Hemi-Nelson algebras

It follows from a straightforward computation.

Let A\in \mathsf{hIL}. We will show that \mathrm{K}(A) satisfies the conditions of Definition 1. In order to prove it, let x=(a,b), y = (c,d) and z = (e,f) be elements of \mathrm{K}(A).

It is immediate that x\rightarrow x = (1,0), so (hN1) is satisfied.

In order to show (hN2), note that

Since a\wedge (a\rightarrow c) \leq a\wedge c \leq a\wedge (b\vee c) then

so (hN2) is satisfied.

A direct computation shows that

But

so

which is (hN3).

Condition (hN4)

follows from the equality

and the fact that a\wedge d\wedge (a\rightarrow c) \leq c\wedge d = 0, i.e., a\wedge d\wedge (a\rightarrow c) = 0.

Now we will show (hN5) and (hN6).

Straightforward computations show that

Besides, a\wedge (a\rightarrow c)\leq a\wedge c\wedge (a\rightarrow c), so

Moreover, taking into account the distributivity of the underlying lattice of A and the facts that a\wedge b = c\wedge d = 0 and a\wedge (a\rightarrow c)\leq c, we get

Thus, a direct computation based on equations (?),…,(?) shows that conditions (hN5) and (hN6)

are satisfied.

Suppose that x\rightarrow y = 1, y\rightarrow x = 1, y\rightarrow z = 1 and z\rightarrow y = 1, so a\rightarrow c = 1, a\wedge d = 0, c\rightarrow a = 1, c\wedge b = 0, c\rightarrow e = 1, c\wedge f = 0, e\rightarrow c = 1 and e\wedge d = 0. In particular, it follows from Lemma 3 that a = c = e, so

However, a\wedge f = e\wedge f = 0. Hence, x\rightarrow z = (1,0). Analogously, we have that z\rightarrow x = (1,0). Thus, (hN7)

is satisfied.

Finally, we will show (hN8) and (hN9).

Suppose that x\rightarrow y = 1 and y\rightarrow x = 1, so a = c, a\wedge d = 0 and c\wedge b = 0. Besides,

which implies that

Since a\wedge e\wedge (d\vee f) = (a\wedge e\wedge d) \vee (a\wedge e \wedge f) and a = c then a\wedge e\wedge (d\vee f) = 0, so

which is (hN8).

In order to see (hN9),

note that

so

Taking into account that (a\vee e)\wedge d\wedge f = (a \wedge d \wedge f) \vee (e\wedge d\wedge f), a = c, a\wedge c = 0 and e\wedge f = 0 we get

Therefore,

which was our aim.

First we will show that condition 1) is satisfied. By (hN2),

1\rightarrow x \leq x.

Now we will show that 2) is verified. Suppose that x\rightarrow y = 1, so by (hN2)

we get x\leq (\sim x\vee y), i.e.,

Finally we will see 3) is satisfied. Assume that x\rightarrow y = 1 and \sim y\rightarrow \sim x = 1. It follows from 2. of this proposition that

so

In particular,

so

Therefore, x\leq y.

The reflexivity and symmetry of \theta are immediate. The transitivity of the relation follows from (hN7).

Thus, \theta is an equivalence relation. The facts that \theta is compatible with \wedge, \vee and \rightarrow follow from (hN8), (hN9) and (hN10).

Let T\in \mathsf{KhIL} and x,y\in T. It follows from (hN1) that

The inequality

follows from (hN5) and (hN6).

Therefore, T/\theta \in \mathsf{hIL}.

We write \rho instead of \rho_T. First we will show that \rho is a well defined map. Let x\in T. Note that x/\theta \wedge \sim x/\theta = 0/\theta if and only if (x\wedge \sim x) \rightarrow 0 = 1 and 0\rightarrow (x\wedge \sim x) = 1. First note that it follows from (hN1) and (hN4)

that

Besides, it follows from (hN1) and (hN3)

that

Thus, \rho is a well defined map.

Straightforward computations show that \rho preserves \wedge, \vee, \sim, 0 and 1. The fact that \rho preserves \rightarrow follows from (hN3) and (hN4).

Hence, \rho is a homomorphism. Finally, in order to show that \rho is an injective map let x,y \in T such that \rho(x) = \rho(y), i.e., x\rightarrow y= 1, y\rightarrow x = 1, \sim x \rightarrow \sim y = 1 and \sim y \rightarrow \sim x = 1. Thus, it follows from Lemma 7 that x = y. Hence, \rho is an injective map. Therefore, \rho is a monomorphism.

Let A\in \mathsf{hIL}. We show that equation (12) holds in \mathrm{K}(A). Take x=(a,b), y=(c,d) and z=(e,f) in \mathrm{K}(A). Then

and

It is straightforward to verify that

Thus equation (12) holds in \mathrm{K}(A).

Finally, by Theorem 11, every T\in \mathsf{KhIL} is isomorphic to a subalgebra of \mathrm{K}(A) for some A\in \mathsf{hIL}, and hence equation (12) holds in T as well.

Suppose that 1 = x \vee (z \wedge \sim z). It follows that 1\leq x \vee \sim z, and by De Morgan’s laws we obtain \sim x \wedge z = 0. Since z \wedge \sim z \leq x \vee \sim x, we have x \vee (z \wedge \sim z) \leq x \vee \sim x, which implies x \vee \sim x = 1. Therefore,

Hence z \leq x, and consequently 1 = x \vee z = x. The converse implication is immediate and the result follows.

Suppose that x \rightarrow y = 1 and y \rightarrow x = 1. From Proposition 7 we obtain x = x \wedge (\sim x \vee y) and y = y \wedge (\sim y \vee x). Let z = (x \wedge \sim x) \vee (y \wedge \sim y). Then z \leq \sim z and z \wedge \sim z = z. Hence,

Since y \leq \sim y \vee x, we have x \vee y \leq \sim y \vee x, and therefore x \vee (z \wedge \sim z) = x \vee y. By the same reasoning, y \vee (z \wedge \sim z) = x \vee y, and thus, y \vee (z \wedge \sim z) = x \vee (z \wedge \sim z).

Now, suppose that there exists z\in T such that x\vee (z\wedge\sim z)=y\vee (z\wedge\sim z). By (hN1) and (12), 1=x\vee (z\wedge\sim z)\rightarrow y\vee (z\wedge\sim z)=(x\rightarrow y)\vee (z\wedge\sim z). By Lemma 13, x\rightarrow y=1. By the same reasoning we obtain y\rightarrow x=1.

We will prove that \mathsf{KhIL} is precisely the class of algebras defined by the equations of Kleene algebras together with conditions (hN1)–(hN6) and equation (12).

On the one hand, from Proposition 12 it follows that every algebra in \mathsf{KhIL} satisfies equation (12).

On the other hand, assume that T satisfies conditions (hN1)–(hN6) and equation (12). From Proposition 14 it follows immediately that T satisfies conditions (hN8), (hN9) and (hN7), since \theta_{T^-} is a lattice congruence of T.

To prove condition (hN10), suppose that x\rightarrow y=y\rightarrow x=1. By Proposition 14, there exists w\in T such that x\vee (w\wedge\sim w)=y\vee (w\wedge\sim w). Then,

and by equation (12), it follows that

Again, by Proposition 14, (z\rightarrow x)\rightarrow (z\rightarrow y)=1 and (z\rightarrow y)\rightarrow (z\rightarrow x)=1. Analogously, it is proved that (y\rightarrow z)\rightarrow (x\rightarrow z)=1 and (x\rightarrow z)\rightarrow (y\rightarrow z)=1. Therefore, T\in \mathsf{KhIL} and the result follows.

Let \theta \in \mathrm{Con}(T). Note that 1/\theta is an upset because \theta is in particular a lattice congruence. In order to show that 1/\theta is an implicative filter, let x,y\in T such that x, x\rightarrow y\in 1/\theta. Then x \wedge (x\rightarrow y)\in 1/\theta. Since 1/\theta is an upset then it follows from (hN2)

that x \wedge (\sim x \vee y) \in 1/\theta. Moreover, since x/\theta = 1/\theta then \sim x/\theta = 0/\theta, so y/\theta = 1/\theta, i.e., y\in 1/\theta. Hence, 1/\theta is an implicative filter. The fact that 1/\theta is an h-implicative filter follows from a direct computation.

Let \theta \in \mathrm{Con}(T) and x,y \in T. It is immediate that if (x,y)\in \theta then x\rightarrow y, \sim y \rightarrow \sim x, y\rightarrow x, \sim x \rightarrow \sim y \in 1/\theta. Conversely, assume that x\rightarrow y, \sim y \rightarrow \sim x, y\rightarrow x, \sim x \rightarrow \sim y \in 1/\theta. Since T/\theta \in \mathsf{KhIL} then it follows from Proposition 7 that x/\theta = y/\theta, i.e., (x,y)\in \theta.

Let \theta,\psi \in \mathrm{Con}(T). It is immediate that if \theta\subseteq \psi then 1/\theta \subseteq 1/\psi. The converse follows from a direct computation based in Lemma 20.

It follows from the definition of implicative filter that 1\in F. From F2), in Remark 18, it follows that F is an upset. Now, let x\in T. In order to show that x\in F if and only if 1\rightarrow x\in F, suppose that x\in F. Then

so 1\rightarrow x\in F. The converse follows from Proposition 7.

Finally, in order to show that F is closed by \wedge, let x, y \in F. Then 1\rightarrow y\in F. Moreover,

because x\in F, so since 1\rightarrow y \in F then x\rightarrow (y\wedge x)\in F. Thus, y\wedge x\in F. Hence, F is a filter.

By Theorem 11 without loss of generality we can assume that T = \mathrm{K}(A) for some A\in \mathsf{hIL}. Let x = (a,b), y = (c,d) and z = (e,f) be elements of \mathrm{K}(A). A direct computation shows that

so

In order to show 1) note that

Since a \wedge (a\leftrightarrow c) = c\wedge (a\leftrightarrow c), a direct computation proves that

The conditions 2) follows from 1) of this lemma and (hN8).

In order to show the condition 3), note that it follows from 1) of this lemma and (hN8)

that

Hence, it follows from (hN7)

that

The condition 4) follows from 3) of this lemma and (hN8).

The condition 5) follows from 1) of this lemma, the distributivity of the underlying lattice of T and (hN9).

In order to show 6), note that

Given (u,v) \in T we define \pi_1(u,v) = u and \pi_2(u,v) = v. It is immediate that \pi_1(x\wedge s(x,y)) = \pi_1(y\wedge s(x,y)), so

Besides, a direct computation shows that

Since a\wedge (a\leftrightarrow c) = c\wedge (a\leftrightarrow c) then

Thus, we have proved 6). A similar argument shows 7).

The conditions 8) and 9) follow from 1) of this lemma and (hN10).

Condition 10) follows from a direct computation.

Let F\in \mathrm{hIF}(T). First we will show that \Theta(F) is an equivalence relation. It is immediate that \Theta(F) is a reflexive and symmetric relation. In order to show the transitivity, let x,y,z\in T such that (x,y), (y,z) \in \Theta(F), i.e., s(x,y), s(y,z) \in F. Let f = s(x,y)\wedge s(y,z). It follows from Lemma 22 that F is a filter, so f\in F. Besides it follows from Lemma 23 that (x\wedge f) \rightarrow (z\wedge f) = 1. But ((x\wedge f) \rightarrow (z\wedge f)) \rightarrow (x\rightarrow z)\in F, so x\rightarrow z \in F. Similarly we have that z\rightarrow x\in, \sim x \rightarrow \sim z, \sim z\rightarrow \sim x \in F. Thus, s(x,z)\in F, i.e., (x,z)\in \Theta(F). Hence, \Theta(F) is a transitive relation. In consequence, \Theta(F) is an equivalence relation.

Finally, let x\in T. Then x\in 1/\Theta(F) if and only if x\rightarrow 1, 1\rightarrow x, 0\rightarrow \sim x and \sim x\rightarrow 0 \in F. Suppose that x\rightarrow 1, 1\rightarrow x, 0\rightarrow \sim x and \sim x\rightarrow 0 \in F. Since 1\rightarrow x\in F then it follows from Lemma 22 that x\in F. Conversely, assume that x\in F. It follows from Lemma 22 that 1\rightarrow x\in F. Besides, ((x\wedge x)\rightarrow (1\wedge x))\rightarrow (x\rightarrow 1) \in F, i.e., 1\rightarrow (x\rightarrow 1) \in F, so x\rightarrow 1 \in F. Moreover, it follows from Lemma 23 that 0\rightarrow (\sim x \wedge x) = 1. Since ((0\wedge x) \rightarrow (\sim x \wedge x)) \rightarrow (0\rightarrow \sim x)\in F then 0\rightarrow \sim x \in F. Finally, since (\sim x \wedge x) \rightarrow 0 = 1 and ((\sim x \wedge x)\rightarrow (0\wedge x))\rightarrow (\sim x \rightarrow 0) \in F then \sim x \rightarrow 0 \in F. Therefore, 1/\Theta(F) = F.

First we will see that \Theta(F) preserves \wedge. Let x,y,z\in T such that (x,y) \in \Theta(F), i.e., s(x,y)\in F. It follows from Lemma 23 that

Thus, it follows from (hN8)

that

But

so (x\wedge z) \rightarrow (y\wedge z) \in F. Similarly, (y\wedge z)\rightarrow (x\wedge z)\in F. In order to show that \sim(x\wedge z) \rightarrow \sim(y\wedge z) \in F, first note that

so it follows from (hN9)

that

i.e.,

Hence, \sim (x\wedge z) \rightarrow \sim(y\wedge z) \in F. Analogously, we get \sim (y\wedge z) \rightarrow \sim(x\wedge z) \in F. Thus, (x\wedge z,y\wedge z) \in \Theta(F).

It follows from that s(x,y) = s(\sim x, \sim y) the fact that \Theta(F) preserves \sim. Since we have also proved that \Theta(F) preserves \wedge and in particular T is a Kleene algebra, then \Theta(F) preserves \vee.

In what follows we will show that \Theta(F) preserves the implication. Let x,y,z\in T such that (x,y)\in \Theta(F), i.e., s(x,y) \in F. Then

are elements of \Theta(F). Since it follows from Lemma 23 that

then the transitivity of \Theta(F) implies that (x\rightarrow z,y\rightarrow z) \in \Theta(F). A similar argument based on Lemma 23 shows that (z\rightarrow x,z\rightarrow y)\in \Theta(F). Hence, \Theta(F) preserves the implication.

Then, T/\Theta(F) \in \mathsf{KhIL}. Therefore, \Theta(F) \in \mathrm{Con}(T).

Let T\in \mathsf{KhIL}. Let us take F an h-implicative filter of T and x, x\rightarrow_N y \in F. In particular, x\rightarrow (x\wedge y) \in F. Then it follows from F2) that ((x\wedge x) \rightarrow (y\wedge x)) \rightarrow (x\rightarrow y)\in F, i.e., (x\rightarrow (x\wedge y)) \rightarrow (x\rightarrow y) \in F. Thus, taking into account that F is an implicative filter and x, x\rightarrow (x\wedge y) \in F we get y\in F. Therefore, F is an N-implicative filter of T.

It follows from Remark 32 that without loss of generality we can assume that T = \mathrm{K}(A) for some A\in \mathsf{hIL}. Let x = (a,b) and y = (d,e) for some a,b,d,e \in A such that a\wedge b = d\wedge e = 0.

In order to show 1), suppose that x\rightarrow y = 1, so a\leq d, i.e., x\vee \mathsf{c} \leq y\vee \mathsf{c}. Now we will show 2). Suppose that x\vee \mathsf{c} = y\vee \mathsf{c}, i.e., a = d. Then x\rightarrow y = (a\rightarrow d,a\wedge e) =(a\rightarrow a, d\wedge e) = (1,0), which is 2). Now, assume that (x\wedge y) \rightarrow 0 = 1, so (a\wedge d) \rightarrow 0 =1. Hence, a\wedge d = 0, which implies that 0\rightarrow (x\wedge y) = 1 and x\wedge y \leq \mathsf{c}. Finally we will prove 4). A direct computation shows that (0,0) \rightarrow (0,1) = (1,0).

Suppose that \rho is surjective. Let x, y \in T such that (x\wedge y)\rightarrow 0 = 1, so it follows from Lemma 33 that 0\rightarrow (x\wedge y) = 1. Thus, x/\theta \wedge y/\theta = 0/\theta. Since \rho is surjective, there exists z\in T such that z/\theta = x/\theta and \sim z/\theta = y/\theta, i.e., z\rightarrow x = 1, x\rightarrow z = 1, \sim z \rightarrow y = 1, y \rightarrow \sim z = 1. Applying again Lemma 33 we get z\vee \mathsf{c} = x\vee \mathsf{c} and \sim z \vee \mathsf{c} = y\vee \mathsf{c}.

Conversely, suppose that \mathsf{(C)} is satisfied and let x,y \in T such that x/\theta \wedge y/\theta = 0/\theta, so (x\wedge y)\rightarrow 0 = 1. It follows from hypothesis that there exists z\in T such that z\vee \mathsf{c} = x\vee \mathsf{c} and \sim z \vee \mathsf{c} = y\vee \mathsf{c}. Then it follows from Lemma 33 that z\rightarrow x = 1, x\rightarrow z = 1, \sim z \rightarrow y = 1 and y\rightarrow \sim z = 1. Then \rho(z) = (x/\theta, y/\theta). Therefore, \rho is surjective.

Suppose that T satisfies \mathsf{(CK)} and let x, y\in T such that (x\wedge y)\rightarrow 0 = 1. It follows from Lemma 33 that x\wedge y \leq \mathsf{c}. Let \hat{x} = x\vee \mathsf{c} and \hat{y} = y\vee \mathsf{c}. Then \hat{x}, \hat{y}\geq \mathsf{c} and \hat{x}\wedge \hat{y} \leq \mathsf{c}, so it follows from hypothesis that there exists z\in T such that z\vee \mathsf{c} = \hat{x} and \sim z \vee \mathsf{c} = \hat{y}, i.e., z\vee \mathsf{c} = x\vee \mathsf{c} and \sim z \vee \mathsf{c} = y\vee \mathsf{c}.

Conversely, suppose that \mathsf{(C)} is satisfied. Let x,y\in T such that x, y\geq \mathsf{c} and x\wedge y\leq \mathsf{c}. It follows that x\wedge y=\mathsf{c}. It follows from Lemma 33 that (x\wedge y)\rightarrow 0 = 1. Thus, it follows from hypothesis that there exists z\in T such that z\vee \mathsf{c} = x\vee \mathsf{c} and \sim z \vee \mathsf{c} = y\vee \mathsf{c}, i.e., z\vee \mathsf{c} = x and \sim z \vee \mathsf{c} = y.

Suppose that T \cong \mathrm{K}(A) for some A\in \mathsf{hIL}. Let x, y \in \mathrm{K}(A) such that x, y \geq \mathsf{c} and x\wedge y \leq \mathsf{c}. Thus, there exist a,b \in A such that x = (a,0), y = (b,0) and a\wedge b=0. The element z = (a,b) \in \mathrm{K}(A) satisfies that z\vee \mathsf{c} = x and \sim z \vee \mathsf{c} = y. Hence, \mathrm{K}(A) satisfies \mathsf{(CK)}. Since this condition is preserved by isomorphisms then T satisfies \mathsf{(CK)}. Furthermore, the fact that \mathrm{K}(A) has center, which is (0,0), implies that T has center. The converse follows from Lemma 34, Lemma 35 and Remark 32.

We write \alpha in place of \alpha_{A}. First we will show that \alpha is a well defined map. Let a\in A. Then a\wedge (a \rightarrow 0) \leq 0, i.e., a\wedge (a\rightarrow 0) = 0. Thus, (a, a\rightarrow 0)/\theta \in \mathrm{K}(A)/\theta. Let a\in A. Then (a,a\rightarrow 0) /\theta = (a,0)/\theta. It is immediate that \alpha is a homomorphism. The injectivity of \alpha is also immediate. In order to show that \alpha is suryective, let y\in \mathsf{C}(\mathrm{K}(A)), so y = (a,b)/\theta for some a, b\in A such that a \wedge b = 0. Moreover, y = (a, a\rightarrow 0)/\theta, so y = \alpha(a). Thus, \alpha is suryective. Therefore, \alpha is an isomorphism.

Let A\in \mathsf{hIL}. It is immediate that if A\in \mathsf{SH} then \mathrm{K}(A) \in \mathsf{SN}. Conversely, suppose that \mathrm{K}(A) \in \mathsf{SN}, so \mathrm{K}(A)/\theta \in \mathsf{SH}, where \theta is the equivalence relation considered in Lemma 9. Then, since it follows from Lemma 37 that A is isomorphic to \mathrm{K}(A)/\theta then A\in \mathsf{SH}, which was our aim.

Let (a,b), (c,d)\in \mathrm{K}(A). Note that (a,b)\rightarrow_N (c,d) = (a\rightarrow (a\wedge c),a\wedge d). Thus, (a,b)\rightarrow_N (c,d) = (1,0) if and only if a\rightarrow (a\wedge c) = 1 and a\wedge d = 0. Taking into account that a\wedge (a\rightarrow (a\wedge c))= a\wedge c and c\wedge d = 0, we have that (a,b)\rightarrow_N(c,d) = (1,0) if and only if a\leq c. Hence, condition 1) is satisfied. Conditions 2) and 3) follows from [SM2, Lemma 9] and the fact that the equation 1\rightarrow x = x is satisfied in every semi-Heyting algebra.

Without loss of generality, we can assume that T = \mathrm{K}(A) for some A\in \mathsf{SH}. Conditions 1) and 2) follow from Lemma 44. The proofs of 3) and 4) follow from a direct computation based on the definition of semi-Heyting algebras.

Let T\in \mathsf{SN}. It follows from Proposition 31 that every h-implicative filter of T is an N-implicative filter of T. The fact that every N-implicative filter of T is an h-implicative filter of T follows from a direct computation based on Remark 30 and Lemma 45.

It is similar to the proof of Proposition 42.

Let T \in \mathsf{SNA}. Without loss of generality we can assume that T takes the form \mathrm{K}(A) for some A\in \mathsf{SRL}. In order to show 1), it is enough to show that for every a,b,c \in A, a\rightarrow (a\wedge c) \leq ((a\wedge c) \rightarrow (b\wedge c)) \rightarrow (a\rightarrow b) (In [LMSM, Remark 3.1] it was shown that if A\in \mathsf{SRL} and (a,b), (c,d)\in \mathrm{K}(A), then (a,b)\rightarrow (c,d) = (1,0) if and only if a\leq c.). The last equality holds if and only if (a\rightarrow (a\wedge c)) \wedge ((a\wedge c) \rightarrow (b\wedge c)) \leq a\rightarrow b. This follows from the fact that a\rightarrow (b\wedge c)\leq a\rightarrow b.

That conditions 2), 3) and 4) are satisfied follows from a direct computation.

Let T\in \mathsf{SNA}. It follows from Proposition 28 that every h-implicative filter of T is an open implicative filter of T. The fact that every open implicative filter of T satisfies F1) and F4) follows from a direct computation based on Lemma 52.

Let F be an open implicative filter of T. To prove that it satisfies F2) suppose that z\in F. Then, x\rightarrow (x\wedge z)=(x\rightarrow x)\wedge (x\rightarrow z)=x\rightarrow z. Since x\leq 1, we obtain 1\rightarrow z\leq x\rightarrow z and from the fact that 1\rightarrow z\in F, we obtain x\rightarrow z\in F. From Lemma 52, (x\rightarrow (x\wedge z)) \rightarrow (((x\wedge z) \rightarrow (y\wedge z)) \rightarrow (x\rightarrow y)) \in F and thus ((x\wedge z) \rightarrow (y\wedge z)) \rightarrow (x\rightarrow y) \in F. To prove that F satisfies F3), suppose that z\in F. A direct computation shows that ((x\wedge \sim y)\rightarrow (x\wedge \sim y \wedge z))\rightarrow( (x\wedge\sim y)\rightarrow \sim ((x\wedge z)\rightarrow y))=1. Using an analogous argument, we obtain x\rightarrow (x\wedge z)\in F and since x\rightarrow (x\wedge z)\leq (x\wedge \sim y)\rightarrow (x\wedge \sim y\wedge z), it follows that (x\wedge \sim y)\rightarrow (x\wedge \sim y\wedge z)\in F. Finally, by Lemma 52, \sim (x\rightarrow y) \rightarrow \sim ((x\wedge z)\rightarrow (y\wedge z))\in F.

Therefore, every open implicative filter of T is an h-implicative filter of T.