nLab copower

Redirected from "tensorings".
Contents

Context

Enriched category theory

Limits and colimits

Contents

Idea

In a closed monoidal category CC the tensor product a⊗ba \otimes b and internal hom [b,c][b,c] are related by the defining natural isomorphism

C(a⊗b,c)≃C(a,[b,c]). C(a \otimes b, c) \simeq C(a, [b,c]) \,.

The notion of copowering generalizes this to the situation where a category CC does not act on itself by tensors, but where another category VV acts on CC.

The dual notion is that of powering.

Definition

Definition

Let VV be a closed monoidal category serving as the cosmos for enrichment.

In a VV-enriched category C\mathbf{C}, the copower of an object c∈Cc \in \mathbf{C} by an object v∈Vv \in V is an object v⋅c∈Cv \cdot c \in \mathbf{C} with a ( V V -enriched-)natural isomorphism

C(v⋅c,c′)≅V(v,C(c,c′)) \mathbf{C}\big( v \cdot c,\, c' \big) \;\;\cong\;\; \mathbf{V}\big( v ,\, \mathbf{C}(c,c') \big)

where

  • C(−,−)\mathbf{C}(-,-) denotes the VV-valued hom-object of CC,

  • V(−,−)\mathbf{V}(-,-) denotes correspondingly the internal hom of VV (i.e. the hom-object with respect to its canonical self-enrichment).

Remark

(terminology)
Copowers are frequently called tensors and a VV-category having all copowers is called tensored, while the word “copower” is reserved for the case V=SetV=Set. However, there seems to be no good reason for making this distinction. Moreover, the word “tensor” is fairly overused, and unfortunate since a tensor (= a copower) is a colimit, while a cotensor (= power) is a limit.

Copowers as colimits

Conical colimits

For a Set-enriched category CC, a copower v⋅cv \cdot c is the colimit of the constant functor v→Cv \to C on cc (where vv is viewed as a discrete category).

Weighted colimits

A copower v⋅cv \cdot c is the colimit of the functor 1→C1 \to C picking out cc weighted by the presheaf 1 op→V1^{op} \to V picking out vv.

Properties

  • As described above, copowers are a kind of weighted colimit. Conversely, all weighted colimits can be constructed from copowers together with conical colimits (i.e., ordinary SetSet-based colimits with an enhanced VV-universal property, although the latter is automatic if powers also exist), assuming these exist. The dual limit notion of a copower is a power.

Examples

Example

Let VV be a Bénabou cosmos.

(e.g.Kelly 1982, Sec. 3.7)

Example

(copowering of small categories over set)
Every locally small category CC with all coproducts is canonically copowered over Set: the copowering functor

⋅:Set×C→C \cdot \,\colon\, Set \times C \to C

sends (S,b)(S,b) to the coproduct of |S||S|-many copies of b∈Cb \in C:

S⋅b≔∐ s∈Sb. S \cdot b \coloneqq \coprod_{s \in S} b \,.

The defining natural isomorphism in this situation is then just the fact that the hom functor sends colimits in its first argument to limits:

C(∐ s∈Sb,c)≃∏ s∈SC(b,c)≃Set(S,C(b,c)). C(\coprod_{s \in S} b , c) \simeq \prod_{s \in S} C(b,c) \simeq Set(S, C(b,c)) \,.

Example

(copowering of the category of monoids)
A particularly illuminating instance of Example occurs when CC is the category of monoids (or that of groups). In this case, the copower X⋅AX\cdot A of a monoid AA by a set XX is the free product of XX copies of AA, which can more concretely be described as a “one-sided version” of the tensor product of commutative monoids. Indeed, X⋅AX\cdot A is the monoid consisting of

  • The set given by the quotient of the free (noncommutative) monoid Free(X×A)Free(X\times A) on X×AX\times A by the congruence relation ∼\sim generated by the relations
    (x,a)(x,b) ∼(x,ab), (x,1 A) ∼() \begin{aligned} (x,a)(x,b) &\sim (x,a b),\\ (x,1_A) &\sim () \end{aligned}

    for each x∈Xx\in X and each a,b∈Aa,b\in A, where ()() is the unit of Free(X×A)Free(X\times A). Here, for each x∈Xx\in X and each a∈Aa\in A, we write x⋅ax\cdot a for the equivalence class of (x,a)(x,a).

  • The product given by concatenation, i.e. by
    (x⋅a)⋅ X⋅A(y⋅b)=(x⋅a)(y⋅b), (x\cdot a)\cdot_{X\cdot A}(y\cdot b) = (x\cdot a)(y\cdot b),

    for each x⋅a,y⋅b∈X⋅Ax\cdot a,y\cdot b\in X\cdot A;

  • The unit given by
    1 X⋅A=x⋅1 A,1_{X\cdot A}=x\cdot 1_A,

    which is independent of xx, as x⋅1 A=()=y⋅1 Ax\cdot 1_A=()=y\cdot 1_A for all x,y∈Ax,y\in A.

Explicitly, ∐ x∈XA\coprod_{x\in X}A is isomorphic to the above monoid via the isomorphism sending (x 1⋅a 1)⋯(x n⋅a n)(x_{1}\cdot a_{1})\cdots(x_{n}\cdot a_{n}) to the element of ∐ x∈XA\coprod_{x\in X}A given by the word a 1 (x 1)⋯a n (x n)a^{(x_{1})}_{1}\cdots a^{(x_{n})}_{n}, where a i (x i)a^{(x_{i})}_{i} is the element a ia_{i} in the x ix_{i}-th copy of AA in the expression ∐ x∈XA\coprod_{x\in X}A.

The universal property of the copower X⋅AX\cdot A states that a morphism of monoids from X⋅AX\cdot A to a monoid BB is the same data as a “left-bilinear” map of sets f:A×X→Bf\colon A\times X\to B, satisfying

f(ab,x) =f(a,x)f(b,x), f(1 A,x) =1 B \begin{aligned} f(a b,x) &= f(a,x)f(b,x),\\ f(1_A,x) &= 1_B \end{aligned}

for each x∈Xx\in X.

The copower ⋅:Set×Mon→Mon\cdot\colon Set\times Mon\to Mon also endows MonMon with skew monoidal structures ◃\triangleleft and ▹\triangleright, given by

A◃B =|B|⋅A, A▹B =|A|⋅B \begin{aligned} A\triangleleft B &= |B| \cdot A,\\ A\triangleright B &= |A| \cdot B \end{aligned}

for each A,B∈Obj(Mon)A,B\in Obj(Mon), where |A||A| and |B||B| are the underlying sets of AA and BB. While monoids in CMonCMon with respect to the tensor product of commutative monoids are semirings, monoids in MonMon with respect to ◃\triangleleft and ▹\triangleright recover left and right near-semirings.

References

Textbook accounts:

A discussion on the advantages of the terminology “copower” over the older terminology “tensor”:

Last revised on May 29, 2026 at 18:21:00. See the history of this page for a list of all contributions to it.