Finite length bounds Hom-dimension from the unit #
In the setting of Deligne's theorem — an abelian ℂ-linear rigid monoidal category whose unit endomorphisms are exactly the scalars — the Hom-space out of the tensor unit is finite dimensional whenever the target has bounded length, with dimension at most the length bound. This is the Hom-finiteness input of Deligne's Proposition 2.1.
The route: a nonzero map φ : 𝟙 ⟶ Z is a monomorphism because the
unit is simple (simple_unit_of_hasScalarUnit), and pulling
subobjects of cokernel φ back along the projection embeds its
subobject chains strictly above the nonzero subobject φ, so the
length bound drops by one on the cokernel. Left-exactness of
Hom (𝟙, −), in the concrete form that a map annihilated by the
projection factors through φ with a scalar coefficient, bounds the
kernel of the induced linear map by one dimension, and induction
along the length bound does the rest.
At the bottom of the induction sits the simple case, recorded
separately: Hom (𝟙, S) vanishes for a simple S not isomorphic
to the unit, and is the line End (𝟙_ A) = ℂ when it is — the
scalar-unit hypothesis read as a ℂ-linear equivalence.
The unit endomorphisms are a line #
The scalar-unit hypothesis says that scaling the identity of the
tensor unit is a bijection ℂ → End (𝟙). Scaling is ℂ-linear, so
the bijection is a ℂ-linear equivalence and End (𝟙) is a line.
Transporting along an isomorphism 𝟙 ≅ S gives the same for
𝟙 ⟶ S.
Scaling the identity of the tensor unit, as a ℂ-linear map.
Equations
- RS.unitScalarMap = { toFun := fun (c : ℂ) => c • CategoryTheory.CategoryStruct.id (CategoryTheory.MonoidalCategoryStruct.tensorUnit A), map_add' := ⋯, map_smul' := ⋯ }
Instances For
The scalars exhaust the unit endomorphisms, as a ℂ-linear
equivalence ℂ ≃ₗ End (𝟙_ A): this is exactly the content of
HasScalarUnit, packaged linearly.
Equations
Instances For
The unit endomorphisms are one dimensional.
An isomorphism 𝟙 ≅ S identifies 𝟙 ⟶ S with the unit
endomorphisms, ℂ-linearly.
Equations
Instances For
Hom (𝟙, S) is a line when S is isomorphic to the unit:
it is then End (𝟙_ A) = ℂ.
Maps from the unit to a simple object are proportional #
Schur's lemma between the simple unit and a simple target: any
nonzero φ : 𝟙 ⟶ S is an isomorphism, so every ψ : 𝟙 ⟶ S is a
scalar multiple of it, the scalar produced by HasScalarUnit.
A nonzero map from the unit to a simple object is an isomorphism, the unit being simple; so a simple object receiving a nonzero map from the unit is the unit up to isomorphism.
Maps from the unit to a simple object not isomorphic to it vanish.
Hom (𝟙, S) is trivial for a simple S not isomorphic to the
unit.
Hom (𝟙, S) vanishes for a simple S not isomorphic to the
unit.
Quotients drop the length bound #
Pulling a subobject of a quotient back along the projection gives a
subobject of the ambient object. The operation is monotone, and
along an epimorphism it reflects inequalities, so it embeds strict
chains; every pullback contains the kernel of the projection, so for
the quotient by a nonzero subobject the embedded chain sits strictly
above ⊥ and the length bound drops by one.
The induction along the length bound #
Hom (𝟙, −) is left exact; concretely, a map 𝟙 ⟶ Z annihilated
by the projection to cokernel φ factors through the kernel φ of
that projection, with coefficient a unit endomorphism — a scalar.
So the kernel of postcomposition by the projection is at most one
dimensional, the quotient lemma above drops the length bound on the
cokernel, and rank-nullity closes the induction.
The theorems #
Finite length bounds Hom-dimension from the unit: in the
setting of Deligne's theorem, a length bound of N on Z makes
𝟙 ⟶ Z a finite dimensional ℂ-module of dimension at most N.
The duality correspondence (X ⟶ Y) ≃ₗ (𝟙 ⟶ Y ⊗ Xᘁ) as a
ℂ-linear equivalence: precomposition by the left unitor followed by
the right-dual adjunction tensorRightHomEquiv. Linearity follows
from bilinearity of composition and of the tensor product.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Finite length bounds every Hom-dimension: in the setting of
Deligne's theorem, a length bound of N on Y ⊗ Xᘁ makes X ⟶ Y
a finite dimensional ℂ-module of dimension at most N, via the
duality correspondence (X ⟶ Y) ≃ₗ (𝟙 ⟶ Y ⊗ Xᘁ).
Finite length, unquantified #
LengthLE Z N is the bound-shaped finite-length predicate of this
development; "Z has finite length" is ∃ N, LengthLE Z N, and
"every object has finite length" is that statement quantified over
all objects. There is no finite-length typeclass here, so the
hypothesis is carried explicitly.
Hom out of the unit is finite dimensional whenever the target has finite length.