betaFeedback

WorksMathematics

Free Gödel algebras over Gödel hoops

Abstract

We give an explicit description of the left adjoint FF of the forgetful functor U ⁣:GAlg→GHoopsU\colon\mathsf{GAlg}\to\mathsf{GHoops} from Gödel algebras to Gödel hoops. If HH is a Gödel hoop with a least element mm, then F(H)F(H) is the direct product of the Gödel algebra obtained by adjoining a new bottom element to HH and of HH itself, regarded as a Gödel algebra with bottom element mm. For an arbitrary Gödel hoop HH, we glue these products along the principal filters of HH, and we describe the operations of F(H)F(H), the action of FF on homomorphisms and the unit of the adjunction. As an application, we show that the unit of the adjunction F⊣UF\dashv U is not cartesian, and that the same holds for BL-algebras and basic hoops. Moreover, the semi-abelian categories of Gödel hoops and of basic hoops are not equivalent to the slice categories of Gödel algebras and of BL-algebras over their initial object. Thus, in contrast with the cases of Wajsberg hoops and of product hoops, the free-forgetful adjunctions between Gödel algebras and Gödel hoops, and between BL-algebras and basic hoops, do not give rise to ideally exact contexts.

Introduction

Among the many-valued logics studied in mathematical fuzzy logic, a central role is played by those whose truth values are the real numbers of the unit interval [0,1][0,1], and in which the conjunction is interpreted by a continuous tt-norm. Recall that a tt-norm is a binary operation ∗\ast on [0,1][0,1] which is associative, commutative, monotone in each argument with respect to the usual order, and has 11 as neutral element. If ∗\ast is continuous, then the implication is interpreted by its residuum

x→y=max⁡{z∈[0,1]∣z∗x≤y}.x \to y = \max\{z \in[0,1] \mid z \ast x \le y\}.

The three fundamental continuous tt-norms are the Łukasiewicz tt-norm x∗y=max⁡{x+y−1,0}x \ast y = \max\{x+y-1,0\}, the Gödel tt-norm x∗y=min⁡{x,y}x \ast y = \min\{x,y\} and the product tt-norm x∗y=xyx \ast y = xy. Every continuous tt-norm is an ordinal sum of isomorphic copies of the Łukasiewicz and product tt-norms, defined on pairwise disjoint open subintervals of [0,1][0,1], and it coincides with the Gödel tt-norm whenever its two arguments do not lie in the same subinterval [14]. The corresponding logics are Łukasiewicz infinite-valued logic, Gödel logic and product logic, whose equivalent algebraic semantics are given by the varieties of MV-algebras [9], Gödel algebras [14] and product algebras [15], respectively. In [14], P. Hájek introduced Basic Logic, a common fragment of these three logics, together with its algebraic semantics given by BL-algebras, and he conjectured that it is the logic of all continuous tt-norms and their residua. This conjecture was proved in [10], and a more algebraic proof was given in [2].

Hoops first appeared in the work of B. Bosbach on complementary semigroups [6], and their structure theory was developed by W. J. Blok and I. M. A. Ferreirim [5]. Basic hoops, Wajsberg hoops, Gödel hoops and product hoops are the 00-free subreducts of BL-algebras, MV-algebras, Gödel algebras and product algebras, respectively [2, 13], and the latter are term equivalent to the corresponding classes of bounded hoops, that is, of hoops with a least element named by a constant 00; 2020 Mathematics Subject Classification. 03G25; 06D20; 06F05; 08B20; 18A40; 18E13.

Key words and phrases. Gödel hoop, Gödel algebra, BL-algebra, basic hoop, free–forgetful adjunction, ideally exact category, semi-abelian category. see [2, 14]. In each case, the functor which forgets the constant 00 has a left adjoint, which freely adjoins a falsum constant. An explicit description of this left adjoint is known for Wajsberg hoops, where it is given by the MV-closure of [1], and for product hoops, where it was obtained in [13] by means of ℓ\ell-groups. On the other hand, it was observed in [13 Section 6] that it is not known how to freely obtain a Gödel algebra from a Gödel hoop, and that the methods developed there for product hoops cannot be directly applied to this case.

In this article, we give such a construction. Let U:GAlg→GHoopsU : \mathrm{GAlg} \to\mathrm{GHoops} be the forgetful functor from Gödel algebras to Gödel hoops, and let FF be its left adjoint. For a Gödel hoop HH, we denote by H+H^{+} the Gödel algebra obtained by adjoining a new bottom element to HH. We show in Lemma 3.1 that, if HH has a least element mm, then F(H)F(H) is the direct product of H+H^{+} and of HH itself, regarded as a Gödel algebra with bottom element mm. Adjoining a new bottom element is not enough, since a homomorphism of Gödel hoops from HH to a Gödel algebra may send mm to 00: for homomorphisms into a Gödel chain, the factor H+H^{+} accounts for those which send mm to a non-zero element, and the factor HH for those which send mm to 00. Since every Gödel hoop is the directed union of its principal filters, each of which has a least element, F(H)F(H) can be obtained in general by gluing together these products. In Theorem 3.3, we make this explicit: we describe F(H)F(H) as a quotient of a set of triples of elements of H+H^{+}, with operations given by explicit terms in the language of HH, and we verify directly that F(H)F(H) is a Gödel algebra with the required universal property. For instance, Examples 3.4 show that F(H)≅H+F(H) \cong H^{+} whenever HH is a totally ordered Gödel hoop without least element. The Gödel algebra H+H^{+} also appears in [12 Section 3.1], in connection with Esakia duality for relative Stone algebras, while Gödel algebras free over distributive lattices, which satisfy a different universal property, were described via Esakia duality in [8 Theorem 3.19]; see also [3] for the finite case. As a consequence, we recover the result of P. Köhler which expresses finitely generated free Gödel algebras in terms of free Gödel hoops; see Corollary 3.5 and [11 Theorem 5.7].

Our motivation comes from the study of internal actions in ideally exact categories. The variety GHoops\mathrm{GHoops} is a semi-abelian category [18, 19], while GAlg\mathrm{GAlg} is an ideally exact category in the sense of G. Janelidze [17]. By [17] and [20 Remark 3.3], a monadic functor from an ideally exact category U\mathcal{U}, with initial object 00 and terminal object 11, to a semi-abelian category is, up to an equivalence, the pullback functor U→(U↓0)\mathcal{U} \to(\mathcal{U} \downarrow0) along the unique morphism 0→10 \to1 if and only if the unit of the corresponding adjunction is cartesian. Monadic adjunctions of this kind, called ideally exact contexts, were used in [20] to introduce coherent and ideal actions, and it was shown in [20 Theorems 6.30 and 6.39] that MV-algebras and Wajsberg hoops, as well as product algebras and product hoops, form ideally exact contexts with a good theory of actions, that is, in which every coherent action is ideal, and so is every morphism between such actions. Internal actions and split extensions of hoops were also studied in [21, 22]. In [23], the ideally exact contexts with a good theory of actions were characterised among those obtained from a semi-abelian variety by freely adding constants, provided that the unit of the corresponding free–forgetful adjunction is cartesian. As observed in [23 Section 4], no explicit description was available for the left adjoints of the forgetful functors GAlg→GHoops\mathrm{GAlg} \to\mathrm{GHoops} and BLAlg→BHoops\mathrm{BLAlg} \to\mathrm{BHoops}, where BLAlg\mathrm{BLAlg} and BHoops\mathrm{BHoops} denote the varieties of BL-algebras and of basic hoops, and it was not known whether the units of the corresponding adjunctions are cartesian.

We show that neither of these units is cartesian, so that the corresponding free–forgetful adjunctions are not ideally exact contexts. For Gödel hoops, this is proved in Proposition 4.1, and the case of basic hoops follows in Corollary 4.2. Moreover, by Proposition 4.3, the semi-abelian categories GHoops and (GAlg ↓\downarrow 2), where 2 denotes the two-element Boolean algebra, are not equivalent, and, by Proposition 4.4, neither are BHoops and (BLAlg ↓\downarrow 2). This corrects the claim, made in the introduction of [21] but not used elsewhere in that paper, that these categories are equivalent.

Preliminaries

Definition 2.1. A Gödel hoop is an algebra (H,∧,∨,→,1)(H, \wedge, \vee, \to, 1) such that (H,∧,∨,1)(H, \wedge, \vee, 1) is a lattice with top element 1 and, for all x,y,z∈Hx, y, z \in H,

x∧y≤z⟺x≤y→z,(1)x \wedge y \le z \quad\Longleftrightarrow\quad x \le y \to z, \tag*{(1)}
(x→y)∨(y→x)=1.(2)(x \to y) \vee(y \to x) = 1. \tag*{(2)}

A Gödel algebra is an algebra (A,∧,∨,→,0,1)(A, \wedge, \vee, \to, 0, 1) such that (A,∧,∨,→,1)(A, \wedge, \vee, \to, 1) is a Gödel hoop and 0≤x0 \le x for every x∈Ax \in A.

Thus, Gödel hoops are the prelinear Brouwerian algebras, also known as relative Stone algebras [11, 12], and Gödel algebras are the prelinear Heyting algebras. Gödel hoops are usually defined as the basic hoops (H,⋅,→,1)(H, \cdot, \to, 1) satisfying x⋅x=xx \cdot x = x [2]: the two presentations are term equivalent, since in such hoops x⋅y=x∧yx \cdot y = x \wedge y and x∨y=((x→y)→y)∧((y→x)→x)x \vee y = ((x \to y) \to y) \wedge((y \to x) \to x). Gödel hoops and Gödel algebras form varieties, which we denote by GHoops and GAlg. Homomorphisms preserve all the operations and constants of the respective signatures. In particular, a homomorphism of Gödel hoops between two Gödel algebras need not preserve their least elements. The underlying lattice of a Gödel hoop is distributive, and we freely use elementary identities such as x∧(x→y)=x∧yx \wedge(x \to y) = x \wedge y, y≤x→yy \le x \to y and x→(y→z)=(x∧y)→zx \to(y \to z) = (x \wedge y) \to z, as well as the fact that →\to is antitone in its first and monotone in its second argument.

The forgetful functor U:GAlg→GHoopsU : \mathrm{GAlg} \to\mathrm{GHoops}, which forgets the constant 0, has a left adjoint FF, since it is induced by an inclusion of signatures between varieties. In a Gödel algebra we write ¬x=x→0\neg x = x \to0. We denote by 2={0<1}\mathbf{2} = \{0 < 1\} the two-element Boolean algebra, which is the initial Gödel algebra, by 1={1}\mathbf{1} = \{1\} the trivial Gödel hoop, which is the zero object of GHoops, and by CnC_n the nn-element chain, regarded as a Gödel algebra or as a Gödel hoop.

Lemma 2.2. Let AA be a Gödel algebra.

(1) For every e∈Ae \in A, the interval ↓e={x∈A∣x≤e}\downarrow e = \{x \in A \mid x \le e\}, with the restricted lattice operations, the bottom element 0, the top element ee and the implication x→ey=e∧(x→y)x \to_e y = e \wedge(x \to y), is a Gödel algebra, and the map pe:A→↓e:x↦x∧ep_e : A \to\downarrow e: x \mapsto x \wedge e is a surjective homomorphism of Gödel algebras.

(2) For every a∈Aa \in A, the elements c=¬ac = \neg a and d=¬¬ad = \neg\neg a satisfy c∧d=0c \wedge d = 0 and c∨d=1c \vee d = 1, and the map A→↓d×↓c:x↦(x∧d,x∧c)A \to\downarrow d \times\downarrow c: x \mapsto(x \wedge d, x \wedge c) is an isomorphism of Gödel algebras, whose inverse is (u,v)↦u∨v(u,v) \mapsto u \vee v.

Proof. (1) For all x,y∈Ax,y \in A and z∈↓ez \in\downarrow e, we have

z≤pe(x→y)⟺z∧x≤y⟺z∧x∧e≤y∧e⟺z≤pe(x)→epe(y),z \le p_e(x \to y) \Longleftrightarrow z \wedge x \le y \Longleftrightarrow z \wedge x \wedge e \le y \wedge e \Longleftrightarrow z \le p_e(x) \to_e p_e(y),

since z∧x≤ez \wedge x \le e. If x,y∈↓ex,y \in\downarrow e, then pe(x)=xp_e(x) = x and pe(y)=yp_e(y) = y, and these equivalences show that ↓e\downarrow e satisfies (2.1) with respect to →e\to_e. For arbitrary x,y∈Ax,y \in A, they show that pe(x→y)=pe(x)→epe(y)p_e(x \to y) = p_e(x) \to_e p_e(y). Clearly, pep_e preserves meets and constants, and it preserves joins by distributivity. Since pep_e is surjective, (2.2) holds in ↓e\downarrow e.

(2) We have c∧d=c∧(c→0)=c∧0=0c \wedge d = c \wedge(c \to0) = c \wedge0 = 0 and a→c=(a∧d)→0=ca \to c = (a \wedge d) \to0 = c. Since a∧c=0a \wedge c = 0, for every z∈Az \in A we have z∧c≤az \wedge c \le a if and only if z∧c=0z \wedge c = 0, so that c→a=c→0=dc \to a = c \to0 = d. Hence, (2.2) applied to aa and cc gives c∨d=1c \vee d = 1. The map in (2) is a homomorphism by (1), and it is bijective, with the stated inverse, since cc and dd are complementary elements of a distributive lattice. ∎

For a Gödel hoop HH, we denote by H+=H⊔{⊥}H^{+}=H\sqcup\{\perp\} the set obtained by adding to HH a new element ⊥\perp, declared to be strictly below every element of HH. We keep the operations of HH, and for u∈H+u\in H^{+} and x∈Hx\in H we put

⊥∧u=u∧⊥=⊥,⊥∨u=u∨⊥=u,⊥→u=1,x→⊥=⊥.\perp\wedge u=u\wedge\perp=\perp,\qquad\perp\vee u=u\vee\perp=u,\qquad\perp\to u=1,\qquad x\to\perp=\perp.

Lemma 2.3. Let HH be a Gödel hoop.

(1) H+H^{+} is a Gödel algebra with bottom element ⊥\perp.

(2) Every homomorphism of Gödel hoops f:H→Kf:H\to K extends to a homomorphism of Gödel algebras f+:H+→K+f^{+}:H^{+}\to K^{+} with f+(⊥)=⊥f^{+}(\perp)=\perp.

(3) Let AA be a Gödel algebra and let g:H→U(A)g:H\to U(A) be a homomorphism of Gödel hoops. Then the extension g♯:H+→Ag^{\sharp}:H^{+}\to A of gg with g♯(⊥)=0g^{\sharp}(\perp)=0 is a homomorphism of Gödel algebras if and only if ¬g(x)=0\neg g(x)=0 for every x∈Hx\in H.

Proof. (1) Condition (2.2) holds whenever one of its arguments is ⊥\perp, since ⊥→u=1\perp\to u=1 for every u∈H+u\in H^{+}. As for (2.1), both sides hold if y=⊥y=\perp, both sides are equivalent to x=⊥x=\perp if y∈Hy\in H and z=⊥z=\perp, and in the remaining cases (2.1) holds in HH or trivially.

(2) This follows from the definitions, since ff maps HH into KK.

(3) The map g♯g^{\sharp} preserves the constants and all the operations, except possibly the implications x→⊥=⊥x\to\perp=\perp with x∈Hx\in H, which are preserved if and only if g(x)→0=0g(x)\to0=0, that is, ¬g(x)=0\neg g(x)=0. □

Finally, let HH be a Gödel hoop and let b∈Hb\in H. The principal filter ↑b\mathord{\uparrow}b of the elements x∈Hx\in H such that b≤xb\leq x is closed under the operations of HH, since y≤x→yy\leq x\to y. We denote by HbH_b the Gödel algebra obtained by regarding ↑b\mathord{\uparrow}b as a Gödel hoop with bottom element bb, and we put Hb+=(↑b)+H_b^{+}=(\mathord{\uparrow}b)^{+}. Notice that the operations ∧\wedge, ∨\vee and →\to of the Gödel algebra Hb+×HbH_b^{+}\times H_b are the restrictions of those of H+×HH^{+}\times H, that its bottom element is (⊥,b)(\perp,b), and that pseudocomplements in HbH_b are given by ¬bx=x→b\neg_b x=x\to b.

The construction of the free functor

Let HH be a Gödel hoop. For every b∈Hb\in H, the principal filter ↑b\mathord{\uparrow}b is a Gödel hoop with least element bb. We first show in Lemma 3.1 that Hb+×HbH_b^{+}\times H_b is the free Gödel algebra on ↑b\mathord{\uparrow}b, and we describe in Lemma 3.2 the homomorphisms between these free algebras induced by the inclusions ↑b⊆↑c\mathord{\uparrow}b\subseteq\mathord{\uparrow}c, for c≤bc\leq b. Since HH is the directed union of its principal filters, F(H)F(H) is then obtained in Theorem 3.3 by gluing the algebras Hb+×HbH_b^{+}\times H_b along these homomorphisms.

Lemma 3.1. Let HH be a Gödel hoop, let b∈Hb\in H, let AA be a Gödel algebra and let g:↑b→U(A)g:\mathord{\uparrow}b\to U(A) be a homomorphism of Gödel hoops. Then the map gˉ:Hb+×Hb→A\bar{g}:H_b^{+}\times H_b\to A defined by

gˉ(u,v)=(g♯(u)∧¬¬g(b))∨(g(v)∧¬g(b))(3)\bar{g}(u,v)=\left(g^{\sharp}(u)\wedge\neg\neg g(b)\right)\vee\left(g(v)\wedge\neg g(b)\right) \tag*{(3)}

is the unique homomorphism of Gödel algebras such that gˉ(x,x)=g(x)\bar{g}(x,x)=g(x) for every x∈↑bx\in\mathord{\uparrow}b.

Proof. Put c=¬g(b)c=\neg g(b) and d=¬¬g(b)d=\neg\neg g(b). By Lemma 2.2(1), pc∘g:↑b→↓cp_c\circ g:\mathord{\uparrow}b\to\mathord{\downarrow}c is a homomorphism of Gödel hoops which sends bb to g(b)∧c=0g(b)\wedge c=0, hence a homomorphism of Gödel algebras Hb→↓cH_b\to\mathord{\downarrow}c. For every x∈↑bx\in\mathord{\uparrow}b we have g(b)≤g(x)∧dg(b)\leq g(x)\wedge d, so that the pseudocomplement d∧¬(g(x)∧d)d\wedge\neg(g(x)\wedge d) of pd(g(x))p_d(g(x)) in ↓d\mathord{\downarrow}d is below d∧c=0d\wedge c=0. Thus, by Lemma 2.3(3), the extension (pd∘g)♯:Hb+→↓d(p_d\circ g)^{\sharp}:H_b^{+}\to\mathord{\downarrow}d is a homomorphism of Gödel algebras. Composing (pd∘g)♯×(pc∘g)(p_d\circ g)^{\sharp}\times(p_c\circ g) with the inverse of the isomorphism of Lemma 2.2(2), we obtain that gˉ\bar{g} is a homomorphism of Gödel algebras. Moreover, gˉ(x,x)=g(x)∧(d∨c)=g(x)\bar{g}(x,x)=g(x)\wedge(d\vee c)=g(x) for every x∈↑bx\in\mathord{\uparrow}b. For the uniqueness, notice that the bottom element of Hb+×HbH_b^+ \times H_b is (⊥,b)(\bot,b), so that ¬(b,b)=(⊥,1)\neg(b,b)=(\bot,1) and ¬¬(b,b)=(1,b)\neg\neg(b,b)=(1,b). Hence, for all u,v∈↑bu,v\in\uparrow b,

(u,v)=((u,u)∧¬¬(b,b))∨((v,v)∧¬(b,b)),(⊥,v)=(v,v)∧¬(b,b).(4)(u,v)=((u,u)\wedge\neg\neg(b,b))\vee((v,v)\wedge\neg(b,b)),\qquad(\bot,v)=(v,v)\wedge\neg(b,b). \tag*{(4)}

Therefore, Hb+×HbH_b^+ \times H_b is generated, as a Gödel algebra, by its diagonal {(x,x)∣x∈↑b}\{(x,x)\mid x\in\uparrow b\}, and the uniqueness follows. □

In other words, Hb+×HbH_b^+ \times H_b, together with the map x↦(x,x)x\mapsto(x,x), is the free Gödel algebra on the Gödel hoop ↑b\uparrow b. The definition of gˉ\bar{g} reflects the decomposition A≅↓¬¬g(b)×↓¬g(b)A\cong\downarrow\neg\neg g(b)\times\downarrow\neg g(b) of Lemma 2.2(2): in the first factor, the image of every x∈↑bx\in\uparrow b has pseudocomplement 00, so that ⊥\bot can be sent to the bottom element, while in the second factor g(b)g(b) becomes the bottom element, so that HbH_b is mapped there as a Gödel algebra.

Let c≤bc\leq b in HH, so that ¬cb=b→c\neg_c b=b\to c and ¬c¬cb=(b→c)→c\neg_c\neg_c b=(b\to c)\to c are the pseudocomplement and the double pseudocomplement of bb in HcH_c. For (u,v)∈Hb+×Hb(u,v)\in H_b^+\times H_b, we put

ρb,c(u,v)=(u∧¬c¬cb)∨(v∧¬cb),(5)\rho_{b,c}(u,v)=(u\wedge\neg_c\neg_c b)\vee(v\wedge\neg_c b), \tag*{(5)}

where the operations are computed in H+H^+. Since c≤v∧¬cbc\leq v\wedge\neg_c b, we have ρb,c(u,v)∈↑c\rho_{b,c}(u,v)\in\uparrow c. For instance, ρb,c(⊥,v)=v∧¬cb\rho_{b,c}(\bot,v)=v\wedge\neg_c b. With respect to the decomposition Hc≅↓¬c¬cb×↓¬cbH_c\cong\downarrow\neg_c\neg_c b\times\downarrow\neg_c b of Lemma 2.2(2), the first component of ρb,c(u,v)\rho_{b,c}(u,v) only depends on uu, and the second one only on vv.

Lemma 3.2. Let HH be a Gödel hoop and let b,c∈Hb,c\in H with c≤bc\leq b. Then the map γb,c:Hb+×Hb→Hc+×Hc\gamma_{b,c}:H_b^+\times H_b\to H_c^+\times H_c defined by

γb,c(u,v)=(u,ρb,c(u,v))\gamma_{b,c}(u,v)=(u,\rho_{b,c}(u,v))

is the unique homomorphism of Gödel algebras such that γb,c(x,x)=(x,x)\gamma_{b,c}(x,x)=(x,x) for every x∈↑bx\in\uparrow b. Moreover, γb,b\gamma_{b,b} is the identity and γc,d∘γb,c=γb,d\gamma_{c,d}\circ\gamma_{b,c}=\gamma_{b,d} whenever d≤c≤bd\leq c\leq b.

Proof. We apply Lemma 3.1 to the homomorphism of Gödel hoops g:↑b→U(Hc+×Hc):x↦(x,x)g:\uparrow b\to U(H_c^+\times H_c):x\mapsto(x,x). Write p=¬cbp=\neg_c b and q=¬c¬cbq=\neg_c\neg_c b. Since the bottom element of Hc+×HcH_c^+\times H_c is (⊥,c)(\bot,c), we have g#(⊥)=(⊥,c)g^\#(\bot)=(\bot,c) and

¬g(b)=(b→⊥,b→c)=(⊥,p),¬¬g(b)=(⊥→⊥,p→c)=(1,q).\neg g(b)=(b\to\bot,b\to c)=(\bot,p),\qquad\neg\neg g(b)=(\bot\to\bot,p\to c)=(1,q).

Hence, by (3.1), we get

gˉ(u,v)=(u,(u∧q)∨(v∧p))if u∈↑b,gˉ(⊥,v)=(⊥,(c∧q)∨(v∧p)).\bar{g}(u,v)=(u,(u\wedge q)\vee(v\wedge p))\quad\text{if }u\in\uparrow b,\qquad \bar{g}(\bot,v)=(\bot,(c\wedge q)\vee(v\wedge p)).

Since c∧q=c≤v∧pc\wedge q=c\leq v\wedge p, in both cases we have gˉ(u,v)=γb,c(u,v)\bar{g}(u,v)=\gamma_{b,c}(u,v). The last assertion follows from the uniqueness in Lemma 3.1, since γb,b\gamma_{b,b}, γc,d∘γb,c\gamma_{c,d}\circ\gamma_{b,c} and γb,d\gamma_{b,d} send (x,x)(x,x) to (x,x)(x,x) for every x∈↑bx\in\uparrow b. □

We are now ready to describe F(H)F(H). Let

T(H)={(b,u,v)∈H×H+×H∣(u,v)∈Hb+×Hb},T(H)=\{(b,u,v)\in H\times H^+\times H\mid(u,v)\in H_b^+\times H_b\},

that is, (b,u,v)∈T(H)(b,u,v)\in T(H) if and only if b≤vb\leq v and either u=⊥u=\bot or b≤ub\leq u. We write (b,u,v)∼(b′,u′,v′)(b,u,v)\sim(b',u',v') if there exists c≤b∧b′c\leq b\wedge b' such that γb,c(u,v)=γb′,c(u′,v′)\gamma_{b,c}(u,v)=\gamma_{b',c}(u',v'), that is, such that u=u′u=u' and ρb,c(u,v)=ρb′,c(u′,v′)\rho_{b,c}(u,v)=\rho_{b',c}(u',v').

Thus, a triple (b,u,v)(b,u,v) represents the element (u,v)(u,v) of the free Gödel algebra Hb+×HbH_b^+\times H_b on ↑b\uparrow b, and two triples are equivalent when they have the same image at some common stage c≤b∧b′c\leq b\wedge b'. Since HH is the directed union of its principal filters and FF, being a left adjoint, preserves directed colimits, F(H)F(H) is the directed colimit of the Gödel algebras Hb+×HbH_b^+\times H_b along the Gödel homomorphisms γb,c\gamma_{b,c}. In the following theorem, we verify directly that the quotient T(H)/∼T(H)/\sim is a Gödel algebra with the required universal property, and we make its structure explicit.

Theorem 3.3. Let HH be a Gödel hoop.

(1) The relation ∼\sim is an equivalence relation on T(H)T(H). If [b,u,v][b,u,v] denotes the class of (b,u,v)(b,u,v), then F(H)=T(H)/∼F(H)=T(H)/\sim is a Gödel algebra with constants 0=[b,⊥,b]0=[b,\bot,b] and 1=[b,1,1]1=[b,1,1], which do not depend on b∈Hb\in H, and with operations

[b,u,v]⋆[b′,u′,v′]=[c,u⋆u′,ρb,c(u,v)⋆ρb′,c(u′,v′)](6)[b,u,v]\star[b',u',v']=[c,u\star u',\rho_{b,c}(u,v)\star\rho_{b',c}(u',v')] \tag*{(6)}

for ⋆∈{∧,∨,→}\star\in\{\wedge,\vee,\to\}, where c=b∧b′c=b\wedge b' and the operations in the second and in the third entry are computed in H+H^+ and in HH, respectively.

(2) The map ηH:H→UF(H):h↦[h,h,h]\eta_H:H\to UF(H):h\mapsto[h,h,h] is a homomorphism of Gödel hoops. For every Gödel algebra AA and every homomorphism of Gödel hoops g:H→U(A)g:H\to U(A), the map g^:F(H)→A\hat{g}:F(H)\to A defined by

g^[b,u,v]=(g+(u)∧¬¬g(b))∨(g(v)∧¬g(b))(7)\hat{g}[b,u,v]=(g^+(u)\wedge\neg\neg g(b))\vee(g(v)\wedge\neg g(b)) \tag*{(7)}

is the unique homomorphism of Gödel algebras such that U(g^)∘ηH=gU(\hat{g})\circ\eta_H=g. Hence, FF is the left adjoint of UU, with unit η\eta.

(3) For every homomorphism of Gödel hoops f:H→Kf:H\to K, the homomorphism F(f):F(H)→F(K)F(f):F(H)\to F(K) is given by

F(f)[b,u,v]=[f(b),f+(u),f(v)].F(f)[b,u,v]=[f(b),f^+(u),f(v)].

Proof. For (b,u,v)∈T(H)(b,u,v)\in T(H), we write X=(u,v)X=(u,v) and [b,X]=[b,u,v][b,X]=[b,u,v], so that (3.4) reads

[b,X]⋆[b′,X′]=[c,γb,c(X)⋆γb′,c(X′)],[b,X]\star[b',X']=[c,\gamma_{b,c}(X)\star\gamma_{b',c}(X')],

where the operation on the right-hand side is computed in the Gödel algebra Hc+×HcH^+_c\times H_c.

(1) The relation ∼\sim is clearly reflexive and symmetric. If (b,X)∼(b′,X′)(b,X)\sim(b',X') and (b′,X′)∼(b′′,X′′)(b',X')\sim(b'',X''), with witnesses cc and c′c' respectively, then d=c∧c′d=c\wedge c' satisfies

γb,d(X)=γc,dγb,c(X)=γc,dγb′,c(X′)=γb′,d(X′)=γc′,dγb′,c′(X′)=γc′,dγb′′,c′(X′′)=γb′′,d(X′′)\begin{aligned} \gamma_{b,d}(X)=\gamma_{c,d}\gamma_{b,c}(X)=\gamma_{c,d}\gamma_{b',c}(X')=\gamma_{b',d}(X') \\ &=\gamma_{c',d}\gamma_{b',c'}(X')=\gamma_{c',d}\gamma_{b'',c'}(X'')=\gamma_{b'',d}(X'') \end{aligned}

by Lemma 3.2, so that ∼\sim is transitive. Moreover, (b,X)∼(c,γb,c(X))(b,X)\sim(c,\gamma_{b,c}(X)) for every c≤bc\le b, that is,

[b,X]=[c,γb,c(X)](8)[b,X]=[c,\gamma_{b,c}(X)] \tag*{(8)}

whenever c≤bc\le b.

Next, we show that the operations (3.4) are well defined. Let d≤c=b∧b′d\le c=b\wedge b'. Since γc,d\gamma_{c,d} is a homomorphism, (3.6) and Lemma 3.2 give

[c,γb,c(X)⋆γb′,c(X′)]=[d,γc,d(γb,c(X)⋆γb′,c(X′))]=[d,γb,d(X)⋆γb′,d(X′)],[c,\gamma_{b,c}(X)\star\gamma_{b',c}(X')]=[d,\gamma_{c,d}(\gamma_{b,c}(X)\star\gamma_{b',c}(X'))]=[d,\gamma_{b,d}(X)\star\gamma_{b',d}(X')],

so that the right-hand side of (3.4) may be computed at any d≤b∧b′d\le b\wedge b'. Now, if (b,X)∼(b1,X1)(b,X)\sim(b_1,X_1) with witness ee, and d=e∧b′d=e\wedge b', then γb,d(X)=γe,dγb,e(X)=γe,dγb1,e(X1)=γb1,d(X1)\gamma_{b,d}(X)=\gamma_{e,d}\gamma_{b,e}(X)=\gamma_{e,d}\gamma_{b_1,e}(X_1)=\gamma_{b_1,d}(X_1), so that [b,X]⋆[b′,X′]=[b1,X1]⋆[b′,X′][b,X]\star[b',X']=[b_1,X_1]\star[b',X']. The same argument applies to the second variable. Similarly, the constants do not depend on bb, since each γb,c\gamma_{b,c} sends the bottom element (⊥,b)(\bot,b) and the top element (1,1)(1,1) of Hb+×HbH^+_b\times H_b to those of Hc+×HcH^+_c\times H_c.

It remains to show that F(H)F(H) is a Gödel algebra. For c∈Hc\in H, let ScS_c be the image of the map

λc:Hc+×Hc→F(H):X↦[c,X].\lambda_c:H^+_c\times H_c\to F(H):X\mapsto[c,X].

By (3.4) and the definition of 00 and 11, λc\lambda_c preserves the operations and the constants. Hence, ScS_c is closed under the operations of F(H)F(H), and it is a homomorphic image of the Gödel algebra Hc+×HcH^+_c\times H_c. Since Gödel algebras form a variety, ScS_c is a Gödel algebra. Moreover, F(H)F(H) is the union of the subsets SbS_b, with b∈Hb\in H, and Sb⊆ScS_b\subseteq S_c whenever c≤bc\le b, by (3.6). Thus, any finitely many elements of F(H)F(H) belong to a common ScS_c, namely for cc the meet of the first components of chosen representatives. Since each condition in Definition 2.1 involves finitely many elements, F(H)F(H) is a Gödel algebra. (2) Let x,y∈Hx,y\in H, and put c=x∧yc=x\wedge y and z=x⋆yz=x\star y, so that c≤zc\leq z (if ⋆\star is →\to, since c≤y≤x→yc\leq y\leq x\to y). By Lemma 3.2 and (3.6), we have

ηH(x)⋆ηH(y)=[c,(x,x)⋆(y,y)]=[c,z,z]=[z,z,z]=ηH(z).\eta_H(x)\star\eta_H(y)=[c,(x,x)\star(y,y)]=[c,z,z]=[z,z,z]=\eta_H(z).

Moreover, ηH(1)=[1,1,1]=1\eta_H(1)=[1,1,1]=1, so that ηH\eta_H is a homomorphism of Gödel hoops.

Let g:H→U(A)g:H\to U(A) be a homomorphism of Gödel hoops and, for b∈Hb\in H, let gˉb:Hb+×Hb→A\bar{g}_b:H_b^+\times H_b\to A be the homomorphism of Lemma 3.1 associated with the restriction of gg to ↑b\uparrow b. If c≤bc\leq b, then gˉc∘γb,c\bar{g}_c\circ\gamma_{b,c} and gˉb\bar{g}_b are homomorphisms of Gödel algebras sending (x,x)(x,x) to g(x)g(x) for every x∈↑bx\in\uparrow b, so that gˉc∘γb,c=gˉb\bar{g}_c\circ\gamma_{b,c}=\bar{g}_b by Lemma 3.1. It follows that g^[b,X]=gˉb(X)\hat{g}[b,X]=\bar{g}_b(X) is well defined, and that g^∘λc=gˉc\hat{g}\circ\lambda_c=\bar{g}_c for every c∈Hc\in H. Since any two elements of F(H)F(H) belong to a common ScS_c, and λc\lambda_c and gˉc\bar{g}_c are homomorphisms, so is g^\hat{g}. Moreover, g^(ηH(h))=gˉh(h,h)=g(h)\hat{g}(\eta_H(h))=\bar{g}_h(h,h)=g(h) for every h∈Hh\in H.

For the uniqueness, notice that λb(x,x)=[b,x,x]=[x,x,x]=ηH(x)\lambda_b(x,x)=[b,x,x]=[x,x,x]=\eta_H(x) for every x∈↑bx\in\uparrow b, by (3.6). Applying the homomorphism λb\lambda_b to (3.2), we get

[b,u,v]=(ηH(u)∧¬¬ηH(b))∨(ηH(v)∧¬ηH(b))(9)[b,u,v]=\left(\eta_H(u)\wedge\neg\neg\eta_H(b)\right)\vee\left(\eta_H(v)\wedge\neg\eta_H(b)\right) \tag*{(9)}

for every (b,u,v)∈T(H)(b,u,v)\in T(H), where ηH(⊥)\eta_H(\bot) stands for 00. Hence, F(H)F(H) is generated by ηH(H)\eta_H(H) as a Gödel algebra, and g^\hat{g} is unique.

(3) By (2), F(f)F(f) is the unique homomorphism of Gödel algebras such that F(f)∘ηH=ηK∘fF(f)\circ\eta_H=\eta_K\circ f. Applying (3.5) to g=ηK∘fg=\eta_K\circ f, for which g+(u)=ηK(f+(u))g^+(u)=\eta_K(f^+(u)) with the convention ηK(⊥)=0\eta_K(\bot)=0, and then (3.7) in F(K)F(K), we get F(f)[b,u,v]=[f(b),f+(u),f(v)]F(f)[b,u,v]=[f(b),f^+(u),f(v)]. □

The homomorphism g^\hat{g} has a particularly simple form when AA is a Gödel chain, since then ¬g(b)\neg g(b) is 11 if g(b)=0g(b)=0, and 00 otherwise. Hence, g^[b,u,v]=g+(u)\hat{g}[b,u,v]=g^+(u) if g(b)≠0g(b)\ne0, while g^[b,u,v]=g(v)\hat{g}[b,u,v]=g(v) if g(b)=0g(b)=0: the third entry vv of a triple (b,u,v)(b,u,v) only matters for the homomorphisms which send bb to 00.

The map πH:F(H)→H+\pi_H:F(H)\to H^+: [b,u,v]↦u[b,u,v]\mapsto u is a surjective homomorphism of Gödel algebras by (3.4), and πH∘ηH\pi_H\circ\eta_H is the inclusion of HH into H+H^+. In particular, the unit ηH\eta_H is injective.

Examples

(1) Suppose that HH has a least element mm, and denote by HmH_m the Gödel algebra obtained by regarding HH as a Gödel hoop with bottom element mm. Then the homomorphism λm:H+×Hm→F(H):(u,v)↦[m,u,v]\lambda_m:H^+\times H_m\to F(H):(u,v)\mapsto[m,u,v] is an isomorphism. Indeed, it is surjective by (3.6), and it is injective since mm is the only element below mm and γm,m\gamma_{m,m} is the identity. Under this isomorphism, ηH(x)\eta_H(x) corresponds to (x,x)(x,x). For instance, F(1)≅1+×1≅2F(1)\cong1^+\times1\cong2 and F(C2)≅C3×C2F(C_2)\cong C_3\times C_2. In particular, if AA is a Gödel algebra, then FU(A)≅A+×AFU(A)\cong A^+\times A, and by (3.5) the counit FU(A)→AFU(A)\to A of the adjunction is the second projection.

(2) Suppose that HH is a totally ordered Gödel hoop without least element. In such a hoop, we have x→y=yx\to y=y whenever y<xy<x. Let (b,u,v),(b′,u′,v′)∈T(H)(b,u,v),(b',u',v')\in T(H), and choose c≤b∧b′c\leq b\wedge b'. Then cb=cc_b=c and ¬c¬b=¬c\neg_c\neg_b=\neg_c, so that ρb,c(u,v)=u∨(v∧c)=u∨v\rho_{b,c}(u,v)=u\vee(v\wedge c)=u\vee v, and similarly ρb′,c(u′,v′)=u′∨v\rho_{b',c}(u',v')=u'\vee v. Hence, [b,u,v]=[b′,u′,v′][b,u,v]=[b',u',v'], and the homomorphism πH\pi_H is an isomorphism. In other words, F(H)F(H) is obtained by adjoining a new bottom element to HH. For instance, F((0,1])F((0,1]) is the standard Gödel algebra [0,1][0,1].

(3) Let C2={m<1}C_2=\{m<1\} and let H=C2×(0,1]H=C_2\times(0,1], which is not totally ordered and has no least element. We write x=(x1,x2)x=(x_1,x_2) for x∈Hx\in H. By (3.6), every element of F(H)F(H) is of the form [b,u,v][b,u,v] with b=(m,t)b=(m,t). If c=(m,s)c=(m,s) with s<ts<t, then ¬cb=(1,s)\neg_c b=(1,s) and ¬c¬cb=(m,1)\neg_c\neg_c b=(m,1), so that ρb,c(u,v)\rho_{b,c}(u,v) is (v1,v2)(v_1,v_2) if u≠⊥u\ne\bot, and (v1,s)(v_1,s) otherwise. Hence, [b,u,v][b,u,v] only depends on u∨vu\vee v, and (3.5), applied to x↦(x,x1)x\mapsto(x,x_1), gives an isomorphism F(H)→H+×2F(H)\to H^+\times2 which sends [b,u,v][b,u,v], with b=(m,t)b=(m,t), to (u,v1)(u,v_1), where C2C_2 is identified with U(2)U(2). In particular, πH\pi_H is not an isomorphism.

Corollary 3.5. Let PP be the free Gödel hoop on a finite set XX, and let m=⋀Xm=\bigwedge X, where m=1m=1 if X=∅X=\varnothing. Then mm is the least element of PP, and the free Gödel algebra on XX is isomorphic to P+×PmP^+ \times P_m. In particular, if PP has kk elements, then the free Gödel algebra on XX has k(k+1)k(k+1) elements.

Proof. The principal filter ↑m\uparrow m is closed under the operations of PP and contains XX, so that ↑m=P\uparrow m=P. Since FF is a left adjoint, F(P)F(P) is the free Gödel algebra on XX, and the result follows from Examples 3.4(1). □\square

This is the description of finitely generated free Gödel algebras given in [11 Theorem 5.7], where it is attributed to P. Köhler. For instance, the free Gödel hoops on one and two generators have 2 and 18 elements, so that the corresponding free Gödel algebras have 6 and 342 elements; cf. [16].

A connection with ideally exact categories

Recall that a natural transformation is cartesian if all its naturality squares are pullbacks.

Proposition 4.1. The unit η:1GHoops⇒UF\eta:1_{\mathrm{GHoops}}\Rightarrow UF of the adjunction F⊣UF\dashv U is not cartesian.

Proof. Let H=C2={m<1}H=C_2=\{m<1\}, let τ:H→1\tau:H\to1 be the unique homomorphism, and consider the naturality square

H→ηHUF(H)τ↓↓UF(τ)1→η1UF(1).\begin{CD} H @>{\eta_H}>> UF(H) \\ @V{\tau}VV @VV{UF(\tau)}V \\ 1 @>{\eta_1}>> UF(1). \end{CD}

By Theorem 3.3(3), we have F(τ)[b,u,v]=[1,τ+(u),1]F(\tau)[b,u,v]=[1,\tau^+(u),1], which is 00 if u=⊥u=\perp and 11 otherwise. Since η1\eta_1 selects the top element of UF(1)UF(1) and pullbacks in GHoops are computed on underlying sets, the pullback of UF(τ)UF(\tau) along η1\eta_1 is the set of the elements [b,u,v][b,u,v] of F(H)F(H) such that u≠⊥u\ne\perp. By Examples 3.4(1), this set has four elements, namely the classes [m,u,v][m,u,v] with u,v∈{m,1}u,v\in\{m,1\}, whereas HH has two elements. Hence, the square is not a pullback. □\square

Let UBL:BLAlg→BHoopsU_{\mathrm{BL}}:\mathrm{BLAlg}\to\mathrm{BHoops} be the functor which forgets the constant 00. As for Gödel algebras, UBLU_{\mathrm{BL}} has a left adjoint FBLF_{\mathrm{BL}}, and we denote by ηBL\eta^{\mathrm{BL}} the unit of the adjunction FBL⊣UBLF_{\mathrm{BL}}\dashv U_{\mathrm{BL}}. Recall that Gödel algebras are the BL-algebras satisfying x⋅x=xx\cdot x=x, in which case x⋅y=x∧yx\cdot y=x\wedge y, and that Gödel hoops are the basic hoops satisfying the same identity [2, 14].

Corollary 4.2. The unit ηBL:1BHoops⇒UBLFBL\eta^{\mathrm{BL}}:1_{\mathrm{BHoops}}\Rightarrow U_{\mathrm{BL}}F_{\mathrm{BL}} of the adjunction FBL⊣UBLF_{\mathrm{BL}}\dashv U_{\mathrm{BL}} is not cartesian.

Proof. Let H=C2={m<1}H=C_2=\{m<1\} and τ:H→1\tau:H\to1 be as in the proof of Proposition 4.1, and regard HH as a basic hoop and F(H)F(H) as a BL-algebra. Let q:FBL(H)→F(H)q:F_{\mathrm{BL}}(H)\to F(H) be the homomorphism of BL-algebras such that q∘ηHBL=ηHq\circ\eta_H^{\mathrm{BL}}=\eta_H. By (3.7), F(H)F(H) is generated by ηH(H)\eta_H(H) as a Gödel algebra, hence also as a BL-algebra, since x∧y=x⋅yx\wedge y=x\cdot y and x∨y=((x→y)→y)∧((y→x)→x)x\vee y=((x\to y)\to y)\wedge((y\to x)\to x) in F(H)F(H). Thus, qq is surjective. Both FBL(1)F_{\mathrm{BL}}(1) and F(1)F(1) are the initial BL-algebra $2$, and F(τ)∘q=FBL(τ)F(\tau)\circ q=F_{\mathrm{BL}}(\tau) by the universal property of ηHBL\eta_H^{\mathrm{BL}}, since both homomorphisms send ηHBL(x)\eta_H^{\mathrm{BL}}(x) to 11 for every x∈Hx\in H. Let KK and KBLK_{\mathrm{BL}} be the preimages of 11 under F(τ)F(\tau) and FBL(τ)F_{\mathrm{BL}}(\tau), respectively. If y∈FBL(H)y\in F_{\mathrm{BL}}(H) such that q(y)∈Kq(y)\in K, then FBL(τ)(y)=F(τ)(q(y))=1F_{\mathrm{BL}}(\tau)(y)=F(\tau)(q(y))=1. Hence, $q

maps KBLK_{\mathrm{BL}} onto KK, which has four elements by the proof of Proposition 4.1. Since KBLK_{\mathrm{BL}} is the pullback of UBLFBL(τ)U_{\mathrm{BL}}F_{\mathrm{BL}}(\tau) along η1BL\eta^{\mathrm{BL}}_1, and HH has two elements, the naturality square of ηBL\eta^{\mathrm{BL}} at τ\tau is not a pullback. □

Recall that a variety is a semi-abelian category if and only if it is pointed and protomodular [7, 18], and that a category is ideally exact if it is Barr-exact and protomodular, it has finite coproducts, and the unique morphism from its initial object to its terminal object is a regular epimorphism [17]. The variety GAlg is protomodular, since the terms witnessing the protomodularity of GHoops [7, 19] also witness that of GAlg. Since the unique morphism from the initial object 22 of GAlg to the trivial Gödel algebra is surjective, GAlg is an ideally exact category [17], and the slice category (GAlg ↓2\downarrow2) is semi-abelian; see [20 Remark 3.3].

Proposition 4.3. The semi-abelian categories GHoops and (GAlg ↓2\downarrow2) are not equivalent.

Proof. Recall that an object of a semi-abelian category is simple if it is not a zero object and every regular epimorphism defined on it is either an isomorphism or a morphism to a zero object. We show that GHoops has, up to isomorphism, exactly one simple object, whereas (GAlg ↓2\downarrow2) has at least two.

The quotients of a Gödel hoop HH correspond to its filters, that is, to the upward closed subsets of HH containing 11 and closed under ∧\wedge [5]. Since every principal filter ↑x\uparrow x is a filter, if HH is simple, then ↑x=H\uparrow x = H for every x∈Hx \in H with x≠1x \ne1. Hence, every x≠1x \ne1 in HH is the least element of HH, and H≅C2H \cong C_2. Conversely, C2C_2 is simple.

The objects of (GAlg ↓2\downarrow2) are the pairs (A,p)(A,p), where p ⁣:A→2p\colon A \to2 is a homomorphism, and the zero object is (2,id⁡2)(2,\operatorname{id}_2). The quotients of (A,p)(A,p) are the surjective homomorphisms f ⁣:A→Bf\colon A \to B through which pp factors. Since the congruences of a Gödel algebra correspond to its filters [14], the quotients of (A,p)(A,p) correspond to the filters of AA contained in P=p−1(1)P=p^{-1}(1), the filter PP itself giving the zero object. If P={1}P=\{1\}, then (A,p)(A,p) is a zero object, since p(a)=0p(a)=0 implies ¬a∈P\neg a \in P, and hence a=a∧¬a=0a=a\wedge\neg a=0. If PP has more than two elements, then it contains some y≠1y\ne1 which is not its least element, and the principal filter ↑y\uparrow y lies strictly between {1}\{1\} and PP. Hence, (A,p)(A,p) is simple if and only if PP has exactly two elements, and (C3,p)(C_3,p), where C3={0<x<1}C_3=\{0<x<1\} and p−1(1)={x,1}p^{-1}(1)=\{x,1\}, and (2×2,π1)(2\times2,\pi_1), where π1\pi_1 is the first projection, are non-isomorphic simple objects of (GAlg ↓2\downarrow2). □

The same arguments show that BLAlg is an ideally exact category with initial object 22, so that (BLAlg ↓2\downarrow2) is semi-abelian.

Proposition 4.4. The semi-abelian categories BHoops and (BLAlg ↓2\downarrow2) are not equivalent.

Proof. We compare the objects with exactly two subobjects, a property which is invariant under equivalences: up to isomorphism, (BLAlg ↓2\downarrow2) has at least two of them, whereas BHoops has only one. The subobjects of a pair (A,p)(A,p) of (BLAlg ↓2\downarrow2) correspond to the subalgebras of AA. Since the only subalgebras of C3C_3 and of 2×22\times2 are {0,1}\{0,1\} and the whole algebra, the pairs (C3,p)(C_3,p) and (2×2,π1)(2\times2,\pi_1) of the proof of Proposition 4.3 are two non-isomorphic examples. It remains to show that every basic hoop with exactly two subhoops is isomorphic to C2C_2.

Let HH be such a hoop, so that HH is non-trivial and its only subhoops are {1}\{1\} and HH. Every filter of HH, that is, every upward closed subset containing 11 and closed under the product, is a subhoop, since y≤x→yy\le x\to y. Hence, the only filters of HH are {1}\{1\} and HH, so that, for all d,w∈Hd,w\in H with d≠1d\ne1, there exists nn such that dn≤wd^n\le w. If HH has an idempotent element e≠1e\ne1, then {e,1}\{e,1\} is a subhoop, and H={e,1}≅C2H=\{e,1\}\cong C_2. Suppose now that 11 is the only idempotent element of HH. Let a,b∈Ha,b\in H with a≠1a\ne1, and put w=a⋅bw=a\cdot b, c=a→wc=a\to w and d=c→bd=c\to b. Then b≤cb\le c,

and the identity x⋅(x→y)=y⋅(y→x)x \cdot(x \to y) = y \cdot(y \to x) gives a⋅c=w⋅(w→a)=wa \cdot c = w \cdot(w \to a) = w and c⋅d=b⋅(b→c)=bc \cdot d = b \cdot(b \to c) = b. Hence, w⋅d=a⋅c⋅d=ww \cdot d = a \cdot c \cdot d = w, so that w⋅dn=ww \cdot d^{n} = w for every nn. If d≠1d \ne1, then dn≤wd^{n} \le w for some nn, and w=w⋅dn≤w⋅w≤ww = w \cdot d^{n} \le w \cdot w \le w shows that ww is idempotent, which is impossible, since w≤a<1w \le a < 1. Thus, d=1d = 1, that is, a→a⋅b=ba \to a \cdot b = b. This identity also holds for a=1a = 1. Finally, let x∈Hx \in H with x≠1x \ne1. Since x2j→x2kx^{2j} \to x^{2k} is equal to x2(k−j)x^{2(k-j)} if j≤kj \le k, and to 11 otherwise, the set {x2k∣k≥0}\{x^{2k} \mid k \ge0\} is a subhoop of HH, and it is non-trivial, since x2≤x<1x^{2} \le x < 1. Therefore, it contains xx, that is, x=x2kx = x^{2k} for some k≥1k \ge1. But then x≤x2≤xx \le x^{2} \le x, so that xx is idempotent, a contradiction. □

The forgetful functor UU is monadic by Beck’s theorem [4 Section 3.3], since it reflects isomorphisms and preserves reflexive coequalisers, which in both varieties are computed on underlying sets. Therefore, by [17 Theorem 2.6], Proposition 4.3 gives another proof of Proposition 4.1, and shows more generally that no monadic functor GAlg →\to GHoops has a left adjoint with cartesian unit. In the terminology of [20], the adjunction F⊣UF \dashv U is not an ideally exact context, in contrast with the cases of Wajsberg hoops and product hoops; see [20 Section 6.14]. By the same argument, UBLU_{\mathrm{BL}} is monadic, and Corollary 4.2 shows that the adjunction FBL⊣UBLF_{\mathrm{BL}} \dashv U_{\mathrm{BL}} is not an ideally exact context either. More generally, by Proposition 4.4, no monadic functor BLAlg →\to BHoops has a left adjoint with cartesian unit. This answers the question raised in [23 Section 4]. A natural question, which we do not address here, is whether the ideally exact contexts associated with the unique morphisms from 22 to the terminal objects of GAlg and of BLAlg have a good theory of actions.

Acknowledgment

The author is supported by the “National Group for Algebraic and Geometric Structures and their Applications” (GNSAGA – INdAM), and he is a Postdoctoral Researcher of the Fonds de la Recherche Scientifique–FNRS.

Use of generative AI tools

Generative AI tools (Anthropic’s Claude Opus 5.5 and OpenAI’s GPT-6 Sol) were used to improve the presentation of the content and to help check the correctness of the proofs. The author reviewed all suggestions made by these tools and takes full responsibility for the content of this article.

References

Email address: [email protected]; [email protected]

Institut de Recherche en Mathématique et Physique, Université catholique de Louvain, chemin du Cyclotron 2 bte L7.01.02, B–1348 Louvain-la-Neuve, Belgium.

Dipartimento di Matematica e Informatica, Università degli Studi di Palermo, Via Archirafi 34, 90123 Palermo, Italy.

References

  1. [1]M. Abad, D. N. Castaño and J. P. Díaz Varela, MV-closures of Wajsberg hoops and applications, Algebra Universalis 64 (2010), no. 1–2, 213–230. DOI: 10.1007/s00012-010-0101-4DOI
  2. [2]P. Aglianò, I. M. A. Ferreirim and F. Montagna, Basic hoops: an algebraic study of continuous t-norms, Studia Logica 87 (2007), no. 1, 73–98. DOI: 10.1007/s11225-007-9078-1
  3. [3]S. Aguzzoli, B. Gerla and V. Marra, Gödel algebras free over finite distributive lattices, Annals of Pure and Applied Logic 155 (2008), no. 3, 183–193. DOI: 10.1016/j.apal.2008.04.003DOI
  4. [4]M. Barr and C. Wells, Toposes, triples and theories, Reprints in Theory and Applications of Categories 12 (2005), 1–287.
  5. [5]W. J. Blok and I. M. A. Ferreirim, On the structure of hoops, Algebra Universalis 43 (2000), no. 2–3, 233–257. DOI: 10.1007/s000120050156DOI
  6. [6]B. Bosbach, Komplementäre Halbgruppen. Axiomatik und Arithmetik, Fundamenta Mathematicae 64 (1969), no. 3, 257–287. DOI: 10.4064/fm-64-3-257-287
  7. [7]D. Bourn and G. Janelidze, Characterization of protomodular varieties of universal algebras, Theory and Applications of Categories 11 (2003), no. 6, 143–147.DOI
  8. [8]L. Carai, Free algebras and coproducts in varieties of Gödel algebras, The Journal of Symbolic Logic (2026), 1–32. DOI: 10.1017/jsl.2026.10194DOI
  9. [9]C. C. Chang, Algebraic analysis of many valued logics, Transactions of the American Mathematical Society 88 (1958), no. 2, 467–490. DOI: 10.1090/S0002-9947-1958-0094302-9DOI
  10. [10]R. Cignoli, F. Esteva, L. Godo and A. Torrens, Basic fuzzy logic is the logic of continuous t-norms and their residua, Soft Computing 4 (2000), no. 2, 106–112. DOI: 10.1007/s005000000044DOI
  11. [11]B. A. Davey, Dualities for equational classes of Brouwerian algebras and Heyting algebras, Transactions of the American Mathematical Society 221 (1976), no. 1, 119–146. DOI: 10.1090/S0002-9947-1976-0412063-9DOI
  12. [12]W. Fussner and N. Galatos, Categories of models of R-mingle, Annals of Pure and Applied Logic 170 (2019), no. 10, 1188–1242. DOI: 10.1016/j.apal.2019.05.003DOI
  13. [13]V. Giustarini, F. Manfucci and S. Ugolini, Free constructions in hoops via ℓ-groups, Studia Logica 113 (2025), no. 5, 1317–1365. DOI: 10.1007/s11225-024-10128-yDOI
  14. [14]P. Hájek, Metamathematics of Fuzzy Logic, Trends in Logic, Vol. 4, Kluwer Academic Publishers, 1998. DOI: 10.1007/978-94-011-5300-3
  15. [15]P. Hájek, L. Godo and F. Esteva, A complete many-valued logic with product-conjunction, Archive for Mathematical Logic 35 (1996), no. 3, 191–208. DOI: 10.1007/BF0268618DOI
  16. [16]A. Horn, Free L-algebras, Journal of Symbolic Logic 34 (1969), no. 3, 475–480. DOI: 10.2307/2270910DOI
  17. [17]G. Janelidze, Ideally exact categories, Theory and Applications of Categories 41 (2024), no. 11, 414–425.arxiv.org/abs/2308.06574
  18. [18]G. Janelidze, L. Márki and W. Tholen, Semi-abelian categories, Journal of Pure and Applied Algebra 168 (2002), no. 2–3, 367–386. DOI: 10.1016/S0022-4049(01)00103-7DOI
  19. [19]S. Lapenta, G. Metere and L. Spada, Relative ideals in homological categories with an application to MV-algebras, Theory and Applications of Categories 41 (2024), no. 27, 878–893.arxiv.org/abs/2208.12597
  20. [20]M. Mancini, G. Metere and F. Piazza, Coherent and ideal actions in ideally exact categories, Theory and Applications of Categories 45 (2026), no. 31, 1280–1320.arxiv.org/abs/2507.06124
  21. [21]M. Mancini, G. Metere and F. Piazza, On actions and split extensions in varieties of hoops: the case of strong section, Studia Logica (2026). DOI: 10.1007/s11225-026-10243-yDOI
  22. [22]M. Mancini, G. Metere, F. Piazza and M. E. Tabacchi, On split extensions of product hoops, in: 2025 IEEE International Conference on Fuzzy Systems (FUZZ), Reims, France, 2025, 1–5. DOI: 10.1109/FUZZ62266.2025.11152177DOI
  23. [23]M. Mancini and F. Piazza, Ideally exact categories, varieties of universal algebras and multi-valued logics, in: B. Vantaggi et al. (eds.), Information Processing and Management of Uncertainty in Knowledge-Based Systems (IPMU 2026), Communications in Computer and Information Science 3020, Springer, Cham, 2026, 262–276. DOI: 10.1007/978-3-032-28997-1_19DOI

Paper details

vlnc:W_s9rfedqh
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 1

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Claude Opus 5.5, GPT-6 Sol

View statement
Looking for related papers…
Contents