Blog Archive

Monday, June 29, 2026

The Architecture Behind the Hom Differential II: Tensor-Hom and DG Enrichment

 

The Architecture Behind the Hom Differential II: Tensor-Hom and DG Enrichment

In the first part, we saw that tensor products and Hom complexes arise from the same general mechanism:

bifunctorssigned double complexestotal complexes.

The tensor product of chain complexes is direct-sum totalization. The Hom complex is product totalization. The Hom complex also contains chain maps and chain homotopies:

Z0Hom(C,D)=chain maps,
H0Hom(C,D)=chain maps modulo chain homotopy.

Now we explain the next structural point.

The tensor product and Hom complex fit into the tensor-Hom adjunction in Ch(R). Then the Hom complex upgrades the category of chain complexes into a dg category.


The Tensor-Hom Adjunction in Chain Complexes

Let R be a commutative ring.

For chain complexes of R-modules, the tensor product is

XRY=Tot(XpRYq).

Thus

(XRY)n=p+q=nXpRYq.

The differential is

d(xy)=dXxy+(1)|x|xdYy.

The Hom complex is

HomR(Y,Z)n=jHomR(Yj,Zj+n),

with differential

(df)j=dZfj(1)|f|fj1dY.

These two constructions satisfy the closed monoidal adjunction

HomR(XRY,Z)HomR(X,HomR(Y,Z)).

This is an isomorphism of chain complexes.

On degree n elements, the isomorphism is ordinary currying.

A degree n element on the left is a family of maps

ϕ=(ϕa,b:XaRYbZa+b+n)a,b.

It corresponds to a degree n element on the right, namely a family

ϕ~=(ϕ~a:XaHomR(Y,Z)a+n)a.

For

xXa,

the element ϕ~a(x) is a degree a+n map from Y to Z. Its b-th component is

YbZb+a+n,

given by

yϕa,b(xy).

Thus

ϕ~a(x)(y)=ϕa,b(xy).

The reason this is an isomorphism of chain complexes is that the tensor differential and the Hom differential are compatible.

The evaluation map

ev:HomR(Y,Z)RYZ

is given without extra signs:

ev(fy)=f(y).

Let f be homogeneous of degree m. The tensor differential gives

d(fy)=dfy+(1)mfdYy.

Applying evaluation gives

(df)(y)+(1)mf(dYy).

Using the Hom differential

df=dZf(1)mfdY,

this becomes

dZ(f(y))(1)mf(dYy)+(1)mf(dYy).

The middle terms cancel, leaving

dZ(f(y)).

Thus evaluation is a chain map.

This is the concrete reason the tensor-Hom adjunction lives in Ch(R), not merely in graded R-modules.

So Ch(R) is a closed monoidal category:

RYHomR(Y,).

Why Enrichment Appears

The Hom complex does more than package chain maps and chain homotopies. It upgrades the whole category of chain complexes into an enriched category.

The guiding idea is simple.

Ordinary categories have Hom sets:

Hom(C,D).

A dg category has Hom complexes:

Map(C,D)Ch(K).

So in a dg category, morphisms are not merely elements of a set. They live inside chain complexes. Composition must therefore also live inside the world of chain complexes.

This is the point that needs to be made precise.


What a Ch(K)-Enriched Category Is

Recall that

Ch(K)

is a monoidal category under the tensor product of chain complexes, with unit object

K[0].

A category enriched over Ch(K) consists of the following data.

First, a class of objects.

Second, for every pair of objects A,B, a chain complex

Map(A,B)Ch(K).

Third, for every triple of objects A,B,C, a composition morphism in Ch(K):

:Map(B,C)Map(A,B)Map(A,C).

This is not just a bilinear map of graded modules. It must be a chain map.

Fourth, for every object A, a unit morphism in Ch(K):

K[0]Map(A,A).

These composition and unit maps must satisfy associativity and unitality, now as identities of morphisms in Ch(K).

A dg category is precisely a category enriched over Ch(K).

So to prove that

Ch(C)

is a dg category, we must define:

Map(C,D),

define composition as a chain map,

Map(D,E)Map(C,D)Map(C,E),

define the unit map,

K[0]Map(C,C),

and check associativity and unit laws.


The Hom Complex as the Enriched Hom Object

Let C be a K-linear additive category.

For chain complexes

C,DCh(C),

define

Map(C,D)=Hom(C,D).

Explicitly,

Map(C,D)n=iHomC(Ci,Di+n).

A homogeneous element

fMap(C,D)n

is a family of morphisms

f=(fi:CiDi+n)i.

The differential is

(df)i=dDfi(1)nfi1dC.

Thus the enriched Hom object from C to D is exactly the Hom complex.

Notice the first important consequence:

Z0Map(C,D)

is the set of ordinary chain maps CD.

Indeed, a degree-zero element is a family

f=(fi:CiDi)i,

and the condition df=0 says

dDfi=fi1dC,

which is precisely the chain map condition.

So the ordinary Hom set in Ch(C) is recovered as

HomCh(C)(C,D)=Z0Map(C,D).

This explains why degree-zero cycles, rather than all degree-zero elements, give the ordinary morphisms.


Defining Composition

Now take three chain complexes

C,D,ECh(C).

We want to define a composition morphism in Ch(K):

:Map(D,E)Map(C,D)Map(C,E).

Let

f=(fi:CiDi+n)iMap(C,D)n,

and

g=(gj:DjEj+m)jMap(D,E)m.

Thus f is a degree n family of morphisms, and g is a degree m family of morphisms.

To compose them, start with Ci:

CifiDi+n.

Now the relevant component of g is the one whose source is Di+n, namely

gi+n:Di+nEi+n+m.

Therefore define

(gf)i=gi+nfi.

Thus

gf=(gi+nfi:CiEi+n+m)i.

So

gfMap(C,E)m+n.

Composition adds degrees:

|gf|=|g|+|f|.

This defines a graded bilinear map

Map(D,E)mMap(C,D)nMap(C,E)m+n.

Equivalently, it defines a morphism of graded K-modules

Map(D,E)Map(C,D)Map(C,E).

But for enrichment over Ch(K), this is not enough. We must prove that this map is a chain map.


Why the Leibniz Rule Means That Composition Is a Chain Map

Let

A=Map(D,E),B=Map(C,D),M=Map(C,E).

The enriched composition is supposed to be a morphism in Ch(K):

μ:ABM.

Here

μ(gf)=gf.

So the source complex is

Map(D,E)Map(C,D),

and the target complex is

Map(C,E).

A morphism in Ch(K) is a chain map. Therefore μ must satisfy

dMμ=μdAB.

Equivalently, for every homogeneous simple tensor

gfAmBn,

we need

dM(μ(gf))=μ(dAB(gf)).

The left-hand side is

dM(μ(gf))=dM(gf)=d(gf).

Now compute the right-hand side.

The tensor product differential on

AB=Map(D,E)Map(C,D)

is

dAB(gf)=dgf+(1)mgdf,

where

m=|g|.

Applying μ gives

μ(dAB(gf))=μ(dgf)+(1)mμ(gdf).

By definition of μ,

μ(dgf)=dgf,

and

μ(gdf)=gdf.

Thus

μ(dAB(gf))=dgf+(1)mgdf.

Therefore the chain map condition

dMμ=μdAB

is exactly the formula

d(gf)=dgf+(1)mgdf.

This is the graded Leibniz rule for composition.

So the role of the Leibniz rule is not mysterious. It is simply the chain map condition for the composition map

Map(D,E)Map(C,D)Map(C,E)

after expanding the tensor product differential on the source.


Verifying the Leibniz Rule

Now we verify

d(gf)=dgf+(1)mgdf.

Let

f=(fi:CiDi+n)iMap(C,D)n,

and

g=(gj:DjEj+m)jMap(D,E)m.

Then

gf=(gi+nfi:CiEi+n+m)i.

Since gf has degree m+n, its differential is

d(gf)i=dE(gi+nfi)(1)m+n(gf)i1dC.

Now

(gf)i1=gi+n1fi1.

Thus

d(gf)i=dEgi+nfi(1)m+ngi+n1fi1dC.

On the other hand,

(dgf)i=(dg)i+nfi.

Since g has degree m,

(dg)i+n=dEgi+n(1)mgi+n1dD.

Hence

(dgf)i=dEgi+nfi(1)mgi+n1dDfi.

Also,

(gdf)i=gi+n1(df)i.

Since f has degree n,

(df)i=dDfi(1)nfi1dC.

Therefore

(gdf)i=gi+n1dDfi(1)ngi+n1fi1dC.

Multiplying by (1)m, we get

(1)m(gdf)i=(1)mgi+n1dDfi(1)m+ngi+n1fi1dC.

Adding the two expressions gives

(dgf)i+(1)m(gdf)i
=dEgi+nfi(1)mgi+n1dDfi+(1)mgi+n1dDfi(1)m+ngi+n1fi1dC.

The middle two terms cancel. Hence

(dgf)i+(1)m(gdf)i=dEgi+nfi(1)m+ngi+n1fi1dC.

This is exactly

d(gf)i.

Therefore

d(gf)=dgf+(1)mgdf.

So composition is a chain map in Ch(K):

Map(D,E)Map(C,D)Map(C,E).

This is the enriched composition.


The Unit Map

For each chain complex C, the identity family

idC=(idCi)i

is a degree-zero element of

Map(C,C).

It is a cycle because

(didC)i=dCidCiidCi1dC=dCdC=0.

Therefore

idCZ0Map(C,C).

Equivalently, it defines a chain map

K[0]Map(C,C),

sending

1idC.

This is the enriched identity.


Associativity and Unitality

Associativity follows from ordinary associativity of composition in C.

Indeed, if

fMap(C,D),gMap(D,E),hMap(E,F),

then componentwise,

(h(gf))i=hi+|f|+|g|(gi+|f|fi),

while

((hg)f)i=(hi+|f|+|g|gi+|f|)fi.

These are equal by associativity in C.

Unitality follows similarly. For any homogeneous

fMap(C,D),

we have

(idDf)i=idDi+|f|fi=fi,

and

(fidC)i=fiidCi=fi.

Thus the enriched composition is associative and unital.

Since the composition and unit maps are chain maps, these identities hold inside Ch(K).

Therefore

Ch(C)

is enriched over

Ch(K).

Equivalently,

Ch(C)

is a dg category.


Taking Z0 Recovers Ordinary Composition

The ordinary category of chain complexes is obtained by taking degree-zero cycles of the enriched Hom complexes.

We have already seen that

Z0Map(C,D)=HomCh(C)(C,D).

Now suppose

fZ0Map(C,D),gZ0Map(D,E).

Then

|f|=|g|=0,df=dg=0.

The enriched composition gives

(gf)i=gifi.

This is exactly the ordinary componentwise composition of chain maps.

Moreover, since composition is a chain map,

d(gf)=dgf+gdf=0.

So

gfZ0Map(C,E).

Thus applying Z0 to the enriched composition recovers the usual composition law

HomCh(C)(D,E)×HomCh(C)(C,D)HomCh(C)(C,E).

In this sense, ordinary composition is the degree-zero shadow of dg-enriched composition.


The Homotopy Category as H0

Taking degree-zero homology gives the homotopy category.

A degree-one element

hMap(C,D)1

is a family

h=(hi:CiDi+1)i.

Its boundary is

(dh)i=dDhi+hi1dC.

This is precisely the formula for a null-homotopic chain map.

Therefore

H0Map(C,D)

is the module of chain maps modulo chain homotopy.

Hence

HomK(C)(C,D)=H0Map(C,D).

So there are three layers:

Map(C,D)0=graded maps,
Z0Map(C,D)=chain maps,
H0Map(C,D)=chain maps modulo chain homotopy.

The dg-enriched category remembers the whole Hom complex. The ordinary category remembers only Z0. The homotopy category remembers only H0.


Endomorphisms Form a DG Algebra

In any enriched category, the endomorphism object of an object is automatically a monoid object.

Apply this to the Ch(K)-enriched category

Ch(C).

For a chain complex C, define

End(C)=Map(C,C)=Hom(C,C).

Composition gives multiplication

End(C)End(C)End(C),

and the identity gives the unit

K[0]End(C).

Therefore

End(C)

is a monoid object in Ch(K).

But monoid objects in Ch(K) are precisely differential graded K-algebras.

Concretely, if

a,bEnd(C)

are homogeneous, then multiplication is composition:

ab=ab.

The differential satisfies

d(ab)=d(a)b+(1)|a|ad(b).

This is exactly the graded Leibniz rule for enriched composition.

Thus the dg algebra of endomorphisms is not an extra structure placed on top of a chain complex. It is forced by enrichment:

End(C)

is the endomorphism monoid object of C inside the dg category Ch(C).


Summary

The tensor product and Hom complex are not isolated constructions.

The tensor product comes from direct-sum totalization:

XRY=Tot(XpRYq).

The Hom complex comes from product totalization:

Hom(C,D)=TotΠ(HomC(Cp,Dq)).

Together they give the closed monoidal structure on Ch(R):

RYHomR(Y,).

The Hom complex also gives the enriched Hom objects for the dg category

Ch(C).

Composition is not merely a graded bilinear operation. It is a chain map:

Map(D,E)Map(C,D)Map(C,E).

Taking Z0 recovers ordinary composition of chain maps.

Taking H0 gives composition in the homotopy category.

Finally, the endomorphism object

End(C)

is automatically a monoid object in Ch(K), hence a dg algebra.

Thus the elementary Hom differential

(df)i=dDfi(1)nfi1dC

is the visible trace of a larger structure:

bifunctorssigned double complexestotal complexesdg enrichment.

No comments:

Post a Comment

Popular Posts