5. The semigroup of cones

The following definition is basic in what follows in this section. Here, unless otherwise provided, we shall assume that 𝒞, 𝒟, etc., are small 𝒮-categories.

Definition 5.1 (Cones). Let d 𝖛𝒞. A map γ : 𝖛𝒞𝒞 is called a cone from the base 𝖛𝒞 to the vertex d (or simply a cone in 𝒞 to d) if γ satisfies the following:

Con-1
γ(c) 𝒞(c,d) for all c 𝖛𝒞.
Con-2
If c c then ȷccγ(c) = γ(c).

Given the cone γ we denote by cγ the the vertex of γ and for each c 𝖛𝒞, the morphism γ(c) : c cγ is called the component of γ at c. The cone γ is said to be proper [normal] if γ satisfies the condition Con-3 [respectively Con-4] below:

Con-3
c𝖛𝒞im(γ(c)) = cγ.
Con-4
There exists c 𝖛𝒞 such that γ(c) : c cγ is an isomorphism.

Since U is an (order) embedding of 𝖛𝒞 into Set, it is clear that the statement Con-3 is equivalent to

c𝖛𝒞U(imγ(c)) = U(cγ). (16)

It is useful to observe that a sufficient condition for the cone γ to be proper is that there exists c 𝖛𝒞 such that γ(c) : c cγ is surjective (that is, γ(c) = γ(c)). It follows that any normal cone is proper.

Let con𝒞 [respectively pcn𝒞, and ncn𝒞] denote the set of all cones [respectively proper and normal] cones in 𝒞. Then

ncn𝒞pcn𝒞con𝒞.

When γ con𝒞 is normal we write

M(γ) = {c 𝖛𝒞 : γ(c) is an isomorphism}. (17)

Therefore the cone γ is normal if and only if M(γ) .

Recall that the preorder 𝖛𝒞 is a subcategory of 𝒞 so that the inclusion J = ȷ𝒞𝖛𝒞 is a functor of 𝖛𝒞 to 𝒞. Any cone in 𝒞 in the sense of the definition above, is a natural transformation from the base J to the constant functor of 𝖛𝒞 at vertex cγ; that is, a cone in the sense of classical category theory (see for example, MacLane [1971]).

Given a cone γ in 𝒞 and a morphism f 𝒞(cγ,d) the map defined, for all c 𝖛𝒞, by:

γ f(c) = γ(c)f (18)

is clearly a cone with vertex d. Furthermore, if f,g 𝒞 is a pair of composable morphisms with domf = cγ,we have:

γ (fg) = γ f (gimf). (19)

The construction given in Equation (18) is especially useful when γ is proper.

Proposition 5.1. Let γ be a proper cone in 𝒞 and f 𝒞(cγ,d). Then the cone γ f is prooper if and only if f is surjective.

Proof. Suppose that η = γ f where f is surjective. Since γ is proper, Con-3 or equivalently, Equation (16) hold. Thus

U(cγ) = c𝖛𝒞im(U(γ(c)))

Since by Equation (18), γ f is the map sending c 𝖛𝒞 to γ(c)f, we have

U(γ f)(c) = U(γ(c))U(f) = U(γ(c)) U(f).

Let y U(d). Since the map U(f) : U(cγ) U(d) is surjective there is x U(cγ) such that xU(f) = y. Since the cone γ is proper, by Cone-3, there is c 𝖛𝒞 such that x imU(γ(c)). This implies that x = uU(γ(c)) for some u U(c). Then y = uU(γ(c))U(f). Therefore y Uim γ(c)f.

 It follows that        U(d) c𝖛𝒞imU γ(c)f.  The inclusion        c𝖛𝒞imU γ(c)f U(d) is obvious.  Hence        U(d) = c𝖛𝒞imU γ(c)f.

Therefore by Cone-3, γ f is proper.

Suppose that the cone γ f is proper. By Cone-3 we have

U(d) = c𝖛𝒞U im(γ(c)f).

Hence, for y U(d) there is c 𝖛𝒞 such that y U im(γ(c)f). Then there exist x U(c) such that y = xU(γ(c)f) = uU(f) where u = xU(γ(c)). Therefore U(f) : U(cγ) U(d) is surjective. The fact that 𝒞 is an 𝒮-category now shows that f : cγ d is surjective.

Recall that pcn𝒞 denote the set of all proper cones in 𝒞. For γ,η pcn𝒞, let

γη = γ η(cγ). (20)

By Proposition 5.1 γη is a unique proper cone with vertex imη(cγ) cη. Therefore this defines a binary operation in pcn𝒞.

Proposition 5.2. Let 𝒞 be a small 𝒮-category. Then pcn𝒞 is a semigroup with respect to the binary composition defined by Equation (20).

Proof. It is sufficient to show that the binary composition defined by Equation (20) is associative. Let α,β,γ pcn𝒞 and c 𝖛𝒞. Then

α(βγ)(c) = α(c) (βγ)(cα) = α(c) (β(cα)) (γ(cβ)) = α(c)(β(cα)(γ(cβ) = α(c)(β(cα))(γ(c αβ))  by (6), sinceȷcβ im β(cα)γ(cβ) = γ(imβ(cα)) = γ(cαβ).  Similarly,     (αβ)γ(c) = (αβ)(c)(γ(cαβ)) = α(c)(β(cα))(γ(c αβ))  Therefore     α(βγ)(c) = (αβ)γ(c)  for all c 𝖛𝒞.

Thus α(βγ) = (αβ)γ. If γ,η pcn𝒞 then by Proposition 5.1 γη pcn𝒞. Thus pcn𝒞 is a semigroup.

Let 𝒞 be an 𝒮-categorywith forgetful functor U. Since 𝖛𝒞 is a subcategory of the small category 𝒞 and Set is (small) complete, the functor 𝖛U : 𝖛𝒞Set has the direct limit [see MacLane1971, page 105]. The direct-limit Lim(𝖛U) of 𝖛U consists of a set, which is denoted by Δ, and a cone (the universal or limiting cone) from the base 𝖛U to the vertex Δ [see Nambooripad1994, page 11 for details]. If U is the inclusion functor 𝒞𝒟, then we write Lim𝒞 for the limit Lim(𝒞𝒟) if the category 𝒟 in which the limit is evaluvated is clear from the context. In particular, unless otherwise stated explicitly, Lim𝒞 will stand for the direct limit of the inclusion functor 𝒞Set (evaluvated in Set). We use these notations in the following lemma which provides a convenient representation for Lim(𝖛U).

Proposition 5.3. Suppose that (𝒞,U) be an 𝒮-category. Define

Δ = c𝖛𝒞U(c), (21a)

and let ȷ be the function defined, for all c 𝖛𝒞, by

ȷ(c) = ȷΔ U(c) (21b)

Then ȷ is a proper cone from the base 𝖛U to the vertex Δ and

Lim(𝖛U) = (Δ,ȷ). (21c)

If η pcn𝒞, then there is a unique map η̃ : Δ U(cη) such that

η(c) = ȷΔ U(c) η̃ for allc 𝖛𝒞. (22)

Furthermore, the cone η is proper if and only if η̃ is surjective.

Proof. Since 𝖛𝒞Set is an order embedding we see that the map ȷ : cȷΔc is a cone from the base 𝖛𝒞 to the vertex Δ. We prove that ȷ is universal. Suppose that α : cα(c) is a cone from the base 𝖛𝒞 to the vertex X. For each x Δ define

xf = xα(c) ifx c.

Since, by Equation (21a), Δ is the union of sets c 𝖛𝒞, xf is defined for all x Δ. If x U(ci), i = 1,2, then by the axiom (S:c) of Definition 3.1 there is d 𝖛𝒞 such that d ci, i = 1,2 and x d. Since, as noted above, 𝖛𝒞Set is inclusion preserving and α is a cone on 𝖛𝒞 we have

xα(c1) = xα(d) = xα(c2).

This shows that f : Δ X is a map. The definition of f shows that, for all c 𝖛𝒞,

fc = ȷΔ c f = ȷ(c) f = α(c).

Thus the following diagram commutes

a = b (23)

for all c 𝖛𝒞. To prove the uniqueness of f, suppose that g : Δ X is another map such that

ȷ(c) g = α(c) for allc 𝖛𝒞.  Then we have gc = α(c) = fc

for all c 𝖛𝒞 which implies that f = g. Therefore the cone ȷ is universal from the base 𝖛𝒞 to the vertex Δ.

Let η pcn𝒞. Since ȷ is universal from 𝖛U to Δ there is a unique map η̃ : Δ U(cη) making the digram Definition 24 commute or equivalently, Equation (22) holds.

a = b (24)

Suppose that S is a smigroup and let 𝒞 be a subcategory of ModS with images. Let U : ModS Set be the natural forgetful functor. By Propositions 4.1 and 4.2, 𝒞 is an 𝒮-category with respect to the foregetful functor U𝒞 = U0 : 𝒞Set. Vertices of 𝒞 are S-modules and so Δ defined by Equation (21a) carries a unique module structure such that ȷΔc is a module morphism for each c 𝖛𝒞. Furthermore, as in Proposition 5.3, the map ȷ : cȷ(c) = ȷΔc is a cone from the base 𝖛𝒞 to the vertex Δ which is universal. It follows that (Δ,ȷ) is the direct limit Lim𝖛𝒞 evaluvated in ModS. If 𝒞 𝕃(S) then we can show similarly that (Δ,ȷ) is the direct limit Lim𝖛𝒞 evaluvated in 𝕃(S). Thus we have:

Proposition 5.4. Let S be a smigroup. Suppose that 𝒞ModS [or 𝒞 𝕃(S)] and that Δ and ȷ are defined by Equations (21a) and (21b). Then Equation (21c) holds in which the direct limit Lim(𝖛U) is evaluvated in ModS if 𝒞ModS [or in 𝕃(S) if 𝒞 𝕃(S)].

As above, let (𝒞,U) be a 𝒮-category with forgetful functor U. If 𝒞ModS [or 𝒞 𝕃(S)], η(c) and ȷ(c) = ȷΔc are module morphisms [translations of ideals] for every c 𝖛𝒞. The equation above shows that the map η̃ : Δ cη is a module morphism [translations of ideals]. Define ϕ = ϕ𝒞 on pcn𝒞 as follows:

ϕ(γ) = γ̃ ȷ(cγ) for allγ pcn𝒞 (25)

It is clear that, for each γ pcn𝒞, ϕ(γ) 𝒯Δ is a transformation of Δ. We have

Proposition 5.5. Let 𝒞 be a 𝒮-category. For each γ pcn𝒞, let ϕ(γ) be defined by Equation (25). Then

ϕ𝒞 = ϕ : pcn𝒞 𝒯Δ,γϕ(γ)

is a faithful representation of the semigroup pcn𝒞 by transformations on Δ.

Proof. We shall show that ϕ : pcn𝒞 𝒯Δ is a homomorphism. Let α,β pcn𝒞. By Equation (20),

(αβ)(c) = α(c)β(cα) forc 𝖛𝒞.

Hence, for all x Δ, we have

xϕ(αβ) = xαβ̃ȷ(cαβ) = x(αβ)(c)ȷ(cαβ) = xα(c)(β(cα))ȷ(c αβ)    by (20). Since 𝖛𝒞Set is an order embedding this gives xϕ(αβ) = xα(c)β(cα))ȷ(cβ) = xα̃β̃ȷ(cβ) = xϕ(α)ϕ(β).

by Equation (25). Therefore ϕ(αβ) = ϕ(α)ϕ(β).

Assume that α,β pcn𝒞 and ϕ(α) = ϕ(β).

For c 𝖛𝒞 let

pcn𝒞(c) = {γ pcn𝒞 : cγ c}.

The result above shows that each f 𝒞(c,d) induces a map

ρf : γγ (fcγ) (26)

of the set pcn𝒞(c) to the set pcn𝒞(imf). The partial transformation ρf of pcn𝒞 will be called the translation induced by f. Notice that ρf = ρf. Hence, in the following, we may restrict our consideration to translations induced by surjections.

Let γ pcn𝒞. For each c 𝖛𝒞 and for f 𝒞(c,d), define

H(γ;c) = {γ f : f 𝒞(c γ,c)},  (27a)  (γ g)H(γ;f) = γ (gf) for all g 𝒞(c γ,c).  (27b) 

Clearly, H(γ;c) is a set for all c 𝖛𝒞 and by the uniqueness of epimorphic component, H(γ;f) is a map of the set H(γ;c) to H(γ;d) for f 𝒞(c,d).

In the following we use some standard constructions and terminologies of category theory; see MacLane [1971, Sections III.2 and III.3] and Nambooripad [1994, Sections I.2 and I.3]. In particular, given two functors F,G : 𝒞𝒟 we write F G if and only if

F(c) G(c) for all c 𝖛𝒞 and the mapcȷG(c) F(c)

is a natural transformation.

Proposition 5.6. For each γ pcn𝒞, Equations (27a) and (27b) defines an inclusion preserving covariant functor H(γ;) : 𝒞Set. Moreover, there is a unique natural isomorphism

ηγ : H(γ;) 𝒞(cγ,) such thatηγ(cγ)(γ) = 1γ.

Consequently, H(γ;) is a representable functor and cγ represents it.

Proof. Let g 𝒞(c,d) and h 𝒞(d,u). Then by Equations (27a) and (27b),

(γ f)H(γ;gh) = γ (fgh) = (γ f)H(γ;g)H(γ;h) for all γ f H([;c)]γ.  Therefore   H(γ;gh) = H(γ;g) H(γ;h).

If c d then it follows readily from Equation (27b) that

H(γ;ȷd c) = ȷH(γ;d) H(γ;c).

In particular, H(γ;1c) = 1H(γ;c). In view of the remarks following Equation (27b), this shows that H(γ;) : 𝒞Set is an inclusion preserving covariant functor.‘

To prove the remaining statements, by Yoneda lemma see [MacLane1971, Section III.2] and Nambooripad [1994, section I.2], it is sufficient to show that γ is universal element for H(γ;) in H(γ;cγ). Thus we must show that given any γ f H(γ;c), there is a unique h : cγ c such that

(γ)H(γ;h) = γ f.

By Equation (27b), the above equation holds for the choice h = f. To see that this choice is unique, assum that k : cγ c also satisfies this equation. Then, again by Equation (27b), we have γ f = γ k. Since γ is proper, D. Choose c D. By Equation (18), we have

γ(c)f = γ(c)k which impliesf = k

since γ(c) is surjective (see Equation (17)). Hence imf = img and so, f = fȷcim f = kȷcim k = k. This completes the proof.

The functor H(γ;:)𝒞Set is called the covariant H-functor etermined by the cone γ pcn𝒞.