Whiskering homogeneous dg natural transformations #
A homogeneous dg natural transformation can be precomposed with a dg functor on the source side and postcomposed with one on the target side. Both operations keep the degree, and neither introduces a sign.
Where the Koszul sign is, and where it is not #
It is tempting to expect a sign in whiskerRight, because it applies a
functor to a morphism of degree n. There is none: a DGFunctor preserves
the graded composition on the nose (DGFunctor.map_comp carries no sign), so
transporting η's graded naturality through K transports the sign
unchanged. Likewise whiskerLeft is naturality of η evaluated at
F.map p f, again with η's own sign and no other.
The sign appears one level up, in the interchange law. For
α : F ⟶ F' of degree n and β : G ⟶ G' of degree m the two ways of
reading the horizontal composite differ by (-1)^(m * n):
(α ◁ G) ≫ (F' ▷ β) = (-1)^(m n) • ((F ▷ β) ≫ (α ◁ G')).
That is exactly graded naturality of β evaluated at the degree-n morphism
α X, which is why the sign is β's degree times α's and not something
new. The two horizontal composites agree on the nose exactly when m * n is
even.
Degrees #
Whiskering does not change the degree, so both operations are additive in the
transformation and are packaged as AddMonoidHoms. Vertical composition is
where degrees add, and whiskerLeft_composition and
whiskerRight_composition say whiskering is a homomorphism for it in every
result degree.
Left whiskering #
Precomposition with a dg functor.
The component at X is η's component at F.obj X, and graded naturality is
η's own, evaluated at F.map p f.
Equations
- CategoryTheory.DGFunctor.HomogeneousNatTrans.whiskerLeft F η = ⟨fun (X : C) => CategoryTheory.DGFunctor.HomogeneousNatTrans.app η (F.obj X), ⋯⟩
Instances For
Left whiskering as an additive map.
Equations
Instances For
Left whiskering takes the identity transformation to the identity.
Left whiskering is a homomorphism for vertical composition, in every result degree.
Left whiskering commutes with the pointwise differential.
A whiskered closed transformation is closed.
Right whiskering #
Postcomposition with a dg functor.
The component at X is K applied to η's component, in the same degree
n. No sign appears: K preserves the graded composition on the nose, so it
carries η's naturality equation across sign and all.
Equations
- CategoryTheory.DGFunctor.HomogeneousNatTrans.whiskerRight η K = ⟨fun (X : C) => (K.map n) (CategoryTheory.DGFunctor.HomogeneousNatTrans.app η X), ⋯⟩
Instances For
Right whiskering as an additive map.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Right whiskering takes the identity transformation to the identity. This
is DGFunctor.map_id, and it is the one place where right whiskering needs
more than additivity of K.map.
Right whiskering is a homomorphism for vertical composition, in every
result degree. This is DGFunctor.map_comp.
Right whiskering commutes with the pointwise differential. This is
DGFunctor.map_d.
A whiskered closed transformation is closed.
Whiskering by a composite #
All three laws hold by rfl, because DGFunctor.comp composes the object maps
and the AddMonoidHoms directly. They are what makes horizontal composition
of dg natural transformations strictly associative.
Whiskering twice on the right is whiskering by the composite.
Whiskering twice on the left is whiskering by the composite.
Whiskering on the two sides commutes.
The interchange law #
The Godement interchange law, componentwise.
Reading the horizontal composite of α and β in the two possible orders
differs by (-1)^(m * n). The identity is nothing but graded naturality of
β evaluated at the degree-n morphism α X, which is why the sign is
β's degree times α's and not something new.
The Godement interchange law.
(α ◁ G) ≫ (F' ▷ β) = (-1)^(m n) • ((F ▷ β) ≫ (α ◁ G')), with n the degree
of α and m the degree of β. Both sides are transformations
F ⋙ G ⟶ F' ⋙ G' of degree r; the two vertical composites are taken in the
two orders, which is why r comes with both n + m = r and m + n = r.