Ordered Locale Skill

"We extend Stone duality between topological spaces and locales to include order." — Heunen & van der Schaaf, 2024

Overview

An ordered locale is a locale (point-free topological space) equipped with a compatible preorder satisfying the open cone condition.

Ordered Locale = Frame + Preorder + Open Cones

Key Definitions

Frame (Locale)

A frame is a complete lattice where finite meets distribute over arbitrary joins:

a ∧ (⋁ᵢ bᵢ) = ⋁ᵢ (a ∧ bᵢ)

Equivalently: a complete Heyting algebra.

Open Cone Condition

For a preorder ≤ on locale L, the open cone condition requires:

↑x = {y ∈ L : x ≤ y}  is an open in L  (upper cone)
↓x = {y ∈ L : y ≤ x}  is an open in L  (lower cone)

This ensures the order is "visible" to the topology.

Ordered Locale (Definition 2.1, Heunen-van der Schaaf)

An ordered locale is a tuple (L, ≤) where:

1. L is a locale (frame of opens)
2. ≤ is a preorder on O(L)
3. The open cone condition holds

Stone Duality Extended

┌────────────────────────┐     adjunction     ┌─────────────────────┐
│ Preordered Topological │ ←───────────────→ │   Ordered Locales   │
│ Spaces (open cones)    │                    │     (spatial)       │
└────────────────────────┘                    └─────────────────────┘

Restricts to an equivalence:

• Spatial ordered locales ≃ Sober T₀-ordered spaces with open cones

Julia Implementation (Catlab.jl)

Frame as Subobject Heyting Algebra

From Catlab's Subobjects.jl:

using Catlab, Catlab.Theories, Catlab.CategoricalAlgebra

# Frame operations via ThSubobjectHeytingAlgebra
@signature ThFrame <: ThSubobjectHeytingAlgebra begin
  # Infinite joins (frame-specific)
  Join(I::TYPE, f::(I → Sub(X)))::Sub(X) ⊣ [X::Ob]
  
  # Frame distributivity: a ∧ (⋁ᵢ bᵢ) = ⋁ᵢ (a ∧ bᵢ)
end

Ordered Locale Schema (ACSet)

using ACSets

@present SchOrderedLocale(FreeSchema) begin
  Open::Ob                    # Opens of the frame
  Arrow::Ob                   # Order relation arrows
  
  src::Hom(Arrow, Open)       # Source of order arrow
  tgt::Hom(Arrow, Open)       # Target of order arrow
  
  # Cone structure
  Cone::Ob
  apex::Hom(Cone, Open)       # Apex of cone
  leg::Hom(Cone, Arrow)       # Legs of cone
  
  # Attributes
  Index::AttrType
  idx::Attr(Open, Index)      # Open index
  is_upper::Attr(Cone, Bool)  # Upper vs lower cone
end

@acset_type OrderedLocale(SchOrderedLocale)

Open Cone Verification

function verify_open_cone_condition(L::OrderedLocale)
  """
  Verify that ↑x and ↓x are opens for all x.
  """
  opens = parts(L, :Open)
  
  for x in opens
    # Upper cone: ↑x = {y : x ≤ y}
    upper_cone = Set{Int}()
    for arr in parts(L, :Arrow)
      if L[arr, :src] == x
        push!(upper_cone, L[arr, :tgt])
      end
    end
    push!(upper_cone, x)  # Reflexive
    
    # Check upper cone is an open (exists in frame)
    if !is_open(L, upper_cone)
      return false, "↑$x is not an open"
    end
    
    # Lower cone: ↓x = {y : y ≤ x}
    lower_cone = Set{Int}()
    for arr in parts(L, :Arrow)
      if L[arr, :tgt] == x
        push!(lower_cone, L[arr, :src])
      end
    end
    push!(lower_cone, x)  # Reflexive
    
    if !is_open(L, lower_cone)
      return false, "↓$x is not an open"
    end
  end
  
  return true, "Open cone condition satisfied"
end

Cone and Cocone Operations

From Catlab's Limits.jl:

using Catlab.Theories: ThCompleteCategory, ThCocompleteCategory

# Cone: A natural transformation Δ(X) → D
# where Δ(X) is the constant diagram at X
struct Cone{Ob, Hom}
  apex::Ob
  legs::Vector{Hom}  # One leg per object in diagram
end

# Cocone: Dual - natural transformation D → Δ(X)
struct Cocone{Ob, Hom}
  apex::Ob
  legs::Vector{Hom}
end

# Limit = universal cone
function limit_cone(diagram::FreeDiagram, model)
  # Find apex and legs such that any other cone factors through
  lim = limit[model](diagram)
  Cone(ob(lim), collect(legs(lim)))
end

# Colimit = universal cocone  
function colimit_cocone(diagram::FreeDiagram, model)
  colim = colimit[model](diagram)
  Cocone(ob(colim), collect(legs(colim)))
end

Frame Operations via Limits/Colimits

# Meet (∧) = pullback = limit of cospan
function frame_meet(L::OrderedLocale, a::Int, b::Int)
  # Pullback in the locale category
  diagram = FreeDiagram([a, b], [(a, top(L)), (b, top(L))])
  lim = limit_cone(diagram, L)
  return lim.apex
end

# Join (∨) = pushout = colimit of span
function frame_join(L::OrderedLocale, a::Int, b::Int)
  # Pushout in the locale category
  bot = bottom(L)
  diagram = FreeDiagram([a, b], [(bot, a), (bot, b)])
  colim = colimit_cocone(diagram, L)
  return colim.apex
end

# Infinite join (frame-specific)
function frame_sup(L::OrderedLocale, opens::Vector{Int})
  # Colimit of discrete diagram
  diagram = FreeDiagram(opens, Pair{Int,Int}[])
  colim = colimit_cocone(diagram, L)
  return colim.apex
end

# Heyting implication: a ⇒ b = ⋁{c : a ∧ c ≤ b}
function frame_implies(L::OrderedLocale, a::Int, b::Int)
  candidates = [c for c in parts(L, :Open) 
                if below_or_eq(L, frame_meet(L, a, c), b)]
  return frame_sup(L, candidates)
end

GF(3) Triads

acsets (-1) ⊗ ordered-locale (0) ⊗ topos-generate (+1) = 0 ✓  [Schema]
sheaf-cohomology (-1) ⊗ ordered-locale (0) ⊗ kan-extensions (+1) = 0 ✓  [Gluing]
directed-interval (-1) ⊗ ordered-locale (0) ⊗ mlx-apple-silicon (+1) = 0 ✓  [Scale]

Example Locales

1. Sierpinski Locale (2-point)

function sierpinski_locale()
  L = OrderedLocale{Bool}()
  add_parts!(L, :Open, 2, idx=[false, true])  # ⊥, ⊤
  add_part!(L, :Arrow, src=1, tgt=2)          # ⊥ < ⊤
  
  # Cones
  add_part!(L, :Cone, apex=2, is_upper=true)  # ↑⊥ = {⊥,⊤} = ⊤
  add_part!(L, :Cone, apex=1, is_upper=false) # ↓⊤ = {⊥,⊤} = ⊤
  L
end

2. Diamond Locale

function diamond_locale()
  L = OrderedLocale{Int}()
  add_parts!(L, :Open, 4, idx=[0, 1, 1, 2])  # ⊥, a, b, ⊤
  add_part!(L, :Arrow, src=1, tgt=2)  # ⊥ < a
  add_part!(L, :Arrow, src=1, tgt=3)  # ⊥ < b
  add_part!(L, :Arrow, src=2, tgt=4)  # a < ⊤
  add_part!(L, :Arrow, src=3, tgt=4)  # b < ⊤
  L
end

3. Scott Topology (Domain Theory)

function scott_locale(poset::OrderedLocale)
  """
  Scott topology: opens are upper sets closed under directed joins.
  """
  scott_opens = Vector{Set{Int}}()
  
  for u in parts(poset, :Open)
    upset = upper_cone(poset, u)
    if is_scott_open(poset, upset)
      push!(scott_opens, upset)
    end
  end
  
  scott_opens
end

function is_scott_open(L, U::Set{Int})
  # U is Scott-open iff:
  # 1. U is upper-closed
  # 2. For any directed D with ⋁D ∈ U, some d ∈ D is in U
  is_upper_closed(L, U) && is_inaccessible_by_directed_joins(L, U)
end

Comparison with Implementation

Aspect	ordered_locale_mlx.py	Proper Ordered Locale
Structure	Adjacency matrix (poset)	Frame + ACSet schema
Opens	Finite indices	Complete Heyting algebra
Meets/Joins	Graph search	Pullback/Pushout (categorical)
Cones	Not implemented	Cone/Cocone with legs
Open Cone Condition	Not checked	verify_open_cone_condition
Infinite joins	No	frame_sup

Applications

1. Spacetime Causality — Order = causal influence, no points
2. Domain Theory — Way-below relation, compact elements
3. Directed Homotopy — Irreversible paths (directed-interval)
4. Modal Logic — Accessibility relations as order
5. Concurrent Computing — Partial order of events

Commands

# Run ordered locale demo
just ordered-locale-demo

# Verify open cone condition
just ordered-locale-verify

# Scale test with MLX
just ordered-locale-mlx

Files

• ordered_locale.jl — Julia implementation with Catlab
• ordered_locale_mlx.py — MLX-accelerated (finite approximation)
• frame.jl — Frame operations via limits/colimits
• cone_cocone.jl — Cone/cocone constructions

References

1. Heunen, C. & van der Schaaf, N. (2024). "Ordered Locales." J. Pure Appl. Algebra 228(7), 107654.
2. Heunen, C. & van der Schaaf, N. (2025). "Causal Coverage in Ordered Locales and Spacetimes." arXiv:2510.17417.
3. Johnstone, P. (1982). Stone Spaces. Cambridge University Press.
4. Catlab.jl — src/theories/Limits.jl, src/categorical_algebra/cats/Subobjects.jl

---

Skill Name: ordered-locale Type: Point-free Topology / Frame Theory / Directed Order Trit: 0 (ERGODIC - coordinator) GF(3): Mediates between schema (-1) and generation (+1) Open Cones: ↑x and ↓x must be opens Duality: Spatial ordered locales ≃ Sober T₀-ordered spaces