Global dimension at most two (proof of Proposition 2.10) #
graded-category.tex: "At a nonprojective label X, the exact right
mesh … give 0 → R_{τX} → ⊕_{Y→X} R_Y → R_X → L_X → 0. At a projective label
P, the same lemma gives a resolution 0 → R_{rad P} → R_P → L_P → 0 of length
at most one. Hence gl.dim Λ ≤ 2." The proof of Lemma 2.9 uses this in the form
"its kernel is a second syzygy and hence projective".
We derive this consequence: every finite-dimensional module has a projective
resolution of length two by modules R_Z (the simple resolutions and the
horseshoe construction, along a composition series); by Schanuel's lemma every
second syzygy is projective; a finite-dimensional projective module is some
R_W (tops and Nakayama's lemma).
A surjection onto a projective module splits: X ≅ ker π × P.
Instances For
Schanuel's lemma: for surjections e : P → M, e' : P' → M from projective
modules, ker e × P' ≅ ker e' × P.
One horseshoe step. Let 0 → A' → A → A'' → 0 be exact
(i injective, p surjective, ker p = range i) and let e' : P' → A',
e'' : P'' → A'' be surjections with P'' projective. Then
e : P' × P'' → A, (a, b) ↦ i (e' a) + λ b with p λ = e'', is surjective,
and its kernel is an extension of ker e'' by ker e'.
An extension of a projective module by a projective module is projective.
A resolution of length two: P₁ → P₀ → M → 0 with projective P₀, P₁ and
projective kernel of P₁ → ker (P₀ → M).
Instances For
Horseshoe: resolutions of length two pass to extensions.
The simple tops have resolutions of length two (the exact right meshes).
Every finite-dimensional module has a resolution of length two.
Second syzygies are projective (gl.dim Λ ≤ 2).
An endomorphism of R_x inducing the identity on the top is invertible.
A split surjection: P ≅ ker r × Q for r ∘ h = id.
Instances For
A finite-dimensional projective module is R_W for a family W.
Kernels of maps between families are families (gl.dim Λ ≤ 2).