The opposite dg category #
The sign convention, and why this one #
The opposite of a dg category reverses composition and inserts a Koszul sign:
f ∘ᵒᵖ g = (-1) ^ (|f| * |g|) • (g ∘ f).
Without the sign, associativity still holds but the Leibniz rule does not, so the sign is forced rather than chosen. Both checks are recorded here because a sign convention that is only implicit is the usual way two dg developments come to disagree without either noticing.
Associativity. With |f| = p, |g| = q, |h| = r:
(f ∘ᵒᵖ g) ∘ᵒᵖ hcarries(-1) ^ ((p + q) * r + p * q)f ∘ᵒᵖ (g ∘ᵒᵖ h)carries(-1) ^ (p * (q + r) + q * r)
and (p + q) * r + p * q = p * (q + r) + q * r.
Leibniz. Under the convention of DerivedAlgGeo.Algebra.Homology.DGCategory/Basic.lean —
Mathlib's,
δ (f · g) = f · (δg) + (-1) ^ q • (δf) · g — the two sides carry
(-1) ^ (p * q) on g · (δf) and (-1) ^ (p * q + p) on (δg) · f, on both
sides, once p * (q + 1) = p * q + p and q + (p + 1) * q = p * q + 2 * q are
used and (-1) ^ (2 * q) = 1 is discharged. The Koszul sign is unchanged from
the earlier draft, which used the mirror convention: only the term-matching
moved.
The units live in ℤˣ via Int.negOnePow, matching Mathlib's convention in
CochainComplex.HomComplex rather than introducing a second one.
The opposite of a dg category: the same objects, Hom-complexes reversed, composition reversed with a Koszul sign.
Equations
- One or more equations did not get rendered due to their size.
dgComp is ℤˣ-equivariant in its first argument, because it is
additive and the ℤˣ action factors through ℤ.
dgComp is ℤˣ-equivariant in its second argument.
A morphism of AddCommGrpCat commutes with the ℤˣ action, which factors
through ℤ. Stated so the sign bookkeeping below can stay in ℤˣ instead of
dropping to ℤ and back.
The opposite dg category.
Equations
- CategoryTheory.DGCategory.op = { toDGCategoryStruct := CategoryTheory.DGCategory.opStruct C, dgComp_assoc := ⋯, dgId_comp := ⋯, dgComp_id := ⋯, dgId_cocycle := ⋯, dgComp_leibniz := ⋯ }