module Homotopy.Space.Suspension.Properties whereProperties of suspensions🔗
Connectedness🔗
This section contains the aforementioned proof that suspension increases the connectedness of a space.
Susp-is-connected
: ∀ {ℓ} {A : Type ℓ} n
→ is-n-connected A n → is-n-connected (Susp A) (suc n)
Susp-is-connected 0 a-conn = inc N
Susp-is-connected 1 a-conn = ∥-∥-out! do
pt ← a-conn
pure $ is-connected∙→is-connected λ where
N → inc refl
S → inc (sym (merid pt))
(merid x i) → is-prop→pathp (λ i → ∥_∥.squash {A = merid x i ≡ N})
(inc refl) (inc (sym (merid pt))) i
Susp-is-connected {A = A} (suc (suc n)) a-conn =
n-Tr-elim
(λ _ → is-n-connected (Susp A) (3 + n))
(λ _ → is-prop→is-hlevel-suc {n = suc n} (hlevel 1))
(λ x → contr (inc N) (n-Tr-elim _ (λ _ → is-hlevel-suc (2 + n) (n-Tr-is-hlevel (2 + n) _ _))
(Susp-elim _ refl (ap n-Tr.inc (merid x)) λ pt →
commutes→square (ap (refl ∙_) (rem₂ .snd _ ∙ sym (rem₂ .snd _))))))
(conn' .centre)
where
conn' : is-contr (n-Tr A (2 + n))
conn' = is-n-connected-Tr (1 + n) a-conn
rem₁ : is-equiv λ b a → b
rem₁ = is-n-connected→n-type-const
{B = n-Tr.inc {n = 3 + n} N ≡ inc S} {A = A}
(suc n) (hlevel (2 + n)) a-conn
rem₂ : Σ (inc N ≡ inc S) (λ p → ∀ x → ap n-Tr.inc (merid x) ≡ p)
rem₂ = _ , λ x → sym (rem₁ .is-eqv _ .centre .snd) $ₚ xAs a direct corollary, the is (remember that our indices are offset by 2).
Sⁿ⁻¹-is-connected : ∀ n → is-n-connected (Sⁿ⁻¹ n) n
Sⁿ⁻¹-is-connected zero = _
Sⁿ⁻¹-is-connected (suc n) = Susp-is-connected n (Sⁿ⁻¹-is-connected n)instance
Connected-Susp : ∀ {ℓ} {A : Type ℓ} {n} → ⦃ Connected A n ⦄ → Connected (Susp A) (suc n)
Connected-Susp {n = n} ⦃ conn-instance c ⦄ = conn-instance (Susp-is-connected n c)Truncatedness🔗
While there is no similarly pleasant characterisation of the truncatedness of suspensions1, we can give a few special cases.
First, the suspension of a contractible type is contractible.
Susp-is-contr
: ∀ {ℓ} {A : Type ℓ}
→ is-contr A → is-contr (Susp A)
Susp-is-contr c .centre = N
Susp-is-contr c .paths N = refl
Susp-is-contr c .paths S = merid (c .centre)
Susp-is-contr c .paths (merid a i) j = hcomp (∂ i ∨ ∂ j) λ where
k (k = i0) → merid (c .centre) (i ∧ j)
k (i = i0) → N
k (j = i0) → N
k (i = i1) → merid (c .centre) j
k (j = i1) → merid (c .paths a k) iNotice the similarity with the proof that the is contractible: that argument is essentially a recursive version of this one, since is its own suspension.
Going up a level, we do not have that the suspension of a proposition is a proposition (think of the suspension of but we do have that the suspension of a proposition is a set.
module _ {ℓ} {A : Type ℓ} (prop : is-prop A) whereWe start by defining a helper induction principle: in order to map out of it suffices to give values at the four poles, and, assuming holds, a square over the meridians with the given corners.
Susp-prop-elim²
: ∀ {ℓ} {B : Susp A → Susp A → Type ℓ}
→ (bnn : B N N) (bns : B N S)
→ (bsn : B S N) (bss : B S S)
→ ((a : A) → (i j : I) → B (merid a i) (merid a j)
[ _ ↦ (λ { (i = i0) (j = i0) → bnn
; (i = i0) (j = i1) → bns
; (i = i1) (j = i0) → bsn
; (i = i1) (j = i1) → bss }) ])
→ ∀ a b → B a b
Susp-prop-elim² bnn bns bsn bss bm = Susp-elim _
(Susp-elim _ bnn bns λ a j → outS (bm a i0 j))
(Susp-elim _ bsn bss λ a j → outS (bm a i1 j))
λ a → funextP (Susp-elim _
(λ i → outS (bm a i i0))
(λ i → outS (bm a i i1))
(subst-prop prop a (λ j i → outS (bm a i j))))Then, we use the usual machinery of identity systems: we define a type family of “codes” of equality for Its values are either for equal poles or for different poles, and we fill the square using univalence.
private
Code : Susp A → Susp A → Type ℓ
Code = Susp-prop-elim² (Lift _ ⊤) A A (Lift _ ⊤)
λ a i j → inS (double-connection (sym (A≡⊤ a)) (A≡⊤ a) i j)
where
A≡⊤ : A → A ≡ Lift _ ⊤
A≡⊤ a = ua (prop-ext prop (hlevel 1) _ (λ _ → a))We’ve defined a reflexive family of propositions:
Code-is-prop : ∀ a b → is-prop (Code a b)
Code-is-prop = Susp-elim-prop (λ _ → hlevel 1)
(Susp-elim-prop (λ _ → hlevel 1) (hlevel 1) prop)
(Susp-elim-prop (λ _ → hlevel 1) prop (hlevel 1))
Code-refl : ∀ a → Code a a
Code-refl = Susp-elim-prop (λ a → Code-is-prop a a) _ _Thus all that remains is to prove that it implies equality. At the
poles, we can use refl and merid.
_ = I-interp decode : ∀ a b → Code a b → a ≡ b
decode = Susp-prop-elim²
(λ _ → refl) (λ c → merid c)
(λ c → sym (merid c)) (λ _ → refl)This time, if holds, we have to fill a cube with the given four edges:
Notice that we have two different meridians:
comes from our assumption that
holds, whereas
comes from the function out of codes we’re trying to build. If
and
were the same, we could simply fill this cube by interpolating
between
and
along
However, we can take a shortcut: since we’ve assumed
holds, and
is a proposition, then
is contractible, and we’ve shown that the suspension of a
contractible type is contractible, so we can trivially extend our
partial system to fill the desired cube!
λ a i j → is-contr→extend
(Π-is-hlevel 0 (λ _ → Path-is-hlevel 0
(Susp-is-contr (is-prop∙→is-contr prop a))))
(∂ i ∧ ∂ j) _This concludes the proof that is a set with
Code-ids : is-identity-system Code Code-refl
Code-ids = set-identity-system Code-is-prop (decode _ _)
opaque
Susp-prop-is-set : is-set (Susp A)
Susp-prop-is-set = identity-system→hlevel 1 Code-ids Code-is-prop
Susp-prop-path : Path (Susp A) N S ≃ A
Susp-prop-path = identity-system-gives-path Code-ids e⁻¹