module Cat.Univalent.Rezk.HIT {o ℓ} (C : Precategory o ℓ) where
Higher inductive Rezk completions🔗
We can define the Rezk completion of a precategory directly as a higher inductive type, without passing through replacement. Importantly, under this construction, the resulting universal functor becomes definitionally fully faithful.
The type of objects of looks a lot like the delooping of a group, but with an inclusion rather than a single basepoint: indeed, we can think of it as delooping all the automorphism groups of at once. Completely analogously, we have a constructor turning isomorphisms in into paths in and we have a generating triangle saying that this constructor respects composition, filling the diagram
data C^ : Type (o ⊔ ℓ) where inc : ⌞ C ⌟ → C^ glue : ∀ {x y} → x ≅ y → inc x ≡ inc y glueᵀ : ∀ {x y z} (f : x ≅ y) (g : y ≅ z) → Triangle (glue f) (glue (g ∘Iso f)) (glue g) squash : is-groupoid C^
Note that, as in the case for simple deloopings, this generating
triangle is sufficient to ensure that glue
is functorial.
glue-∘ : (e : x ≅ y) (e' : y ≅ z) → Path (Path C^ (inc x) (inc z)) (glue e ∙ glue e') (glue (e' ∘Iso e)) glue-∘ e e' = sym (triangle→commutes (glueᵀ e e')) glue-id : Path (Path C^ (inc x) (inc x)) (glue id-iso) refl glue-id = glue id-iso ≡⟨ ∙-intror (∙-invr _) ⟩≡ glue id-iso ∙ glue id-iso ∙ sym (glue id-iso) ≡⟨ ∙-pulll (glue-∘ _ _ ∙ ap C^.glue (ext (idl _))) ⟩≡ glue id-iso ∙ sym (glue id-iso) ≡⟨ ∙-invr _ ⟩≡ refl ∎
We will need an elimination principle for C^
into sets, saying that it
suffices to handle the point inclusions and the generating paths; the
generating triangle is handled automatically, as is the truncation. Of
course, we can also eliminate C^
into propositions, in which case the generating
paths are also handled automatically.
C^-elim-set : ∀ {ℓ} {P : C^ → Type ℓ} ⦃ _ : ∀ {x} → H-Level (P x) 2 ⦄ → (pi : ∀ x → P (inc x)) → (pg : ∀ {x y} (e : x ≅ y) → PathP (λ i → P (glue e i)) (pi x) (pi y)) → ∀ x → P x C^-elim-set pi pg (inc x) = pi x C^-elim-set pi pg (glue x i) = pg x i C^-elim-set {P = P} pi pg (glueᵀ {x} f g i j) = is-set→squarep (λ i j → hlevel {T = P (glueᵀ f g i j)} 2) (λ i → pi x) (λ i → pg f i) (λ i → pg (g ∘Iso f) i) (λ i → pg g i) i j C^-elim-set {P = P} pi pg (squash x y p q α β i j k) = is-hlevel→is-hlevel-dep 2 (λ x → is-hlevel-suc 2 (hlevel {T = P x} 2)) (go x) (go y) (λ i → go (p i)) (λ i → go (q i)) (λ i j → go (α i j)) (λ i j → go (β i j)) (squash x y p q α β) i j k where go = C^-elim-set {P = P} pi pg
abstract C^-elim-prop : ∀ {ℓ} {P : C^ → Type ℓ} ⦃ _ : ∀ {x} → H-Level (P x) 1 ⦄ → (∀ x → P (inc x)) → ∀ x → P x C^-elim-prop pi = C^-elim-set ⦃ hlevel-instance (is-prop→is-set (hlevel 1)) ⦄ pi (λ e → prop!) instance H-Level-C^ : ∀ {n} → H-Level C^ (3 + n) H-Level-C^ = basic-instance 3 squash Inductive-C^ : ∀ {ℓ ℓm} {P : C^ → Type ℓ} ⦃ i : Inductive (∀ x → P (inc x)) ℓm ⦄ → ⦃ _ : ∀ {x} → H-Level (P x) 1 ⦄ → Inductive (∀ x → P x) ℓm Inductive-C^ ⦃ i ⦄ = record { methods = i .Inductive.methods ; from = λ f → C^-elim-prop (i .Inductive.from f) }
Defining
spans over
C^
🔗
We now turn to the problem of defining the This turns out to have a lot of “cases”, but we can break them down intuitively as follows: to define a function where is a groupoid, we can start by giving a function
record C^-corr {ℓ'} (P : Type ℓ') : Type (o ⊔ ℓ ⊔ ℓ') where field has-is-groupoid : is-groupoid P base : ⌞ C ⌟ → ⌞ C ⌟ → P
Then, we must give actions of the isomorphisms on both the left and the right of making it into a sort of “profunctor”; and, correspondingly, these actions must be “profunctorial”. In particular, they should respect composition and commute past each other, which we can state concisely in terms of triangles and squares.
lmap : ∀ {x x' y} (e : x ≅ x') → base x y ≡ base x' y rmap : ∀ {x y y'} (e : y ≅ y') → base x y ≡ base x y' lcoh : ∀ {x x' x'' y} (e : x ≅ x') (e' : x' ≅ x'') → Triangle (lmap {y = y} e) (lmap (e' ∘Iso e)) (lmap e') rcoh : ∀ {x y y' y''} (e : y ≅ y') (e' : y' ≅ y'') → Triangle (rmap {x = x} e) (rmap (e' ∘Iso e)) (rmap e') comm : ∀ {x x' y y'} (e : x ≅ x') (e' : y ≅ y') → Square (rmap e) (lmap e') (lmap e') (rmap e)
This is sufficient to discharge all the cases when writing a binary
function from
into a groupoid; we leave the formalisation in this <details>
block because it is rather fiddly.
private instance _ : H-Level P 3 _ = hlevel-instance has-is-groupoid go₀ : (x : ⌞ C ⌟) (y : C^) → P go₀ ξ (inc x) = base ξ x go₀ ξ (glue x i) = rmap {ξ} x i go₀ ξ (glueᵀ f g i j) = rcoh {ξ} f g i j go₀ ξ (squash x y p q α β i j k) = let go = go₀ ξ in is-hlevel→is-hlevel-dep 2 (λ _ → hlevel 3) (go x) (go y) (λ i → go (p i)) (λ i → go (q i)) (λ i j → go (α i j)) (λ i j → go (β i j)) (squash x y p q α β) i j k lmap' : ∀ {x x'} (e : x ≅ x') → go₀ x ≡ go₀ x' lmap' e = funextP $ C^-elim-set (λ _ → lmap e) λ e' → comm e' e lcoh' : ∀ {w x y : ⌞ C ⌟} (e : w ≅ x) (e' : x ≅ y) → Triangle (lmap' e) (lmap' (e' ∘Iso e)) (lmap' e') lcoh' e e' = funext-square $ C^-elim-prop λ x → lcoh e e' C^-rec₂ : (x y : C^) → P C^-rec₂ (inc x) z = go₀ x z C^-rec₂ (glue x i) z = lmap' x i z C^-rec₂ (glueᵀ x y i j) z = lcoh' x y i j z C^-rec₂ (squash x y p q α β i j k) z = let go : C^ → P go x = C^-rec₂ x z in is-hlevel→is-hlevel-dep 2 (λ _ → hlevel 3) (go x) (go y) (λ i → go (p i)) (λ i → go (q i)) (λ i j → go (α i j)) (λ i j → go (β i j)) (squash x y p q α β) i j k
Since is already a profunctor over we can show straightforwardly that it extends to a binary type family over
hom^ : C^ → C^ → Type ℓ hom^ x y = ⌞ C^-rec₂ methods x y ⌟ where methods : C^-corr (Set ℓ) methods .has-is-groupoid = hlevel 3 methods .base x y = el! (Hom x y) methods .lmap e = n-ua (dom-iso→hom-equiv e) methods .rmap e = n-ua (cod-iso→hom-equiv e) methods .lcoh e e' = n-ua-triangle (ext λ h → assoc _ _ _) methods .rcoh e e' = n-ua-triangle (ext λ h → sym (assoc _ _ _)) methods .comm e e' = n-ua-square (ext λ h → sym (assoc _ _ _))
-- To make sure that composition in Rzk is injective in the objects, we -- wrap the hom^ family defined above in a record. record Hom^ (x y : C^) : Type ℓ where constructor lift field lower : hom^ x y open Hom^ public instance H-Level-Hom^ : ∀ {x y n} → H-Level (Hom^ x y) (2 + n) H-Level-Hom^ = basic-instance 2 (Iso→is-hlevel 2 (Hom^.lower , (iso lift (λ x → refl) (λ x → refl))) (hlevel 2)) Inductive-Hom^ : ∀ {x y ℓ ℓm} {P : Hom^ x y → Type ℓ} ⦃ i : Inductive (∀ h → P (lift h)) ℓm ⦄ → Inductive (∀ x → P x) ℓm Inductive-Hom^ ⦃ i ⦄ = record { methods = i .Inductive.methods ; from = λ { m (lift x) → i .Inductive.from m x } } private lifthom^ : ∀ {x y} → hom^ x y → Hom^ x y lifthom^ = lift
Making a category🔗
Since our
is valued in sets, we can use the eliminator defined above to construct
the identity morphisms and the composition operation by
These will consist of lifting the corresponding operation from
and then showing that they respect the action of glue
on hom^
, which we have
defined to be pre- and post-composition with the given isomorphism.
Therefore, while there is a lot of code motion to put these together,
they are conceptually very simple.
For a worked-out example, the necessary coherence for lifting the identity morphisms from to asks simply that if then we have which is a short calculation.
id^ : ∀ x → hom^ x x id^ = C^-elim-set (λ _ → id) coh where abstract coh : ∀ {x y} (j : x ≅ y) → PathP (λ i → hom^ (glue j i) (glue j i)) id id coh z = path→ua-pathp _ $ z .to ∘ id ∘ z .from ≡⟨ ap (z .to ∘_) (idl _) ⟩≡ z .to ∘ z .from ≡⟨ z .invl ⟩≡ id ∎
Lifting the composition operation is similar, but more involved, since
we now have to do recursion on C^
thrice.
∘^ : ∀ x y z → hom^ y z → hom^ x y → hom^ x z ∘^ = C^-elim-set f₁ coh₂ where mutual f₀ : ∀ x y z → hom^ (inc y) z → hom^ (inc x) (inc y) → hom^ (inc x) z f₀ x y = C^-elim-set (λ z → _∘_) (coh₀ x y) f₁ : ∀ x (y z : C^) → hom^ y z → hom^ (inc x) y → hom^ (inc x) z f₁ x = C^-elim-set (f₀ x) (coh₁ x) abstract coh₀ : ∀ x y {z z'} (j : z ≅ z') → PathP (λ i → hom^ (inc y) (glue j i) → hom^ (inc x) (inc y) → hom^ (inc x) (glue j i)) _∘_ _∘_ coh₀ x y j = ua→ (λ f → funextP λ g → path→ua-pathp _ (assoc _ _ _)) coh₁ : ∀ x {y z} (j : y ≅ z) → PathP (λ i → (y : C^) → hom^ (glue j i) y → hom^ (inc x) (glue j i) → hom^ (inc x) y) (f₀ x y) (f₀ x z) coh₁ x j = funextP (C^-elim-prop (λ z → ua→ λ h → ua→ λ i → ap (h ∘_) (insertl (j .invr)) ∙ pulll refl)) coh₂ : ∀ {x y} (j : x ≅ y) → PathP (λ i → (y z : C^) → hom^ y z → hom^ (glue j i) y → hom^ (glue j i) z) (f₁ x) (f₁ y) coh₂ j = funextP $ elim! λ y → funextP $ elim! λ z → funextP λ f → ua→ λ g → path→ua-pathp _ (sym (assoc _ _ _))
To show that this forms a precategory, it suffices to lift the corresponding proofs also from Since we’re showing a proposition, this is very straightforward: it’s just some un/wrapping.
Rzk : Precategory (o ⊔ ℓ) ℓ Rzk .P.Ob = C^ Rzk .P.Hom x y = Hom^ x y Rzk .P.Hom-set x y = hlevel 2 Rzk .P.id {x} = lift (id^ x) Rzk .P._∘_ {w} {x} {y} f g = lift (∘^ w x y (f .lower) (g .lower)) Rzk .P.idr {x} {y} (lift f) = ap lift (idr^ x y f) where abstract idr^ : ∀ x y (f : hom^ x y) → ∘^ x x y f (id^ x) ≡ f idr^ = elim! λ x y f → idr f Rzk .P.idl {x} {y} (lift f) = ap lift (idl^ x y f) where abstract idl^ : ∀ x y (f : hom^ x y) → ∘^ x y y (id^ y) f ≡ f idl^ = elim! λ x y f → idl f Rzk .P.assoc {w} {x} {y} {z} (lift f) (lift g) (lift h) = ap lift (assoc^ w x y z f g h) where abstract assoc^ : ∀ w x y z (f : hom^ y z) (g : hom^ x y) (h : hom^ w x) → ∘^ w y z f (∘^ w x y g h) ≡ ∘^ w x z (∘^ x y z f g) h assoc^ = elim! λ w x y z f g h → assoc f g h
We can give the Rezk completion functor.
complete : Functor C Rzk complete .Functor.F₀ = inc complete .Functor.F₁ f = lift f complete .Functor.F-id = refl complete .Functor.F-∘ f g = refl complete-is-ff : is-fully-faithful complete complete-is-ff = is-iso→is-equiv (iso Hom^.lower (λ x → refl) λ x → refl) complete-is-eso : is-eso complete complete-is-eso = elim! λ x → inc (x , Rzk.id-iso)
Univalence of the Rezk completion🔗
To show that
is univalent, we do one last monster
induction, to show that, for a fixed
the space of “isomorphs of
”
is contractible. This automatically handles one coherence, since
contractibility is a proposition; however, when actually lifting the
glue
constructor, we do
still have one coherence to show.
to-path^ : (a : C^) → is-contr (Σ[ b ∈ C^ ] a Rzk.≅ b) to-path^ = C^-elim-set (λ x → record { paths = λ (b , e) → wrap x b e }) (λ e → prop!) where module _ (a : ⌞ C ⌟) where T : C^ → Type _ T x = (e : inc a Rzk.≅ x) → Path (Σ[ b ∈ C^ ] inc a Rzk.≅ b) (inc a , Rzk.id-iso) (x , e) glue' : (x : ⌞ C ⌟) → T (inc x) glue' x e = glue (hat.iso.from e) ,ₚ Rzk.≅-pathp _ _ (apd (λ i → lifthom^) (path→ua-pathp _ (idr _)))
However, this coherence is essentially the generating triangle glueᵀ
, and so the proof
goes through without much stress.
coh : {x y : Ob} (e : x ≅ y) → PathP (λ i → T (glue e i)) (glue' x) (glue' y) coh e = funext-dep λ {x₀} {x₁} α → let open Rzk using (to) α' : glue (e ∘Iso hat.iso.from x₀) ≡ glue (hat.iso.from x₁) α' = ap C^.glue (ext (ua-pathp→path _ (apd (λ _ x → x .to .lower) α))) in Σ-set-square (λ _ → hlevel 2) $ glueᵀ (hat.iso.from x₀) e ▷ α' wrap : ∀ b → T b wrap = C^-elim-set glue' coh Rzk-is-category : is-category Rzk Rzk-is-category = contr→identity-system (λ {a} → to-path^ a)