Λ = Λ P X for the pre-approach structure λ \lambda λ given by λ F ( x ) = ⋁ { v ∈ V : x ∈ Λ F ( v ) } \lambda\mathcal{F}(x)=\bigvee\{v\in V:x\in\Lambda\mathcal{F}(v)\} λ F ( x ) = ⋁ { v ∈ V : x ∈ Λ F ( v )} , by Theorem 37. If it satisfies (6.5) as well, then in view of (6.4), (6.5) rephrases as
{ adh λ A ≥ v ⊗ ϵ } ⊃ { adh λ ( A ( ϵ ) ) ≥ v } , \{\operatorname{adh}_{\lambda}A\geq v\otimes\epsilon\}\supset\{\operatorname{adh}_{\lambda}(A^{(\epsilon)})\geq v\}, { adh λ A ≥ v ⊗ ϵ } ⊃ { adh λ ( A ( ϵ ) ) ≥ v } , yielding (5.4). □
Adherence in ( L , Λ ) (L,\Lambda) ( L , Λ ) vs adherence in pt L \operatorname{pt}L pt L
The correspondence (6.4) between adh λ \operatorname{adh}_{\lambda} adh λ and adh Λ P X \operatorname{adh}_{\Lambda_{\mathbb{P}X}} adh Λ P X extends partially to adh pt L \operatorname{adh}_{\operatorname{pt}L} adh pt L and adh Λ \operatorname{adh}_{\Lambda} adh Λ in C H C a p \mathsf{CH}^{\mathrm{Cap}} CH Cap :
Proposition 39. Let C \mathsf{C} C be such that points of C C a p \mathsf{C}^{\mathrm{Cap}} C Cap are principal. Let ( L , Λ ) (L,\Lambda) ( L , Λ ) be a C C a p \mathsf{C}^{\mathrm{Cap}} C Cap -object. Let H ∈ F P ( pt L ) \mathcal{H}\in\mathbb{F}\mathbb{P}(\operatorname{pt}L) H ∈ FP ( pt L ) , x ∈ pt L x\in\operatorname{pt}L x ∈ pt L , and v ∈ V v\in V v ∈ V . Then
adh pt L H ( x ) ≥ v ⟹ adh Λ H ∘ ( v ) ∈ x . \operatorname{adh}_{\operatorname{pt}L}\mathcal{H}(x)\geq v\Longrightarrow\operatorname{adh}_{\Lambda}\mathcal{H}^{\circ}(v)\in x. adh pt L H ( x ) ≥ v ⟹ adh Λ H ∘ ( v ) ∈ x . If moreover L L L is weakly spatial and points of C C a p \mathsf{C}^{\mathrm{Cap}} C Cap are principal filters of completely prime elements (e.g., for C = C H \mathsf{C}=\mathsf{CH} C = CH ), then
adh Λ H ∘ ( v ) ∈ x ⟹ adh pt L H ⧫ ( x ) ≥ v . \operatorname{adh}_{\Lambda}\mathcal{H}^{\circ}(v)\in x\Longrightarrow\operatorname{adh}_{\operatorname{pt}L}\mathcal{H}^{\blacklozenge}(x)\geq v. adh Λ H ∘ ( v ) ∈ x ⟹ adh pt L H ⧫ ( x ) ≥ v . Proof. Assume that
adh pt L H ( x ) = ⋁ U ∈ β ( H ) Λ U ∘ ( w ) ∈ x w ≥ v . \operatorname{adh}_{\operatorname{pt}L}\mathcal{H}(x)=\bigvee_{\substack{U\in\beta(\mathcal{H})\\\Lambda U^{\circ}(w)\in x}}w\geq v. adh pt L H ( x ) = U ∈ β ( H ) Λ U ∘ ( w ) ∈ x ⋁ w ≥ v . For every ϵ < 1 \epsilon<1 ϵ < 1 , there is U ∈ β ( H ) U\in\beta(\mathcal{H}) U ∈ β ( H ) with Λ U ∘ ( v ⊗ ϵ ) ∈ x \Lambda U^{\circ}(v\otimes\epsilon)\in x Λ U ∘ ( v ⊗ ϵ ) ∈ x . As U ∘ ⊃ H ∘ U^{\circ}\supset\mathcal{H}^{\circ} U ∘ ⊃ H ∘ , ⋀ L ⊃ H ∘ Λ L ( v ⊗ ϵ ) ∈ x \bigwedge_{L\supset\mathcal{H}^{\circ}}\Lambda L(v\otimes\epsilon)\in x ⋀ L ⊃ H ∘ Λ L ( v ⊗ ϵ ) ∈ x and x x x is principal, so that adh Λ H ∘ ( v ) = ⋀ ϵ < 1 ⋁ L ⊃ H ∘ Λ L ( v ⊗ ϵ ) ∈ x \operatorname{adh}_{\Lambda}\mathcal{H}^{\circ}(v)=\bigwedge_{\epsilon<1}\bigvee_{L\supset\mathcal{H}^{\circ}}\Lambda L(v\otimes\epsilon)\in x adh Λ H ∘ ( v ) = ⋀ ϵ < 1 ⋁ L ⊃ H ∘ Λ L ( v ⊗ ϵ ) ∈ x .
Suppose adh Λ H ∘ ( v ) ∈ x \operatorname{adh}_{\Lambda}\mathcal{H}^{\circ}(v)\in x adh Λ H ∘ ( v ) ∈ x and x x x is a principal completely prime filter. Since adh Λ H ∘ ( v ) = ⋀ ϵ < 1 ⋁ L ⊃ H ∘ Λ L ( v ⊗ ϵ ) \operatorname{adh}_{\Lambda}\mathcal{H}^{\circ}(v)=\bigwedge_{\epsilon<1}\bigvee_{L\supset\mathcal{H}^{\circ}}\Lambda L(v\otimes\epsilon) adh Λ H ∘ ( v ) = ⋀ ϵ < 1 ⋁ L ⊃ H ∘ Λ L ( v ⊗ ϵ ) , ⋁ L ⊃ H ∘ Λ L ( v ⊗ ϵ ) ≥ x \bigvee_{L\supset\mathcal{H}^{\circ}}\Lambda L(v\otimes\epsilon)\geq x ⋁ L ⊃ H ∘ Λ L ( v ⊗ ϵ ) ≥ x for every ϵ < 1 \epsilon<1 ϵ < 1 . Hence, by complete primeness of ⋁ x \mathord{\bigvee}x ⋁ x , there is L ϵ ⊃ H ∘ L_{\epsilon}\supset\mathcal{H}^{\circ} L ϵ ⊃ H ∘ with Λ L ϵ ( v ⊗ ϵ ) ∈ x \Lambda L_{\epsilon}(v\otimes\epsilon)\in x Λ L ϵ ( v ⊗ ϵ ) ∈ x . In view of Lemma 22, there is U ϵ ∈ β ( H ⧫ ) U_{\epsilon}\in\beta(\mathcal{H}^{\blacklozenge}) U ϵ ∈ β ( H ⧫ ) with U ϵ ∘ ⊃ L ϵ ∪ H ∘ U_{\epsilon}^{\circ}\supset L_{\epsilon}\cup\mathcal{H}^{\circ} U ϵ ∘ ⊃ L ϵ ∪ H ∘ and thus Λ U ϵ ∘ ( v ⊗ ϵ ) ∈ x \Lambda U_{\epsilon}^{\circ}(v\otimes\epsilon)\in x Λ U ϵ ∘ ( v ⊗ ϵ ) ∈ x . Hence
adh pt L H ⧫ ( x ) = ⋁ U ∈ β ( H ⧫ ) λ pt L L U ( x ) = ⋁ U ∈ β ( H ⧫ ) Λ U ∘ ( w ) ∈ x w ≥ ⋁ ϵ < 1 v ⊗ ϵ = v . \operatorname{adh}_{\operatorname{pt}L}\mathcal{H}^{\blacklozenge}(x)=\bigvee_{U\in\beta(\mathcal{H}^{\blacklozenge})}\lambda_{\operatorname{pt}L}LU(x)=\bigvee_{\substack{U\in\beta(\mathcal{H}^{\blacklozenge})\\\Lambda U^{\circ}(w)\in x}}w\geq\bigvee_{\epsilon<1}v\otimes\epsilon=v. adh pt L H ⧫ ( x ) = U ∈ β ( H ⧫ ) ⋁ λ pt L LU ( x ) = U ∈ β ( H ⧫ ) Λ U ∘ ( w ) ∈ x ⋁ w ≥ ϵ < 1 ⋁ v ⊗ ϵ = v .
Corollary 40. Let C C C be such that points of C C a p C^{\mathrm{Cap}} C Cap are principal. Let ( L , Λ ) (L,\Lambda) ( L , Λ ) be a C C a p C^{\mathrm{Cap}} C Cap -object. If A ⊂ p t L A \subset\mathrm{pt}L A ⊂ pt L then ( { A } ↑ ) ∘ \left(\{A\}^{\uparrow}\right)^{\circ} ( { A } ↑ ) ∘ is principal. Given ϵ ∈ V \epsilon\in V ϵ ∈ V , let a ϵ : = ⋀ ( { A ( ϵ ) } ↑ ) ∘ a_{\epsilon}:=\bigwedge\left(\{A^{(\epsilon)}\}^{\uparrow}\right)^{\circ} a ϵ := ⋀ ( { A ( ϵ ) } ↑ ) ∘ and a : = ⋀ ( { A } ↑ ) ∘ a:=\bigwedge\left(\{A\}^{\uparrow}\right)^{\circ} a := ⋀ ( { A } ↑ ) ∘ . Then
(1)
a ϵ ≤ adh Λ a ( ϵ ) ; (49) a_{\epsilon}\leq\operatorname{adh}_{\Lambda}a(\epsilon); \tag*{(49)} a ϵ ≤ adh Λ a ( ϵ ) ; (49) (2) If moreover, ( L , Λ ) (L,\Lambda) ( L , Λ ) satisfies (6.5) then
x ∈ ( adh Λ a ( adh p t L A ( ϵ ) ( x ) ⊗ ϵ ) ) ∙ , (50) x\in\left(\operatorname{adh}_{\Lambda}a\left(\operatorname{adh}_{\mathrm{pt}\,L}A^{(\epsilon)}(x)\otimes\epsilon\right)\right)^{\bullet}, \tag*{(50)} x ∈ ( adh Λ a ( adh pt L A ( ϵ ) ( x ) ⊗ ϵ ) ) ∙ , (50) for every x ∈ p t L x\in\mathrm{pt}L x ∈ pt L ;
(3) If moreover L L L is weakly spatial and points of C C a p C^{\mathrm{Cap}} C Cap are principal completely prime filters, then
adh p t L a ∙ ( x ) ≥ adh p t L A ( ϵ ) ( x ) ⊗ ϵ . (51) \operatorname{adh}_{\mathrm{pt}\,L}a^{\bullet}(x)\geq\operatorname{adh}_{\mathrm{pt}\,L}A^{(\epsilon)}(x)\otimes\epsilon. \tag*{(51)} adh pt L a ∙ ( x ) ≥ adh pt L A ( ϵ ) ( x ) ⊗ ϵ . (51) Proof. (1). Since ℓ ∈ ( { A } ↑ ) ∘ \ell\in\left(\{A\}^{\uparrow}\right)^{\circ} ℓ ∈ ( { A } ↑ ) ∘ if and only if A ⊂ ℓ ∙ A\subset\ell^{\bullet} A ⊂ ℓ ∙ , we have, in view of Lemma 15(4),
A ⊂ ⋀ ℓ ∈ ( { A } ↑ ) ∘ ℓ ∙ = ( ⋀ ℓ ∈ ( { A } ↑ ) ∘ ℓ ) ∙ A\subset\bigwedge_{\ell\in\left(\{A\}^{\uparrow}\right)^{\circ}}\ell^{\bullet}=\left(\bigwedge_{\ell\in\left(\{A\}^{\uparrow}\right)^{\circ}}\ell\right)^{\bullet} A ⊂ ℓ ∈ ( { A } ↑ ) ∘ ⋀ ℓ ∙ = ℓ ∈ ( { A } ↑ ) ∘ ⋀ ℓ ∙ so that a : = ⋀ ℓ ∈ ( { A } ↑ ) ∘ ℓ ∈ ( { A } ↑ ) ∘ a:=\bigwedge_{\ell\in\left(\{A\}^{\uparrow}\right)^{\circ}}\ell\in\left(\{A\}^{\uparrow}\right)^{\circ} a := ⋀ ℓ ∈ ( { A } ↑ ) ∘ ℓ ∈ ( { A } ↑ ) ∘ , and ( { A } ↑ ) ∘ \left(\{A\}^{\uparrow}\right)^{\circ} ( { A } ↑ ) ∘ is thus principal. Recall that A ( ϵ ) = { t ∈ p t L : adh p t L A ( t ) ≥ ϵ } A^{(\epsilon)}=\{t\in\mathrm{pt}\,L:\operatorname{adh}_{\mathrm{pt}\,L}A(t)\geq\epsilon\} A ( ϵ ) = { t ∈ pt L : adh pt L A ( t ) ≥ ϵ } . For each t ∈ A ( ϵ ) t\in A^{(\epsilon)} t ∈ A ( ϵ ) , Proposition 39 applies to the effect that
adh Λ a ( ϵ ) = adh Λ ( { A } ↑ ) ∘ ( ϵ ) ∈ t , \operatorname{adh}_{\Lambda}a(\epsilon)=\operatorname{adh}_{\Lambda}\left(\{A\}^{\uparrow}\right)^{\circ}(\epsilon)\in t, adh Λ a ( ϵ ) = adh Λ ( { A } ↑ ) ∘ ( ϵ ) ∈ t , that is, t ∈ ( adh Λ a ( ϵ ) ) ∙ t\in\left(\operatorname{adh}_{\Lambda}a(\epsilon)\right)^{\bullet} t ∈ ( adh Λ a ( ϵ ) ) ∙ . In other words, A ( ϵ ) ⊂ ( adh Λ a ( ϵ ) ) ∙ A^{(\epsilon)}\subset\left(\operatorname{adh}_{\Lambda}a(\epsilon)\right)^{\bullet} A ( ϵ ) ⊂ ( adh Λ a ( ϵ ) ) ∙ , that is, adh Λ a ( ϵ ) ∈ ( { A ( ϵ ) } ↑ ) ∘ = ↑ a ϵ \operatorname{adh}_{\Lambda}a(\epsilon)\in\left(\{A^{(\epsilon)}\}^{\uparrow}\right)^{\circ}=\uparrow a_{\epsilon} adh Λ a ( ϵ ) ∈ ( { A ( ϵ ) } ↑ ) ∘ =↑ a ϵ , which yields the conclusion.
(2). Let x ∈ p t L x\in\mathrm{pt}\,L x ∈ pt L and v = adh p t L A ( ϵ ) ( x ) v=\operatorname{adh}_{\mathrm{pt}\,L}A^{(\epsilon)}(x) v = adh pt L A ( ϵ ) ( x ) . By Proposition 39,
adh Λ a ϵ ( v ) = adh Λ ( { A ( ϵ ) } ↑ ) ∘ ( v ) ∈ x \operatorname{adh}_{\Lambda}a_{\epsilon}(v)=\operatorname{adh}_{\Lambda}\left(\{A^{(\epsilon)}\}^{\uparrow}\right)^{\circ}(v)\in x adh Λ a ϵ ( v ) = adh Λ ( { A ( ϵ ) } ↑ ) ∘ ( v ) ∈ x and, as adh Λ : L → L V \operatorname{adh}_{\Lambda}:L\to L^{V} adh Λ : L → L V is monotone, its application to (6.6) yields
adh Λ a ϵ ( v ) ≤ adh Λ ( adh Λ a ( ϵ ) ) ( v ) ≤ adh Λ a ( v ⊗ ϵ ) , \operatorname{adh}_{\Lambda}a_{\epsilon}(v)\leq\operatorname{adh}_{\Lambda}\left(\operatorname{adh}_{\Lambda}a(\epsilon)\right)(v)\leq\operatorname{adh}_{\Lambda}a(v\otimes\epsilon), adh Λ a ϵ ( v ) ≤ adh Λ ( adh Λ a ( ϵ ) ) ( v ) ≤ adh Λ a ( v ⊗ ϵ ) , where the second inequality follows from (6.5). Since the left-hand side is in the filter x x x , so is the right-hand side and the conclusion follows.
(3). (6.7) rephrases as adh Λ ( { A } ↑ ) ∘ ( v ) ∈ x \operatorname{adh}_{\Lambda}\left(\{A\}^{\uparrow}\right)^{\circ}(v)\in x adh Λ ( { A } ↑ ) ∘ ( v ) ∈ x for v = adh p t L A ( ϵ ) ( x ) ⊗ ϵ v=\operatorname{adh}_{\mathrm{pt}\,L}A^{(\epsilon)}(x)\otimes\epsilon v = adh pt L A ( ϵ ) ( x ) ⊗ ϵ . Under the new assumptions, the second part of Proposition 39 applies to H = { A } ↑ \mathcal{H}=\{A\}^{\uparrow} H = { A } ↑ to the effect that
adh p t L ( ( { A } ↑ ) ∘ ) ∙ ( x ) = adh p t L a ∙ ( x ) ≥ adh p t L A ( ϵ ) ( x ) ⊗ ϵ . \operatorname{adh}_{\mathrm{pt}\,L}\left(\left(\{A\}^{\uparrow}\right)^{\circ}\right)^{\bullet}(x)=\operatorname{adh}_{\mathrm{pt}\,L}a^{\bullet}(x)\geq\operatorname{adh}_{\mathrm{pt}\,L}A^{(\epsilon)}(x)\otimes\epsilon. adh pt L ( ( { A } ↑ ) ∘ ) ∙ ( x ) = adh pt L a ∙ ( x ) ≥ adh pt L A ( ϵ ) ( x ) ⊗ ϵ .
Depth of spaces of points
In view of Theorem 37, D \mathcal{D} D -deep C C a p C^{\mathrm{Cap}} C Cap -objects naturally generalize PrAp and PsAp objects. Moreover, we can consider D \mathcal{D} D -depth in C a p \mathrm{Cap} Cap as well. Indeed, given a class D \mathcal{D} D of collections of filters, a convergence approach space ( X , λ ) (X,\lambda) ( X , λ ) is called D \mathcal{D} D -deep if
λ ( ⋂ D ∈ D D ) = ⋀ D ∈ D λ D , \lambda\left(\bigcap_{\mathcal{D}\in\mathcal{D}}\mathcal{D}\right)=\bigwedge_{\mathcal{D}\in\mathcal{D}}\lambda\mathcal{D}, λ ( D ∈ D ⋂ D ) = D ∈ D ⋀ λ D ,
for every D ∈ P ( P X ) \mathbb{D} \in\mathcal{P}(\mathrm{PX}) D ∈ P ( PX ) . Pre-approach spaces are exactly D \mathbb{D} D -deep convergence approach spaces where D ( L ) = P ( F L ) \mathbb{D}(L) = \mathcal{P}(\mathrm{FL}) D ( L ) = P ( FL ) while pseudo-approach spaces are exactly D \mathbb{D} D -deep convergence approach spaces where D \mathbb{D} D is the class of finitely covering sets of filters 4 .
Let us say that D \mathbb{D} D is pt-compatible if { D ∘ : D ∈ D } ∈ D ( L ) \{\mathbb{D}^{\circ} : \mathbb{D} \in\mathbb{D}\} \in\mathbb{D}(L) { D ∘ : D ∈ D } ∈ D ( L ) whenever D ∈ D ( P ( p t L ) ) \mathbb{D} \in\mathcal{D}(\mathcal{P}(\mathrm{pt}\,L)) D ∈ D ( P ( pt L )) .
Lemma 41. The class D \mathbb{D} D given by D ( L ) = P ( F L ) \mathbb{D}(L) = \mathcal{P}(\mathrm{FL}) D ( L ) = P ( FL ) and the class of finitely covering sets of filters are both pt-compatible.
Proof. It is obvious for D ( L ) = P ( F L ) \mathbb{D}(L) = \mathcal{P}(\mathrm{FL}) D ( L ) = P ( FL ) . Let now D ⊂ F P ( p t L ) \mathbb{D} \subset\mathbb{F}\mathbb{P}(\mathrm{pt}\,L) D ⊂ FP ( pt L ) be finitely covering and let ( ℓ D ) D ∈ D ∈ ∏ D ∈ D D ∘ (\ell_{\mathbb{D}})_{\mathbb{D}\in\mathbb{D}} \in\prod_{\mathbb{D}\in\mathbb{D}}\mathbb{D}^{\circ} ( ℓ D ) D ∈ D ∈ ∏ D ∈ D D ∘ , that is, ℓ D ∈ D \ell_{\mathbb{D}} \in\mathbb{D} ℓ D ∈ D for every D ∈ D \mathbb{D} \in\mathbb{D} D ∈ D . Because D \mathbb{D} D is finitely covering, there is a finite subset D 0 \mathbb{D}_0 D 0 of D \mathbb{D} D with ⋁ D ∈ D 0 ( ℓ D ) ∈ ⋂ D ∈ D D \bigvee_{\mathbb{D}\in\mathbb{D}_0}(\ell_{\mathbb{D}}) \in\bigcap_{\mathbb{D}\in\mathbb{D}}\mathbb{D} ⋁ D ∈ D 0 ( ℓ D ) ∈ ⋂ D ∈ D D . In view of Lemma 15,
⋁ D ∈ D 0 ( ℓ D ) ∙ = ( ⋁ D ∈ D 0 ℓ D ) ∙ ∈ ( ⋂ D ∈ D D ) ∘ = ⋂ D ∈ D D ∘ . \bigvee_{\mathbb{D}\in\mathbb{D}_0}(\ell_{\mathbb{D}})^{\bullet} = \left(\bigvee_{\mathbb{D}\in\mathbb{D}_0}\ell_{\mathbb{D}}\right)^{\bullet} \in\left(\bigcap_{\mathbb{D}\in\mathbb{D}}\mathbb{D}\right)^{\circ} = \bigcap_{\mathbb{D}\in\mathbb{D}}\mathbb{D}^{\circ}. D ∈ D 0 ⋁ ( ℓ D ) ∙ = ( D ∈ D 0 ⋁ ℓ D ) ∙ ∈ ( D ∈ D ⋂ D ) ∘ = D ∈ D ⋂ D ∘ .
and { D ∘ : D ∈ D } \{\mathbb{D}^{\circ} : \mathbb{D} \in\mathbb{D}\} { D ∘ : D ∈ D } is finitely covering. □
Remarkably, when D \mathbb{D} D is pt-compatible and C \mathbb{C} C is fully admissible, D \mathbb{D} D -depth is preserved by the point-functor:
Theorem 42. Let C \mathbb{C} C be fully admissible and let D \mathbb{D} D be a pt-compatible class of sets of filters. The space of points of a D \mathbb{D} D -deep C C a p \mathbb{C}^{\mathrm{Cap}} C Cap -object is a D \mathbb{D} D -deep convergence approach space.
Proof. Let ( L , Λ ) (L,\Lambda) ( L , Λ ) be D \mathbb{D} D -deep. Let D ∈ D ( P p t L ) \mathbb{D} \in\mathcal{D}(\mathcal{P}\,\mathrm{pt}\,L) D ∈ D ( P pt L ) . As λ p t L ( ⋂ D ∈ D D ) ( x ) = ⋁ { v ∈ V : Λ ( ⋂ D ∈ D D ) ( v ) ∈ x } \lambda_{\mathrm{pt}\,L}(\bigcap_{\mathbb{D}\in\mathbb{D}}\mathbb{D})(x) = \bigvee\{v \in V : \Lambda(\bigcap_{\mathbb{D}\in\mathbb{D}}\mathbb{D})(v) \in x\} λ pt L ( ⋂ D ∈ D D ) ( x ) = ⋁ { v ∈ V : Λ ( ⋂ D ∈ D D ) ( v ) ∈ x } and ( ⋂ D ∈ D D ) ∘ = ⋂ D ∈ D D ∘ (\bigcap_{\mathbb{D}\in\mathbb{D}}\mathbb{D})^{\circ} = \bigcap_{\mathbb{D}\in\mathbb{D}}\mathbb{D}^{\circ} ( ⋂ D ∈ D D ) ∘ = ⋂ D ∈ D D ∘ , that Λ \Lambda Λ be D \mathbb{D} D -deep yields Λ ( ⋂ D ∈ D D ) ∘ ( v ) = ⋀ D ∈ D Λ D ∘ ( v ) \Lambda(\bigcap_{\mathbb{D}\in\mathbb{D}}\mathbb{D})^{\circ}(v) = \bigwedge_{\mathbb{D}\in\mathbb{D}}\Lambda\mathbb{D}^{\circ}(v) Λ ( ⋂ D ∈ D D ) ∘ ( v ) = ⋀ D ∈ D Λ D ∘ ( v ) . If C \mathbb{C} C is fully admissible, then points are principal filters and thus ⋀ D ∈ D Λ D ∘ ( v ) ∈ x \bigwedge_{\mathbb{D}\in\mathbb{D}}\Lambda\mathbb{D}^{\circ}(v) \in x ⋀ D ∈ D Λ D ∘ ( v ) ∈ x if and only if Λ D ∘ ( v ) ∈ x \Lambda\mathbb{D}^{\circ}(v) \in x Λ D ∘ ( v ) ∈ x for every D ∈ D \mathbb{D} \in\mathbb{D} D ∈ D . In other words,
{ v ∈ V : Λ ( ⋂ D ∈ D D ) ∘ ( v ) ∈ x } = ⋂ D ∈ D { v ∈ V : Λ D ∘ ( v ) ∈ x } . \{v \in V : \Lambda(\bigcap_{\mathbb{D}\in\mathbb{D}}\mathbb{D})^{\circ}(v) \in x\} = \bigcap_{\mathbb{D}\in\mathbb{D}}\{v \in V : \Lambda\mathbb{D}^{\circ}(v) \in x\}. { v ∈ V : Λ ( D ∈ D ⋂ D ) ∘ ( v ) ∈ x } = D ∈ D ⋂ { v ∈ V : Λ D ∘ ( v ) ∈ x } .
Moreover, for each D ∈ D \mathbb{D} \in\mathbb{D} D ∈ D , in view of (2.10),
Λ D ∘ ( ⋁ { v ∈ V : Λ D ∘ ( v ) ∈ x } ) = ⋀ { v ∈ V : Λ D ∘ ( v ) ∈ x } Λ D ∘ ( v ) \Lambda\mathbb{D}^{\circ}\left(\bigvee\{v \in V : \Lambda\mathbb{D}^{\circ}(v) \in x\}\right) = \bigwedge_{\{v\in V:\Lambda\mathbb{D}^{\circ}(v)\in x\}}\Lambda\mathbb{D}^{\circ}(v) Λ D ∘ ( ⋁ { v ∈ V : Λ D ∘ ( v ) ∈ x } ) = { v ∈ V : Λ D ∘ ( v ) ∈ x } ⋀ Λ D ∘ ( v )
belongs to x x x because x x x is principal, so that each set { v ∈ V : Λ D ∘ ( v ) ∈ x } \{v \in V : \Lambda\mathbb{D}^{\circ}(v) \in x\} { v ∈ V : Λ D ∘ ( v ) ∈ x } has a greatest element. Since these sets are down-closed, the infimum of these greatest elements is in ⋂ D ∈ D { v ∈ V : Λ D ∘ ( v ) ∈ x } \bigcap_{\mathbb{D}\in\mathbb{D}}\{v \in V : \Lambda\mathbb{D}^{\circ}(v) \in x\} ⋂ D ∈ D { v ∈ V : Λ D ∘ ( v ) ∈ x } , so that
⋀ D ∈ D ⋁ { v ∈ V : Λ D ∘ ( v ) ∈ x } ≤ ⋁ ⋂ D ∈ D { v ∈ V : Λ D ∘ ( v ) ∈ x } . \bigwedge_{\mathbb{D}\in\mathbb{D}}\bigvee\{v \in V : \Lambda\mathbb{D}^{\circ}(v) \in x\} \le\bigvee\bigcap_{\mathbb{D}\in\mathbb{D}}\{v \in V : \Lambda\mathbb{D}^{\circ}(v) \in x\}. D ∈ D ⋀ ⋁ { v ∈ V : Λ D ∘ ( v ) ∈ x } ≤ ⋁ D ∈ D ⋂ { v ∈ V : Λ D ∘ ( v ) ∈ x } .
λ ( F ) = ⋀ U ∈ β F λ U = ⋀ D ∈ D ⋀ U ∈ β D λ U = ⋀ D ∈ D λ D . \lambda(\mathcal{F}) = \bigwedge_{U\in\beta\mathcal{F}}\lambda U = \bigwedge_{\mathbb{D}\in\mathbb{D}}\bigwedge_{U\in\beta\mathbb{D}}\lambda U = \bigwedge_{\mathbb{D}\in\mathbb{D}}\lambda\mathbb{D}. λ ( F ) = U ∈ β F ⋀ λ U = D ∈ D ⋀ U ∈ β D ⋀ λ U = D ∈ D ⋀ λ D .
while the reverse inequality is trivial. Thus
λ p t L ( ⋂ D ∈ D D ) ( x ) = ⋁ D ∈ D ⋂ { v ∈ V : Λ D ∘ ( v ) ∈ x } = ⋀ D ∈ D ⋁ { v ∈ V : Λ D ∘ ( v ) ∈ x } = ⋀ D ∈ D λ p t L D ( x ) , \lambda_{\mathrm{pt}\,L}\left(\bigcap_{\mathcal{D}\in\mathbb{D}}\mathcal{D}\right)(x)
=
\bigvee_{\mathcal{D}\in\mathbb{D}}\bigcap\{v\in V:\Lambda\mathcal{D}^{\circ}(v)\in x\}
=
\bigwedge_{\mathcal{D}\in\mathbb{D}}\bigvee\{v\in V:\Lambda\mathcal{D}^{\circ}(v)\in x\}
=
\bigwedge_{\mathcal{D}\in\mathbb{D}}\lambda_{\mathrm{pt}\,L}\mathcal{D}(x), λ pt L ( D ∈ D ⋂ D ) ( x ) = D ∈ D ⋁ ⋂ { v ∈ V : Λ D ∘ ( v ) ∈ x } = D ∈ D ⋀ ⋁ { v ∈ V : Λ D ∘ ( v ) ∈ x } = D ∈ D ⋀ λ pt L D ( x ) ,
that is, ( p t L , λ p t L ) (\mathrm{pt}\,L,\lambda_{\mathrm{pt}\,L}) ( pt L , λ pt L ) is a D \mathbb{D} D -deep convergence approach space. □
Alternatives for pointfree versions of PsAp, PrAp, and Ap
Let us assume that C is a fully admissible category of lattices. In view of the results of Section 6.2, we have two natural paths to consider pointfree versions of PsAp, PrAp, and Ap: via depth, or via adherence determinedness. In view of Theorem 42 , we will privilege depth for analogues of PrAp and PsAp, and denote by C P r A p C^{\mathrm{PrAp}} C PrAp and C P s A p C^{\mathrm{PsAp}} C PsAp the full subcategories of C C a p C^{\mathrm{Cap}} C Cap whose objects are deep and fc-deep respectively. In view of Theorem 37 , Theorem 42 , and Lemma 41 :
Corollary 43. The adjunction C a p ⇄ P p t L ( C C a p ) o p \mathrm{Cap}\underset{\mathrm{pt}\,L}{\overset{\mathbb{P}}{\rightleftarrows}}(C^{\mathrm{Cap}})^{\mathrm{op}} Cap pt L ⇄ P ( C Cap ) op restricts to P r A p ⇄ P p t L ( C P r A p ) o p \mathrm{PrAp}\underset{\mathrm{pt}\,L}{\overset{\mathbb{P}}{\rightleftarrows}}(C^{\mathrm{PrAp}})^{\mathrm{op}} PrAp pt L ⇄ P ( C PrAp ) op and to P s A p ⇄ P p t L ( C P s A p ) o p \mathrm{PsAp}\underset{\mathrm{pt}\,L}{\overset{\mathbb{P}}{\rightleftarrows}}(C^{\mathrm{PsAp}})^{\mathrm{op}} PsAp pt L ⇄ P ( C PsAp ) op whenever C is fully admissible.
On the other hand, we will denote by C A p C^{\mathrm{Ap}} C Ap the full subcategory of A d h F 0 \mathrm{Adh}_{F_0} Adh F 0 whose objects satisfy (6.5) and C A p p C^{\mathrm{App}} C App the full subcategory of C P r A p C^{\mathrm{PrAp}} C PrAp whose objects satisfy (6.5). Here, as the defining property (6.5) is in terms of the operator adh Λ : L → L V \operatorname{adh}_{\Lambda}:L\to L^{V} adh Λ : L → L V , we privilege adherence-determinedness, that is, C A p C^{\mathrm{Ap}} C Ap . Indeed, C A p C^{\mathrm{Ap}} C Ap -objects are determined by their operator adh Λ : L → L V \operatorname{adh}_{\Lambda}:L\to L^{V} adh Λ : L → L V , while C A p p C^{\mathrm{App}} C App -objects do not need to be. Proposition 35 (2) yields that A d h F \mathrm{Adh}_{F} Adh F embeds into C P s A p C^{\mathrm{PsAp}} C PsAp for C a category of bounded distributive lattices, such as the categories of frames and of coframes.
Similarly, if C is a category of frames, then A d h F 0 \mathrm{Adh}_{F_0} Adh F 0 embeds into C P r A p C^{\mathrm{PrAp}} C PrAp and C A p C^{\mathrm{Ap}} C Ap into C A p p C^{\mathrm{App}} C App , by Proposition 35 (1). Hence the dotted arrows below are only valid when C is a category of frames, while the solid ones are valid in any fully admissible category C of distributive lattices. Note that asking a category of frames to be fully admissible means that C-morphisms preserve arbitrary joins and meets, so that C is a subcategory of CH. As a result,
Corollary 44. The adjunction C a p ⇄ P p t L ( C H C a p ) o p \mathrm{Cap}\underset{\mathrm{pt}\,L}{\overset{\mathbb{P}}{\rightleftarrows}}(\mathrm{CH}^{\mathrm{Cap}})^{\mathrm{op}} Cap pt L ⇄ P ( CH Cap ) op restricts to P r A p ⇄ P p t L ( A d h F 0 ) o p \mathrm{PrAp}\underset{\mathrm{pt}\,L}{\overset{\mathbb{P}}{\rightleftarrows}}(\mathrm{Adh}_{F_0})^{\mathrm{op}} PrAp pt L ⇄ P ( Adh F 0 ) op and C a p ⇄ P p t L ( C C a p ) o p \mathrm{Cap}\underset{\mathrm{pt}\,L}{\overset{\mathbb{P}}{\rightleftarrows}}(C^{\mathrm{Cap}})^{\mathrm{op}} Cap pt L ⇄ P ( C Cap ) op restricts to P s A p ⇄ P p t L ( A d h F ) o p \mathrm{PsAp}\underset{\mathrm{pt}\,L}{\overset{\mathbb{P}}{\rightleftarrows}}(\mathrm{Adh}_{F})^{\mathrm{op}} PsAp pt L ⇄ P ( Adh F ) op whenever C is a fully
admissible category of distributive lattices.
C C a p ↑ A d h F 0 → A d h F → C P s A p → C C a p ↑ ↑ ↑ C A p → C A p p → C P r A p → C P s A p → C a p \begin{CD}
\mathrm{C}^{\mathrm{Cap}} \\
@A{}AA @. @. \\
\mathrm{Adh}_{\mathbb{F}_{0}} @>>> \mathrm{Adh}_{\mathbb{F}} @>>> \mathrm{C}^{\mathrm{PsAp}} @>>> \mathrm{C}^{\mathrm{Cap}} \\
@A{}AA @. @AA{}A @AA{}A \\
\mathrm{C}^{\mathrm{Ap}} @>>> \mathrm{C}^{\mathrm{App}} @>>> \mathrm{C}^{\mathrm{PrAp}} @>>> \mathrm{C}^{\mathrm{PsAp}} @>>> \mathrm{Cap}
\end{CD} C Cap ⏐ ↑ Adh F 0 ⏐ ↑ C Ap Adh F C App C PsAp ⏐ ↑ C PrAp C Cap ⏐ ↑ C PsAp Cap
Adherence-determinedness of spaces of points
Let WS denote the full subcategory of C C a p \mathrm{C}^{\mathrm{Cap}} C Cap whose objects are weakly spatial, i.e., satisfy (3.4).
We will say that a class D \mathbb{D} D of filters (determined by its instantiations D ( L ) ⊂ F ( L ) \mathbb{D}(L) \subset\mathbb{F}(L) D ( L ) ⊂ F ( L ) ) is pt-compatible if H ∙ ∈ D ( P p t L ) \mathcal{H}^{\bullet} \in\mathbb{D}(\mathbb{P}\,\mathrm{pt}\,L) H ∙ ∈ D ( P pt L ) whenever H ∈ D ( L ) \mathcal{H} \in\mathbb{D}(L) H ∈ D ( L ) . It is clear that:
Lemma 45. The classes F \mathbb{F} F of all filters and F 0 \mathbb{F}_{0} F 0 of principal filters are both pt-compatible.
Theorem 46. If C \mathrm{C} C is fully admissible and D \mathbb{D} D is pt-compatible, then the space of points of a weakly spatial A d h D \mathrm{Adh}_{\mathbb{D}} Adh D -object is a D \mathbb{D} D -adherence determined convergence approach space. In particular, the adjunction C a p ⇄ P p t L ( C C a p ) o p \mathrm{Cap} \underset{\mathrm{pt}\,L}{\stackrel{\mathbb{P}}{\rightleftarrows}} (\mathrm{C}^{\mathrm{Cap}})^{\mathrm{op}} Cap pt L ⇄ P ( C Cap ) op restricts to
P r A p ⇄ P p t L ( A d h F 0 ∩ W S ) o p . \mathrm{PrAp} \underset{\mathrm{pt}\,L}{\stackrel{\mathbb{P}}{\rightleftarrows}} (\mathrm{Adh}_{\mathbb{F}_{0}} \cap\mathrm{WS})^{\mathrm{op}}. PrAp pt L ⇄ P ( Adh F 0 ∩ WS ) op .
Proof. We want to show that if ( L , Λ ) (L,\Lambda) ( L , Λ ) is D \mathbb{D} D -adherence determined then
⋀ D ( P p t L ) ∋ D # F adh p t L D ( x ) ≤ λ p t L F ( x ) = ⋁ { v ∈ V : Λ F ∘ ( v ) ∈ x } . \bigwedge_{\substack{\mathbb{D}(\mathbb{P}\,\mathrm{pt}\,L)\ni\mathcal{D}\#\mathcal{F}}}\operatorname{adh}_{\mathrm{pt}\,L}\mathcal{D}(x) \leq\lambda_{\mathrm{pt}\,L}\mathcal{F}(x) = \bigvee\{v\in V:\Lambda\mathcal{F}^{\circ}(v)\in x\}. D ( P pt L ) ∋ D # F ⋀ adh pt L D ( x ) ≤ λ pt L F ( x ) = ⋁ { v ∈ V : Λ F ∘ ( v ) ∈ x } .
Because of (3.5), it is enough to show it under the assumption that F = F ♮ ∙ \mathcal{F}=\mathcal{F}^{\natural\bullet} F = F ♮ ∙ .
Let v ≤ ⋀ D ∋ D # F adh p t L D ( x ) v\leq\bigwedge_{\mathbb{D}\ni\mathcal{D}\#\mathcal{F}}\operatorname{adh}_{\mathrm{pt}\,L}\mathcal{D}(x) v ≤ ⋀ D ∋ D # F adh pt L D ( x ) . Hence, v ≤ adh p t L D ( x ) v\leq\operatorname{adh}_{\mathrm{pt}\,L}\mathcal{D}(x) v ≤ adh pt L D ( x ) for every D ∈ D ( P p t L ) \mathcal{D}\in\mathbb{D}(\mathbb{P}\,\mathrm{pt}\,L) D ∈ D ( P pt L ) with D # F \mathcal{D}\#\mathcal{F} D # F , so that ⋀ D ∋ D # F D ∘ ( v ) ∈ x \bigwedge_{\mathbb{D}\ni\mathcal{D}\#\mathcal{F}}\mathcal{D}^{\circ}(v)\in x ⋀ D ∋ D # F D ∘ ( v ) ∈ x by Proposition 39, and ⋁ D ∋ D # F adh Λ D ∘ ( v ) ∈ x \bigvee_{\mathbb{D}\ni\mathcal{D}\#\mathcal{F}}\operatorname{adh}_{\Lambda}\mathcal{D}^{\circ}(v)\in x ⋁ D ∋ D # F adh Λ D ∘ ( v ) ∈ x because x x x is principal. Let H ∈ D ( L ) \mathcal{H}\in\mathbb{D}(L) H ∈ D ( L ) with H # F ∘ \mathcal{H}\#\mathcal{F}^{\circ} H # F ∘ . By Lemma 22, H ∙ # F \mathcal{H}^{\bullet}\#\mathcal{F} H ∙ # F because F = F ♮ ∙ \mathcal{F}=\mathcal{F}^{\natural\bullet} F = F ♮ ∙ and L L L is weakly spatial. By Lemma 21, ( H ∙ ) ∘ # F ∘ (\mathcal{H}^{\bullet})^{\circ}\#\mathcal{F}^{\circ} ( H ∙ ) ∘ # F ∘ and H ⊂ ( H ∙ ) ∘ \mathcal{H}\subset(\mathcal{H}^{\bullet})^{\circ} H ⊂ ( H ∙ ) ∘ so that adh Λ ( H ∙ ) ∘ ≤ adh Λ H \operatorname{adh}_{\Lambda}(\mathcal{H}^{\bullet})^{\circ}\leq\operatorname{adh}_{\Lambda}\mathcal{H} adh Λ ( H ∙ ) ∘ ≤ adh Λ H . Moreover, D H = H ∙ ∈ D ( P p t L ) \mathcal{D}_{\mathcal{H}}=\mathcal{H}^{\bullet}\in\mathbb{D}(\mathbb{P}\,\mathrm{pt}\,L) D H = H ∙ ∈ D ( P pt L ) because D \mathbb{D} D is pt-compatible, so that
⋀ D ( P p t L ) ∋ D # F adh Λ D ∘ ( v ) ≤ ⋀ D ( L ) ∋ H # F ∘ adh Λ H ( v ) = Λ F ∘ ( v ) . \bigwedge_{\mathbb{D}(\mathbb{P}\,\mathrm{pt}\,L)\ni\mathcal{D}\#\mathcal{F}}\operatorname{adh}_{\Lambda}\mathcal{D}^{\circ}(v)
\leq
\bigwedge_{\mathbb{D}(L)\ni\mathcal{H}\#\mathcal{F}^{\circ}}\operatorname{adh}_{\Lambda}\mathcal{H}(v)
=
\Lambda\mathcal{F}^{\circ}(v). D ( P pt L ) ∋ D # F ⋀ adh Λ D ∘ ( v ) ≤ D ( L ) ∋ H # F ∘ ⋀ adh Λ H ( v ) = Λ F ∘ ( v ) .
Hence x ∈ Λ F ∘ ( v ) x\in\Lambda\mathcal{F}^{\circ}(v) x ∈ Λ F ∘ ( v ) , so that v ≤ λ p t L F ( x ) v\leq\lambda_{\mathrm{pt}\,L}\mathcal{F}(x) v ≤ λ pt L F ( x ) and the conclusion follows.
The particulars follow from Lemma 45. □
Note that in view of Proposition 35 and Corollary 43, weak spatiality is actually superfluous in Theorem 46 for D = F \mathbb{D}=\mathbb{F} D = F when C \mathrm{C} C -objects are distributive lattices, and for D = F 0 \mathbb{D}=\mathbb{F}_{0} D = F 0 when C \mathrm{C} C -objects are frames, like when C = C H \mathrm{C}=\mathrm{CH} C = CH . However, for D = F 0 \mathbb{D}=\mathbb{F}_{0} D = F 0 in the category C \mathrm{C} C of coframes, weak spatiality is essential:
Example 47 (A (non weakly spatial) C A p \mathrm{C}^{\mathrm{Ap}} C Ap -object whose space of points fails to be a pre-approach space). Let C C C denote the category of coframes, which is fully admissible. Let L = C R L=C_{\mathbb{R}} L = C R be the coframe of closed subsets of R \mathbb{R} R . Let c : C R → C R c:C_{\mathbb{R}}\to C_{\mathbb{R}} c : C R → C R be given by c ( F ) = cl R ( F ∖ { 0 } ) ∪ { 1 } c(F)=\operatorname{cl}_{\mathbb{R}}(F\setminus\{0\})\cup\{1\} c ( F ) = cl R ( F ∖ { 0 }) ∪ { 1 } if F ⊄ { 0 } F\not\subset\{0\} F ⊂ { 0 } and c ( F ) = ∅ c(F)=\varnothing c ( F ) = ∅ if F ⊂ { 0 } F\subset\{0\} F ⊂ { 0 } . Let
Λ H ( v ) = ⋂ F ∈ H # c ( F ) \Lambda\mathcal{H}(v)=\bigcap_{F\in\mathcal{H}^{\#}}c(F) Λ H ( v ) = F ∈ H # ⋂ c ( F ) for v > 0 v>0 v > 0 and Λ H ( 0 ) = R \Lambda\mathcal{H}(0)=\mathbb{R} Λ H ( 0 ) = R . Then, for every v > 0 v>0 v > 0 , Λ ( ↑ L { 0 } ) ( v ) ⊂ c ( { 0 } ) = ∅ \Lambda(\mathord{\uparrow}_{L}\{0\})(v)\subset c(\{0\})=\varnothing Λ ( ↑ L { 0 }) ( v ) ⊂ c ({ 0 }) = ∅ , while Λ ( ↑ L { x } ) ( v ) = ⋂ F ∋ x c ( F ) = { x , 1 } \Lambda(\mathord{\uparrow}_{L}\{x\})(v)=\bigcap_{F\ni x}c(F)=\{x,1\} Λ ( ↑ L { x }) ( v ) = ⋂ F ∋ x c ( F ) = { x , 1 } if x ≠ 0 x\ne0 x = 0 . Thus pt L = R ∖ { 0 } \operatorname{pt}L=\mathbb{R}\setminus\{0\} pt L = R ∖ { 0 } and { 0 } ∙ = ∅ \{0\}^{\bullet}=\varnothing { 0 } ∙ = ∅ so that L L L is not weakly spatial. For every F 0 ∈ L = C R F_{0}\in L=C_{\mathbb{R}} F 0 ∈ L = C R and v > 0 v>0 v > 0 ,
adh Λ F 0 ( v ) = ⋁ F 0 ∈ H # Λ H ( v ) ≤ c ( F 0 ) . \operatorname{adh}_{\Lambda}F_{0}(v)=\bigvee_{F_{0}\in\mathcal{H}^{\#}}\Lambda\mathcal{H}(v)\le c(F_{0}). adh Λ F 0 ( v ) = F 0 ∈ H # ⋁ Λ H ( v ) ≤ c ( F 0 ) . If F 0 ⊂ { 0 } F_{0}\subset\{0\} F 0 ⊂ { 0 } then c ( F 0 ) = adh Λ F 0 ( v ) = ∅ c(F_{0})=\operatorname{adh}_{\Lambda}F_{0}(v)=\varnothing c ( F 0 ) = adh Λ F 0 ( v ) = ∅ . On the other hand, if F 0 ⊄ { 0 } F_{0}\not\subset\{0\} F 0 ⊂ { 0 } , there is x ≠ 0 x\ne0 x = 0 with x ∈ F 0 x\in F_{0} x ∈ F 0 and thus
adh Λ F 0 ( v ) ≥ ⋁ x ∈ F 0 ∖ { 0 } Λ ( ↑ L { x } ) ( v ) = ⋁ x ∈ F 0 ∖ { 0 } { x , 1 } = cl R ( F 0 ∖ { 0 } ) ∪ { 1 } = c ( F 0 ) \operatorname{adh}_{\Lambda}F_{0}(v)\ge\bigvee_{x\in F_{0}\setminus\{0\}}\Lambda(\mathord{\uparrow}_{L}\{x\})(v)=\bigvee_{x\in F_{0}\setminus\{0\}}\{x,1\}=\operatorname{cl}_{\mathbb{R}}(F_{0}\setminus\{0\})\cup\{1\}=c(F_{0}) adh Λ F 0 ( v ) ≥ x ∈ F 0 ∖ { 0 } ⋁ Λ ( ↑ L { x }) ( v ) = x ∈ F 0 ∖ { 0 } ⋁ { x , 1 } = cl R ( F 0 ∖ { 0 }) ∪ { 1 } = c ( F 0 ) and thus adh Λ F 0 ( v ) = c ( F 0 ) \operatorname{adh}_{\Lambda}F_{0}(v)=c(F_{0}) adh Λ F 0 ( v ) = c ( F 0 ) . Thus Λ \Lambda Λ is F 0 F_{0} F 0 -adherence determined and satisfies (48) because c c c is idempotent, so that, ( L , Λ ) (L,\Lambda) ( L , Λ ) is a C A p \mathrm{C}^{\mathrm{Ap}} C Ap -object. On pt L = R ∖ { 0 } \operatorname{pt}L=\mathbb{R}\setminus\{0\} pt L = R ∖ { 0 } , consider the filter F = { ( 0 , 1 ) } ↑ = ⋂ x ∈ ( 0 , 1 ) { x } ↑ \mathcal{F}=\{(0,1)\}^{\uparrow}=\bigcap_{x\in(0,1)}\{x\}^{\uparrow} F = {( 0 , 1 ) } ↑ = ⋂ x ∈ ( 0 , 1 ) { x } ↑ . For every x ∈ ( 0 , 1 ) x\in(0,1) x ∈ ( 0 , 1 ) , ( { x } ↑ ) ∘ = ↑ L { x } (\{x\}^{\uparrow})^{\circ}=\mathord{\uparrow}_{L}\{x\} ({ x } ↑ ) ∘ = ↑ L { x } , hence 1 ∈ Λ ( { x } ↑ ) ( 1 ) 1\in\Lambda(\{x\}^{\uparrow})(1) 1 ∈ Λ ({ x } ↑ ) ( 1 ) and thus λ pt L ( { x } ↑ ) ( 1 ) = 1 \lambda_{\operatorname{pt}L}(\{x\}^{\uparrow})(1)=1 λ pt L ({ x } ↑ ) ( 1 ) = 1 , so that adh pt L B ( 1 ) = 1 \operatorname{adh}_{\operatorname{pt}L}B(1)=1 adh pt L B ( 1 ) = 1 whenever B ∈ F # B\in\mathcal{F}^{\#} B ∈ F # . Thus, ⋁ B ∈ F # adh pt L B ( 1 ) = 1 \bigvee_{B\in\mathcal{F}^{\#}}\operatorname{adh}_{\operatorname{pt}L}B(1)=1 ⋁ B ∈ F # adh pt L B ( 1 ) = 1 , while λ pt L F ( 1 ) = 0 \lambda_{\operatorname{pt}L}\mathcal{F}(1)=0 λ pt L F ( 1 ) = 0 because F ∘ = ↑ L [ 0 , 1 ] \mathcal{F}^{\circ}=\mathord{\uparrow}_{L}[0,1] F ∘ = ↑ L [ 0 , 1 ] and { 0 } ∈ ( F ∘ ) # \{0\}\in(\mathcal{F}^{\circ})^{\#} { 0 } ∈ ( F ∘ ) # so that Λ F ∘ ( 1 ) ⊂ c ( { 0 } ) = ∅ \Lambda\mathcal{F}^{\circ}(1)\subset c(\{0\})=\varnothing Λ F ∘ ( 1 ) ⊂ c ({ 0 }) = ∅ . As a result, pt L \operatorname{pt}L pt L fails to be a pre-approach space.
(6.5) and spaces of points
Recalling § 1.4 and Remark 17, we see that the structure of pt L \operatorname{pt}L pt L becomes more rigid in C H C a p \mathrm{CH}^{\mathrm{Cap}} CH Cap :
Lemma 48. Let L L L be a C H C a p \mathrm{CH}^{\mathrm{Cap}} CH Cap -object and let ℓ ∈ L \ell\in L ℓ ∈ L , x ∈ pt L x\in\operatorname{pt}L x ∈ pt L , A ⊂ pt L A\subset\operatorname{pt}L A ⊂ pt L and a : = ⋀ ( { A } ↑ ) ∘ a:=\bigwedge(\{A\}^{\uparrow})^{\circ} a := ⋀ ({ A } ↑ ) ∘ . Then
(1)
pt L ∖ ℓ ∙ = ( ℓ ∗ ) ∙ ; \operatorname{pt}L\setminus\ell^{\bullet}=(\ell^{*})^{\bullet}; pt L ∖ ℓ ∙ = ( ℓ ∗ ) ∙ ; (2)
( ⋀ x ) ∙ = { x } ; (\bigwedge x)^{\bullet}=\{x\}; ( ⋀ x ) ∙ = { x } ; (3) a = ⋁ x ∈ A ⋀ x a=\bigvee_{x\in A}\bigwedge x a = ⋁ x ∈ A ⋀ x ;
(4) a ∙ = A a^{\bullet}=A a ∙ = A .
Proof. (1). Because in C H C a p \mathrm{CH}^{\mathrm{Cap}} CH Cap points of an object L L L are CH-morphism φ : L → 2 \varphi:L\to2 φ : L → 2 , we have in particular φ ( ℓ ∗ ) = ( φ ( ℓ ) ) ∗ \varphi(\ell^{*})=(\varphi(\ell))^{*} φ ( ℓ ∗ ) = ( φ ( ℓ ) ) ∗ , that is,
ℓ ∈ x ⟺ ℓ ∗ ∉ x , (52) \ell\in x\Longleftrightarrow\ell^{*}\notin x, \tag*{(52)} ℓ ∈ x ⟺ ℓ ∗ ∈ / x , (52)
when the point is seen as a filter x x x on L L L , and thus
pt L ∖ ℓ ∙ = ( ℓ ∗ ) ∙ . \operatorname{pt}L\setminus\ell^{\bullet}=(\ell^{*})^{\bullet}. pt L ∖ ℓ ∙ = ( ℓ ∗ ) ∙ .
(2). Suppose y ∈ ( ⋀ x ) ∙ y\in(\bigwedge x)^{\bullet} y ∈ ( ⋀ x ) ∙ . Then ⋀ y ≤ ⋀ x \bigwedge y\le\bigwedge x ⋀ y ≤ ⋀ x . Moreover, ( ⋀ y ) ∉ x (\bigwedge y)\notin x ( ⋀ y ) ∈ / x for otherwise ⋀ y ≤ ⋀ x ≤ ( ⋀ y ) ∗ \bigwedge y\le\bigwedge x\le(\bigwedge y)^{*} ⋀ y ≤ ⋀ x ≤ ( ⋀ y ) ∗ , so that ⋀ y ∈ x \bigwedge y\in x ⋀ y ∈ x by (6.9). Thus ⋀ y = ⋀ x \bigwedge y=\bigwedge x ⋀ y = ⋀ x and x = y x=y x = y . (3). For every x ∈ A x \in A x ∈ A and ℓ ∈ ( { A } ↑ ) ∘ \ell\in(\{A\}^{\uparrow})^{\circ} ℓ ∈ ({ A } ↑ ) ∘ , ℓ ∈ x \ell\in x ℓ ∈ x and x x x is principal so that a ∈ x a \in x a ∈ x , that is, ⋀ x ∈ A Λ x ≤ a \bigwedge_{x \in A} \Lambda^{x} \leq a ⋀ x ∈ A Λ x ≤ a and thus ⋁ x ∈ A Λ x ≤ a \bigvee_{x \in A} \Lambda^{x} \leq a ⋁ x ∈ A Λ x ≤ a . The reverse inequality follows from ⋁ x ∈ A Λ x ∈ ( { A } ↑ ) ∘ \bigvee_{x \in A} \Lambda^{x} \in(\{A\}^{\uparrow})^{\circ} ⋁ x ∈ A Λ x ∈ ({ A } ↑ ) ∘ , which in turns follows from
( ⋁ x ∈ A Λ x ) ∙ = ⋃ x ∈ A ( Λ x ) ∙ = A \left(\bigvee_{x \in A} \Lambda^{x}\right)^{\bullet} = \bigcup_{x \in A}(\Lambda^{x})^{\bullet} = A ( x ∈ A ⋁ Λ x ) ∙ = x ∈ A ⋃ ( Λ x ) ∙ = A
by Lemma 15 and (2) above.
(4).
a ∙ = ( ⋁ x ∈ A Λ x ) ∙ = ⋃ x ∈ A ( Λ x ) ∙ = A . a^{\bullet} = \left(\bigvee_{x \in A} \Lambda^{x}\right)^{\bullet} = \bigcup_{x \in A}(\Lambda^{x})^{\bullet} = A. a ∙ = ( x ∈ A ⋁ Λ x ) ∙ = x ∈ A ⋃ ( Λ x ) ∙ = A .
Recall that by Proposition 35(1), C H A p \mathrm{CH}^{\mathrm{Ap}} CH Ap is a subcategory of C H A p p \mathrm{CH}^{\mathrm{App}} CH App .
Theorem 49. Let L L L be a weakly spatial C H A p p \mathrm{CH}^{\mathrm{App}} CH App -object. Then pt L \operatorname{pt} L pt L is an approach space. In other words, the adjunction
Cap ⇄ pt L ( C H C a p ) o p \operatorname{Cap} \underset{\operatorname{pt} L}{\rightleftarrows} (\mathrm{CH}^{\mathrm{Cap}})^{\mathrm{op}} Cap pt L ⇄ ( CH Cap ) op restricts to
A p ⇄ pt L ( C H A p p ∩ W S ) o p . \mathrm{Ap} \underset{\operatorname{pt} L}{\rightleftarrows} (\mathrm{CH}^{\mathrm{App}} \cap\mathrm{WS})^{\mathrm{op}}. Ap pt L ⇄ ( CH App ∩ WS ) op .
Proof. By Corollary 43, pt L \operatorname{pt} L pt L is a pre-approach space. We now want to show that pt L \operatorname{pt} L pt L satisfies (5.4) if L L L satisfies (6.5). Under these assumptions, given A ⊂ pt L A \subset\operatorname{pt} L A ⊂ pt L and ϵ ∈ V \epsilon\in V ϵ ∈ V , Corollary 40 yields (6.8), which combined with Lemma 48 (4), yields (5.4).
We cannot relax the assumption on C \mathcal{C} C to fully admissible.
Example 50 (A weakly spatial C A p p \mathcal{C}^{\mathrm{App}} C App -object, which is not a C A p \mathcal{C}^{\mathrm{Ap}} C Ap -object, with a space of points failing to be approach, for C \mathcal{C} C the category of coframes). Like in Example 47, let C \mathcal{C} C denote the category of coframes and let L = C R L = \mathcal{C}_{\mathbb{R}} L = C R be the coframe of closed subsets of R \mathbb{R} R . Let N L ( x ) = ↑ L { x } \mathcal{N}_{L}(x) = \mathord{\uparrow}_{L}\{x\} N L ( x ) = ↑ L { x } if x ∈ R ∖ { 0 , 1 } x \in\mathbb{R} \setminus\{0,1\} x ∈ R ∖ { 0 , 1 } , N L ( 0 ) = { F ∈ C R : ∃ δ > 0 , [ 0 , δ ] ⊂ F } \mathcal{N}_{L}(0) = \{F \in\mathcal{C}_{\mathbb{R}} : \exists\delta> 0,\ [0,\delta] \subset F\} N L ( 0 ) = { F ∈ C R : ∃ δ > 0 , [ 0 , δ ] ⊂ F } , and N L ( 1 ) = ↑ L { 0 , 1 } \mathcal{N}_{L}(1) = \mathord{\uparrow}_{L}\{0,1\} N L ( 1 ) = ↑ L { 0 , 1 } . Now, let Λ F ( v ) = { x ∈ R : N L ( x ) ⊂ F } \Lambda F(v) = \{x \in\mathbb{R} : \mathcal{N}_{L}(x) \subset F\} Λ F ( v ) = { x ∈ R : N L ( x ) ⊂ F } if v > 0 v > 0 v > 0 and Λ F ( 0 ) = R \Lambda F(0) = \mathbb{R} Λ F ( 0 ) = R . Since F F F contains at most one filter ↑ L { x } \mathord{\uparrow}_{L}\{x\} ↑ L { x } , Λ F ( v ) \Lambda F(v) Λ F ( v ) has at most 3 elements, hence is finite and thus belongs to L = C R L = \mathcal{C}_{\mathbb{R}} L = C R . Note that pt L = R \operatorname{pt} L = \mathbb{R} pt L = R and F ∙ = F F^{\bullet} = F F ∙ = F for every F ∈ L F \in L F ∈ L , so that ( L , Λ ) (L,\Lambda) ( L , Λ ) is weakly spatial. Moreover, ( L , Λ ) (L,\Lambda) ( L , Λ ) is deep. Indeed, for any D ⊂ F L \mathcal{D} \subset\mathcal{F}L D ⊂ F L , N L ( x ) ⊂ ⋂ D ∈ D D \mathcal{N}_{L}(x) \subset\bigcap_{D \in\mathcal{D}} D N L ( x ) ⊂ ⋂ D ∈ D D if and only if N L ( x ) ⊂ D \mathcal{N}_{L}(x) \subset D N L ( x ) ⊂ D for every D ∈ D D \in\mathcal{D} D ∈ D so that Λ ( ⋂ D ∈ D D ) ( v ) = ⋁ D ∈ D Λ D ( v ) \Lambda(\bigcap_{D \in\mathcal{D}}D)(v) = \bigvee_{D \in\mathcal{D}}\Lambda D(v) Λ ( ⋂ D ∈ D D ) ( v ) = ⋁ D ∈ D Λ D ( v ) for every v > 0 v > 0 v > 0 . It is easily checked that
adh Λ F ( v ) = { F ∪ { 1 } if 0 ∈ F , F if 0 ∉ F \operatorname{adh}_{\Lambda} F(v) =
\begin{cases}
F \cup\{1\} & \text{if } 0 \in F,\\
F & \text{if } 0 \notin F
\end{cases} adh Λ F ( v ) = { F ∪ { 1 } F if 0 ∈ F , if 0 ∈ / F for every F ∈ C R F \in\mathcal{C}_{\mathbb{R}} F ∈ C R and v > 0 v > 0 v > 0 . In particular, 0 ∈ adh Λ F ( ϵ ) 0 \in\operatorname{adh}_{\Lambda} F(\epsilon) 0 ∈ adh Λ F ( ϵ ) if and only if 0 ∈ F 0 \in F 0 ∈ F and thus
adh Λ ( adh Λ F ( ϵ ) ) ( v ) = adh Λ F ( v ⊗ ϵ ) \operatorname{adh}_{\Lambda}(\operatorname{adh}_{\Lambda} F(\epsilon))(v) = \operatorname{adh}_{\Lambda} F(v \otimes\epsilon) adh Λ ( adh Λ F ( ϵ )) ( v ) = adh Λ F ( v ⊗ ϵ ) for every F ∈ C R F \in\mathcal{C}_{\mathbb{R}} F ∈ C R and v , ϵ ∈ V v,\epsilon\in V v , ϵ ∈ V . Hence ( L , Λ ) (L,\Lambda) ( L , Λ ) is a weakly spatial C A p p \mathcal{C}^{\mathrm{App}} C App -object. Note, however, that it is not F 0 \mathbb{F}_{0} F 0 -adherence determined, hence not a C A p \mathcal{C}^{\mathrm{Ap}} C Ap -object. Indeed, the filter F = N R ( 0 ) ∩ C R \mathcal{F} = \mathcal{N}_{\mathbb{R}}(0) \cap\mathcal{C}_{\mathbb{R}} F = N R ( 0 ) ∩ C R satisfies Λ F ( 1 ) = ∅ \Lambda\mathcal{F}(1) = \varnothing Λ F ( 1 ) = ∅ while ⋂ C ∈ F # adh Λ C ( 1 ) = { 0 , 1 } \bigcap_{C \in\mathcal{F}^{\#}}\operatorname{adh}_{\Lambda} C(1) = \{0,1\} ⋂ C ∈ F # adh Λ C ( 1 ) = { 0 , 1 } . To see that pt L \operatorname{pt} L pt L is not an approach space, let A = ( 0 , 1 ) ⊂ pt L = R A = (0,1) \subset\operatorname{pt} L = \mathbb{R} A = ( 0 , 1 ) ⊂ pt L = R and ϵ = 1 \epsilon= 1 ϵ = 1 . The filter G ∈ F P ( R ) \mathcal{G} \in\mathbb{F}\mathbb{P}(\mathbb{R}) G ∈ FP ( R ) generated by { ( 0 , δ ) : 0 < δ < 1 } \{(0,\delta) : 0 < \delta< 1\} {( 0 , δ ) : 0 < δ < 1 } meshes with A A A and G ∘ ⊃ N L ( 0 ) \mathcal{G}^{\circ} \supset\mathcal{N}_{L}(0) G ∘ ⊃ N L ( 0 ) so that 0 ∈ A ( 1 ) 0 \in A^{(1)} 0 ∈ A ( 1 ) . Note that a proper filter finer than { 0 , 1 } ↑ = N L ( 1 ) \{0,1\}^{\uparrow} = \mathcal{N}_{L}(1) { 0 , 1 } ↑ = N L ( 1 ) cannot
mesh with A A A so that 1 ∉ A ( 1 ) = [ 0 , 1 ) 1 \notin A^{(1)} = [0,1) 1 ∈ / A ( 1 ) = [ 0 , 1 ) . On the other hand, adh p t L A ( 1 ) ( 1 ) = 1 \operatorname{adh}_{\mathrm{pt}\ L} A^{(1)}(1)=1 adh pt L A ( 1 ) ( 1 ) = 1 because 0 ∈ A ( 1 ) 0\in A^{(1)} 0 ∈ A ( 1 ) and { 0 } ∘ = ↑ L { 0 } ⊋ N L ( 1 ) \{0\}^{\circ}=\mathop{\uparrow}\nolimits_L\{0\}\supsetneq N_L(1) { 0 } ∘ = ↑ L { 0 } ⊋ N L ( 1 ) . Hence
adh p t L A ( 1 ) = 0 < 1 = adh p t L A ( 1 ) ( 1 ) ⊗ 1 \operatorname{adh}_{\mathrm{pt}\ L} A(1)=0<1=\operatorname{adh}_{\mathrm{pt}\ L} A^{(1)}(1)\otimes1 adh pt L A ( 1 ) = 0 < 1 = adh pt L A ( 1 ) ( 1 ) ⊗ 1
and (5.4) fails.
Measure of compactness
To illustrate how, in contrast to C C o n v \mathbf{C}^{\mathrm{Conv}} C Conv , index analysis in the spirit of [11 ] can be conducted in C C a p \mathbf{C}^{\mathrm{Cap}} C Cap , let us consider a measure of compactness in C H C a p \mathbf{CH}^{\mathrm{Cap}} CH Cap , in which we can get results generalizing the classical results on compactness. We include an illustrative example generalizing Proposition 1 and deeper results such as generalizations of Tychonoff’s and Mrowka-Kuratowski theorems will be the subject of future research.
[1 ] introduced a notion of compactness in C C o n v \mathbf{C}^{\mathrm{Conv}} C Conv and provided among other things a generalization of Proposition 1 to that context. On the other hand, [14 Theorem 11] provides a generalization of Proposition 1 to C a p \mathbf{Cap} Cap in terms of measure of compactness. In particular, if ( X , λ ) (X,\lambda) ( X , λ ) is a convergence approach space, D \mathbb{D} D is a transferable class of (set-theoretic) filters, F ∈ F P X \mathcal{F}\in\mathbb{F}PX F ∈ F P X and A ⊂ X A\subset X A ⊂ X then the measure of D \mathbb{D} D -compactness of F \mathcal{F} F at A A A is given by
c D A ( F ) = ⋀ D ∋ D # F ⋁ a ∈ A adh λ D ( a ) , c_{\mathbb{D}}^{A}(\mathcal{F})=\bigwedge_{\mathbb{D}\ni\mathcal{D}\#\mathcal{F}}\bigvee_{a\in A}\operatorname{adh}_{\lambda}\mathcal{D}(a), c D A ( F ) = D ∋ D # F ⋀ a ∈ A ⋁ adh λ D ( a ) ,
and [14 Theorem 11] implies
Proposition 51. If ( X , λ X ) (X,\lambda_X) ( X , λ X ) and ( Y , λ Y ) (Y,\lambda_Y) ( Y , λ Y ) are convergence approach spaces, D \mathbb{D} D is a transferable class of filters, F ∈ F P X \mathcal{F}\in\mathbb{F}PX F ∈ F P X , A ⊂ X A\subset X A ⊂ X and f ∈ C ( X , Y ) f\in\mathcal{C}(X,Y) f ∈ C ( X , Y ) then
c D f ( A ) ( f [ F ] ) ≥ c D A ( F ) . c_{\mathbb{D}}^{f(A)}(f[\mathcal{F}])\geq c_{\mathbb{D}}^{A}(\mathcal{F}). c D f ( A ) ( f [ F ]) ≥ c D A ( F ) .
We are now in a position to propose a generalization of Proposition 51 to C C a p \mathbf{C}^{\mathrm{Cap}} C Cap , in a way that extends Proposition 51 even in C a p \mathbf{Cap} Cap . Consider the measure of D \mathbb{D} D -compactness of F \mathcal{F} F at ℓ \ell ℓ given by
c D ℓ ( F ) = ⋁ { v ∈ V : ∀ D ∈ D L , D # F ⇒ ℓ ∈ ( adh D ( v ) ) # } . c_{\mathbb{D}}^{\ell}(\mathcal{F})=\bigvee\{v\in V:\forall\mathcal{D}\in\mathbb{D}L,\ \mathcal{D}\#\mathcal{F}\mathrel{\Rightarrow}\ell\in(\operatorname{adh}\mathcal{D}(v))^{\#}\}. c D ℓ ( F ) = ⋁ { v ∈ V : ∀ D ∈ D L , D # F ⇒ ℓ ∈ ( adh D ( v ) ) # } .
Let us call a class D \mathbb{D} D of filters lattice-transferable if
D ∈ D L ⇒ φ ( D ) ∈ D L ′ \mathcal{D}\in\mathbb{D}L\mathrel{\Rightarrow}\varphi(\mathcal{D})\in\mathbb{D}L' D ∈ D L ⇒ φ ( D ) ∈ D L ′
whenever φ ∈ CH ( L , L ′ ) \varphi\in\operatorname{CH}(L,L') φ ∈ CH ( L , L ′ ) . Recall the definition (5.6) of M ( φ ) \operatorname{M}(\varphi) M ( φ ) .
Theorem 52. Let φ ∈ CH ( L , L ′ ) \varphi\in\operatorname{CH}(L,L') φ ∈ CH ( L , L ′ ) , F ∈ F L ′ \mathcal{F}\in\mathbb{F}L' F ∈ F L ′ and m ′ ∈ L ′ m'\in L' m ′ ∈ L ′ and D \mathbb{D} D a lattice-transferable class of filters. Then
c D m ′ ( F ) ⊗ M ( φ ) ≤ c D φ ℓ ( m ′ ) ( φ ℓ ( F ) ) . c_{\mathbb{D}}^{m'}(\mathcal{F})\otimes\operatorname{M}(\varphi)\leq c_{\mathbb{D}}^{\varphi\ell(m')}(\varphi_{\ell}(\mathcal{F})). c D m ′ ( F ) ⊗ M ( φ ) ≤ c D φ ℓ ( m ′ ) ( φ ℓ ( F )) . Proof. Let v ∈ V v\in V v ∈ V with adh L ′ L ( v ) # m ′ \operatorname{adh}_{L'}\mathcal{L}(v)\#m' adh L ′ L ( v ) # m ′ whenever L ∈ D L ′ \mathcal{L}\in\mathbb{D}L' L ∈ D L ′ with L # F \mathcal{L}\#\mathcal{F} L # F . Let now D ∈ D L \mathcal{D}\in\mathbb{D}L D ∈ D L with D # φ ℓ ( F ) \mathcal{D}\#\varphi_{\ell}(\mathcal{F}) D # φ ℓ ( F ) , equivalently, φ ( D ) # F \varphi(\mathcal{D})\#\mathcal{F} φ ( D ) # F by Lemma 6. As D \mathbb{D} D is transferable, φ ( D ) ∈ D L ′ \varphi(\mathcal{D})\in\mathbb{D}L' φ ( D ) ∈ D L ′ so that adh L ′ φ ( D ) ( v ) # m ′ \operatorname{adh}_{L'}\varphi(\mathcal{D})(v)\#m' adh L ′ φ ( D ) ( v ) # m ′ . Moreover, in view of Theorem 30,
adh L ′ φ ( D ) ( v ) ≤ φ ( adh L φ ℓ ( φ ( D ) ) ( v ⊗ M ( φ ) ) ) ≤ φ ( adh L D ( v ⊗ M ( φ ) ) ) \operatorname{adh}_{L'}\varphi(\mathcal{D})(v)\leq\varphi(\operatorname{adh}_{L}\varphi_{\ell}(\varphi(\mathcal{D}))(v\otimes\operatorname{M}(\varphi)))\leq\varphi(\operatorname{adh}_{L}\mathcal{D}(v\otimes\operatorname{M}(\varphi))) adh L ′ φ ( D ) ( v ) ≤ φ ( adh L φ ℓ ( φ ( D )) ( v ⊗ M ( φ ))) ≤ φ ( adh L D ( v ⊗ M ( φ ))) because D ⊂ φ ℓ ( φ ( D ) ) \mathcal{D}\subset\varphi_{\ell}(\varphi(\mathcal{D})) D ⊂ φ ℓ ( φ ( D )) . Hence φ ( adh L D ( v ⊗ M ( φ ) ) ) # m ′ \varphi(\operatorname{adh}_{L}\mathcal{D}(v\otimes\operatorname{M}(\varphi)))\#m' φ ( adh L D ( v ⊗ M ( φ ))) # m ′ , equivalently, adh L D ( v ⊗ M ( φ ) ) # φ ℓ ( m ′ ) \operatorname{adh}_{L}\mathcal{D}(v\otimes\operatorname{M}(\varphi))\#\varphi_{\ell}(m') adh L D ( v ⊗ M ( φ )) # φ ℓ ( m ′ ) which yields the conclusion. □ \square □
Remark 53. Recall that the classical case corresponds to φ = f − 1 : P Y → P X \varphi=f^{-1}: \mathrm{P}Y \to\mathrm{P}X φ = f − 1 : P Y → P X , in which case φ ℓ = f : P X → P Y \varphi_\ell=f:\mathrm{P}X \to\mathrm{P}Y φ ℓ = f : P X → P Y .
Translated in Cap, we obtain the following generalization of Proposition 51 that accounts for maps that are not morphisms:
Corollary 54. If ( X , λ X ) (X,\lambda_X) ( X , λ X ) and ( Y , λ Y ) (Y,\lambda_Y) ( Y , λ Y ) are convergence approach spaces, D \mathbb{D} D is a transferable class of filters, F ∈ F P X \mathcal{F}\in\mathbb{F}\mathrm{P}X F ∈ F P X , A ⊂ X A\subset X A ⊂ X and f : X → Y f:X\to Y f : X → Y then
D D A ( F ) ⊗ m ( f ) ≤ c D f ( A ) ( f [ F ] ) , \mathcal{D}_{\mathbb{D}}^{A}(\mathcal{F})\otimes\mathrm{m}(f)\leq c_{\mathbb{D}}^{f(A)}(f[\mathcal{F}]), D D A ( F ) ⊗ m ( f ) ≤ c D f ( A ) ( f [ F ]) , where m ( f ) \mathrm{m}(f) m ( f ) is defined by (5.5).
Department of Mathematics, Kean University, Jersey City campus, 2039 Kennedy Blvd, Jersey City, NJ 07305
Department of Science and Math, Arkansas State University, Campus Querétaro, Querétaro, México
Email address: eapzfc@gmail.com
Email address: fmynard@kean.edu