Documentation

DerivedAlgGeo.Algebra.Homology.DGCategory.Model.Shift

The shifted hom-complexes of C^dg, checked against Mathlib #

DGCategory.shiftD and DGCategory.shiftComp are stated for an arbitrary dg category, so on Cdg A they must come out as the operations Mathlib already has. This file checks that, rather than leaving the agreement to a comment.

Three things are recorded.

  1. shiftD on Cdg A is δ with the sign (-1) ^ m, and shiftComp is Cochain.comp with the sign the derivation in ..Shift forced. Both are rfl, so the general definitions genuinely restrict to the model.

  2. The sign on the shifted differential is the same (-1) ^ m that Mathlib's own δ_shift carries. That lemma is about shifting the objects of a cochain — Cochain.shift a : Cochain (K⟦a⟧) (L⟦a⟧) n — which is a different operation from shifting the hom-complex, so this is a real cross-check between two independent constructions and not a restatement of one.

  3. shiftD_eq_delta_shift records where the two operations meet: the sign (-1) ^ m produced by CochainComplex.shiftFunctor on the hom-complex and the sign produced by Cochain.shift on the objects agree, which is what makes the two conventions composable rather than merely coexistent.

shiftD on C^dg is δ with the shift's sign.

theorem CategoryTheory.Cdg.shiftComp_eq {A : Type u} [Category.{v, u} A] [Preadditive A] {K L M : Cdg A} (m n a b c : ℤ) (h : a + b = c) (f : ↑((DGCategoryStruct.dgHom K L).X a)) (g : ↑((DGCategoryStruct.dgHom L M).X b)) :

shiftComp on C^dg is Cochain.comp with the Koszul sign that DGCategory.shiftComp's derivation forced.

The cross-check. Cochain.shift shifts the objects of a cochain and δ_shift says that costs (-1) ^ m. shiftD shifts the hom-complex and costs (-1) ^ m by CochainComplex.shiftFunctor. The two constructions are independent; the signs agree.