We present a category equivalent to that of semi-Nelson algebras. The objects in this category are pairs consisting of a semi-Heyting algebra and one of its filters. The filters must contain all the dense elements of the semi-Heyting algebra and satisfy an additional technical condition. We also show that in the case of dually hemimorphic semi-Nelson algebras, the filters are not necessary and the category is equivalent to that of dually hemimorphic semi-Heyting algebras.
A semi-Heyting algebra is an algebra \mathbf{H}=\langle H; \land,\lor,\Rightarrow,0,1 \rangle of type (2,2,2,0,0) such that the following conditions are satisfied for all x,y,z\in A:
\langle H,\land,\lor,0,1 \rangle is a bounded lattice,
x\land(x\Rightarrow y)=x\land y,
x\land(y\Rightarrow z)=x\land((x\land y)\Rightarrow(x\land z)),
x\Rightarrow x=1.
A semi-Heyting algebra is a Heyting algebra if it satisfies the identity:
(H) (x\land y)\Rightarrow x=1.
A Nelson algebra is an algebra \mathbf{A}=\langle A; \land,\lor,\to,\sim,1 \rangle of type (2,2,2,1,0) such that the following conditions are satisfied for all x,y,z\in A:
x\land(x\lor y)=x,
x\land(y\lor z)=(z\land x)\lor(y\land x),
\sim\sim x=x,
\sim(x\land y)=\sim x\lor \sim y,
x\land \sim x=(x\land \sim x)\land (y\lor \sim y),
x\to x=1,
x\to(y\to z)=(x\land y)\to z,
x\land (x\to y)=x\land(\sim x\lor y).
A semi-Nelson algebra is an algebra \mathbf{A}=\langle A; \land, \lor,\to, \sim, 1 \rangle of type (2,2,2,1,0) such that the following conditions are satisfied for all x,y,z\in A:
x\land(x\lor y)=x,
x\land(y\lor z)=(z\land x) \lor(y\land x),
\sim\sim x=x,
\sim(x\land y)=\sim x\lor\sim y,
x\land\sim x=(x\land\sim x)\lor(y\lor\sim y),
x\land(x\mbox{ _N } y)=x\land(\sim x\lor y),
x\mbox{ _N }(y\mbox{ _N } z)=(x\land y)\mbox{ _N } z,
(x\mbox{ _N } y)\mbox{ _N }[(y\mbox{ _N } x)\mbox{ _N }[(x\to z)\mbox{ _N }(y\to z)]]=1,
(x\mbox{ _N } y)\mbox{ _N }[(y\to_N x)\mbox{ _N }[(z\to x)\mbox{ _N }(z\to y)]]=1,
(\sim(x\to y))\mbox{ _N }(x\land\sim y) = 1,
(x\land\sim y)\mbox{ _N }(\sim(x\to y)) = 1,
where x \mbox{ _N } y := x \to (x \land y).
An algebra \mathbf{A}=\langle A; \land,\lor,^*,0,1\rangle is a pseudocomplemented lattice if the following conditions hold:
\langle A; \land,\lor,0,1\rangle is a bounded lattice,
x\land(x\land y)^*=x\land y^*,
0^*=1 and 1^*=0.
If \mathbf A=\langle A; \land,\lor,^*,0,1\rangle is a pseudocomplemented lattice, then for every x, y \in A, if x\land y=0, then x\leq y^*. It also holds that (x\lor y)^*=x^*\land y^*.
Let \mathbf{A}=\langle A; \land,\lor,^*,0,1\rangle be a pseudocomplemented lattice. We say that an element a\in A is dense if a^*=0. We denote by \mathcal{D}_s(A) the set of all dense elements of A.
If \mathbf H\in\mbox{SH}, then x\in\mathcal{D}_s(H) if and only if x=y\lor y^* for some y\in H.
An algebra \mathbf{A}=\langle A; \land,\lor,\to,^\dagger,0,1 \rangle of type (2,2,2,1,0,0) is said to be a dually hemimorphic semi-Heyting algebra if \langle A; \land,\lor,\to,0,1 \rangle is a semi-Heyting algebra and the following equations are satisfied:
0^\dagger=1,
1^\dagger=0,
(x\land y)^\dagger=x^\dagger\lor y^\dagger.
An algebra \mathbf{A}=\langle A; \land,\lor,\to,\sim,',1 \rangle of type (2,2,2,1,0) is said to be a dually hemimorphic semi-Nelson algebra if \langle A; \land,\lor,\to,\sim,1 \rangle is a semi-Nelson algebra and the following equations are satisfied:
(\sim 1)'=1,
1'\to(\sim1)=1,
((x\to y)\land(y\to x)\land x')\to((x\to y)\land(y\to x)\land y)'=1,
\sim x'\to(\sim x\land(x'\to x))=1,
(\sim x\land(x'\to x))\to\sim x'=1,
(x\land y)'\to(x'\lor y')=1,
(x'\lor y')\to(x\land y)'=1.
We use the convention that the unary operation ' has higher priority than \sim, so the expression \sim x' means \sim(x').
From the previous definitions we can deduce the rule
[cornejo16semiNelson, Lemma 2.7 (h) and (l)] Let \mathbf{A}\in\mbox{SN}, and a,b,c\in A. Then the following properties hold:
a\mbox{ _N } b=b\mbox{ _N } a=1 if and only if a\to b=b\to a=1,
If a\mbox{ _N } b=1=b\mbox{ _N } c=1, then a\mbox{ _N } c=1.
([cornejo16semiNelson, Theorem 5.3 and Corollary 5.2]) If \mathbf{A}\in\mbox{SH}, then \mathbf{A} is isomorphic to \mathbf{sH}(\mathbf{V_k}(\mathbf A)) through the isomorphism g(a)=[\![(a,a^*)]\!]. Also, if \mathbf{A}\in\mbox{SN}, then \mathbf{A} is isomorphic to a subalgebra of \mathbf{V_k(sH(\mathbf A))} through the application h(a)=([\![a]\!],[\![\sim a]\!]).
Let \mathbf{H}\in\mbox{SH}. A subset F\subseteq H is an i-filter, if the following conditions hold:
F is a lattice filter of \mathbf H.
\mathcal{D}_s(H)\subseteq F.
If z\lor t\in F, then for all x\in H, (x\Rightarrow z)\lor(x\land t)\in F.
Consider the semi-Heyting algebra \mathbf{A}=\langle A; \land,\lor, \Rightarrow,1 \rangle, with the operation \Rightarrow indicated in the table:
1mm
The subset F=\{1\} of \mathbf A is a filter, and it satisfies IF2) (since the only dense element is 1). But F does not satisfy IF3) because 1\lor 0=1\in F, but (0\Rightarrow 1)\lor(0\land0)=0\lor0=0\notin F.
If we consider the structure \mathbf{ N(H,F)}, with \mathbf{H}\in\mbox{SH}, an i-filter F, and (a,b),(c,d)\in N(H,F), then (1) c\land d=0 and (2) c\lor d\in F. Notice that
Also, using (2) and IF3) we have that (a\Rightarrow c)\lor(a\land d)\in F, so the pair (a\Rightarrow c,a\land d) is in N(H,F) and we can use it to define (a,b)\to(c,d)=(a\Rightarrow c,a\land d).
If \mathbf{H}\in\mbox{SH} and F is an i-filter of \mathbf H, then the system \langle N(H,F); \sqcap,\sqcup,\to,\sim,\top \rangle with the operations defined as in (V1)-(V5) is a semi-Nelson algebra.
Using that F is a filter we can prove that the operations \sqcap, \sqcup, \top and \sim are well defined in N(H,F) (see [monteiro19construction]). Also, by observation 3.3, it follows that the operation \to is well defined. Furthermore, following the proof of Theorem 4.1 in[cornejo16semiNelson], \mathbf{N(H,F)} verifies the axioms SN1)-SN11) so it is is a semi-Nelson algebra.
Let \mathbf{H}\in\mbox{SH}. If \mathbf S is a subalgebra of \mathbf{V_k}(\mathbf{H}), such that \pi_1(S)=H (with \pi_1 the projection over the first component of {V_k}(H)), then there exists an i-filter E of \mathbf H such that S=N(H,E).
Let E=\{x\in H : x=a\lor b \text{ for some } (a,b)\in S\}. We are going to prove that E is an i-filter.
It was shown in [viglizzo99algebras] that if \mathbf H\in\mbox{H}, then E is a filter of \mathbf H. But that proof can be extended naturally to the variety \mbox{SH}, because it only uses lattice properties.
Consider a dense element x\in H. Since \pi_1(S)=H, then there exists y\in H such that (x,y)\in S. Besides, since S is a subalgebra of \mathbf{V_k(H)}, (0,1)\in S, and therefore (x,y)\to(0,1)=(x\Rightarrow0,x\land1)=(x^*,x)=(0,x)\in S. Hence, x\in E because x=x\lor0, and we conclude that E satisfies IF2).
Suppose now that for z,t\in H we have z\lor t\in E,
Since z\lor t\in E, there exists a pair (a,b)\in S such that z\lor t=a\lor b. But S is a subalgebra of \mathbf{V_k(H)}, so \sim(a,b)=(b,a)\in S, and therefore (a,b)\cup(b,a)=(a\lor b,a\land b)=(a\lor b,0)\in S. This means that (z\lor t,0)\in S, and hence
Using the hypothesis \pi_1(S)=H, and x,z,t\in H, it follows that there exist some x',z',t'\in H such that (x,x'), (z,z'), (t,t')\in S. This implies that the next term belongs to S:
Since (x\land z')\land(x'\lor t')=(x\land z'\land x')\lor(x\land z'\land t')=0\lor(x\land z'\land t')=x\land z'\land t', (?) yields
Combining (?) with (?) we get
Since x\land z'\land t'\land(z\lor t)=(x\land z'\land t'\land z)\land(x\land z'\land t'\land t)=(x\land t'\land(z\land z'))\lor(x\land z\land(t\land t'))=0\lor0=0, from (?) it follows that
By the definition of E it follows that
Hence, E is an i-filter.
Let us prove now that N(H,E)=S. It is clear that S\subseteq N(H,E). If we have (a,b)\in N(H,E), then a\land b=0, and a\lor b\in E, so there exists (c,d)\in S such that c\lor d=a\lor b. Then, as before (c,d)\cup(d,c)=(c\lor d,0)=(a\lor b,0)\in S. On the other hand, since \pi_1(S)=H, there exists some b'\in H such that (b,b')\in S, and therefore (b,b')\to(0,1)=(b^*,b)\in S. By properties of the pseudocomplement, a\land b=0 implies that a\leq b^* and hence (a\lor b)\land b^*=(a\land b^*)\lor(b\land b^*)=a\lor 0=a, so we can conclude that (a\lor b,0)\cap(b^*,b)=((a\lor b)\land b^*,0\lor b)=(a,b)\in S.
Every semi-Nelson algebra \mathbf{A} is isomorphic to an algebra of the form \mathbf{N(H,F)}, for some semi-Heyting algebra \mathbf{H} and some i-filter F of \mathbf{H}.
If \mathbf{H}\in\mbox{SH}, and F is an i-filter of \mathbf H, then \pi_1(N(H,F))=H.
The application \alpha:\mathbf{sHF}\longrightarrow\mathbf{sN} defined on the objects of \mathbf{sHF} by \alpha((\mathbf H,F))=\mathbf{N(H,F)}, and on morphisms by \alpha(f)(a,b)=(f(a),f(b)) for all (a,b)\in N(H,F), is a functor from \mathbf{sHF} to \mathbf{sN}.
Let us check that \alpha is well defined. If (\mathbf H,F) is an object of \mathbf{sHF}, then \alpha((\mathbf H,F))=\mathbf{N(H,F)} is a semi-Nelson algebra due to Lemma 3.4.
Now we take a morphism f:(\mathbf H,F)\longrightarrow(\mathbf H',F'). Since f is a Heyting homomorphism, we have by straightforward calculations that: \alpha(f)((a,b)\to(c,d))=(f(a\Rightarrow c),f(a\land d))=(f(a),f(b))\to((f(c),f(d)))= \alpha(f)(a,b)\to\alpha(f)(c,d), and similarly for the rest of the operations.
For all (a,b) \in N(H,F), (a,b)=(Id_{(H,F)}(a),Id_{(H,F)}(b))), therefore \alpha(Id_{(\mathbf H,F)})=Id_{\alpha(\mathbf H,F)}.
Finally, for any two given morphisms f:(\mathbf H',F')\longrightarrow(\mathbf H'',F''), and g:(\mathbf H,F)\longrightarrow(\mathbf H',F'), we have that for all (a,b)\in N(H,F), \alpha(f\circ g)(a,b)=((f\circ g)(a),(f\circ g)(b))=(f(g(a)),f(g(b)))=\alpha(f)((g(a),g(b)))=((\alpha(f))\circ(\alpha(g)))(a,b).
Let A\in\mbox{SN}. We say that an element a\in A is positive if \sim a\leq a. We denote by A^+ the set of positive elements of A.
If A\in\mbox{SN}, then A^+=\{x\lor\sim x : x\in A\}. Indeed, if x\in A^+, then \sim x\leq x, so x=x\lor\sim x. On the other hand, if y=x\lor\sim x, by SN?) and SN?) it follows that y=x\lor\sim x\geq x\land\sim x=\sim(\sim x\lor x)=\sim y.
If \mathbf{A}\in\mbox{SN}, then [\![A^+]\!] is an i-filter of \mathbf{sH}(A).
By Theorem 2.12, there exists a subalgebra \mathbf S of \mathbf{V_k}(\mathbf{sH}(A)) such that \mathbf A is isomorphic to \mathbf S, and the isomorphism is given by h(a)=([\![a]\!],[\![\sim a]\!]) for all a\in A. Therefore S=\{([\![a]\!],[\![\sim a]\!]) : a\in A\}. Then, it is immediate that \pi_1(S)=\mathbf{sH}(A).
By Lemma 3.5, S=N(\mathbf{sH}(A),E), where E=\{[\![a]\!]\lor[\![\sim a]\!] : a\in A\} is an i-filter of \mathbf{sH}(\mathbf A). But E=[\![A^+]\!], due to observation 3.10, which completes the proof.
The application \beta:\mathbf{sN}\longrightarrow\mathbf{sHF} defined on the objects of \mathbf{sN} by \beta(\mathbf{A})=(\mathbf{sH}(\mathbf A),[\![A^+]\!]), and on morphisms f:A\longrightarrow A' by \beta(f)([\![a]\!])=[\![f(a)]\!] for all a\in A, is a functor from \mathbf{sN} to \mathbf{sHF}.
Let us check that \beta is well defined. For an object \mathbf{A} of \mathbf{sN}, \mathbf{sH}(\mathbf{A}) is a semi-Heyting algebra, and by Lemma 3.11 we have that [\![A^+]\!] is an i-filter of \mathbf{sH}(\mathbf{A}). Therefore, (\mathbf{sH}(A),[\![A^+]\!]) is an object in \mathbf{sHF}.
For any semi-Nelson morphism f:A\longrightarrow A', we calculate: \beta(f)([\![a\to b]\!])= [\![f(a\to b)]\!]= [\![f(a)\to f(b)]\!]= [\![f(a)]\!]\Rightarrow[\![f(b)]\!]= \beta(f)([\![a]\!])\Rightarrow\beta(f)([\![b]\!]), and similarly for the other operations.
Also, if [\![x]\!]\in [\![A^+]\!], then by observation 3.10 we can write [\![x]\!]= [\![y\lor \sim y]\!], for some y\in A, and so \beta(f)([\![x]\!])=\beta(f)([\![y\lor\sim y]\!])=\beta(f)([\![y]\!])\cup\beta(f)([\![\sim y]\!])=[\![f(y)]\!]\cup[\![f(\sim y)]\!]= [\![f(y)]\!]\cup[\![\sim f(y)]\!]= [\![f(y)\lor\sim f(y)]\!]. Since f(y)\in f(A)\subseteq A', it follows that \beta(f)([\![x]\!])\in [\![A'^+]\!]. This shows that \beta(f)([\![A^+]\!])\subseteq [\![A'^+]\!], so \beta(f) is a morphism in \mathbf{sHF}.
It is straightforward to check that \beta preserves identities and compositions, so it is indeed a functor.
The functors \alpha and \beta establish an equivalence between the categories \mathbf{sHF} and \mathbf{sN}.
Notice that \beta\alpha(\mathbf{A},F)=\beta(\mathbf{N(\mathbf{A},F)})=(\mathbf{sH(N(\mathbf{A},F))},[\![N(A,F)^+]\!]), with \mathbf{sH(N(\mathbf{A},F))} =\{[\![(x,y)]\!] : (x,y)\in N(A,F)\}= \{[\![(x,y)]\!] : x,y\in A, x\land y=0, x\lor y\in F \}.
We define now \eta_A(x)=[\![(x,x^*)]\!], for each object \mathbf{A} of \mathbf{sHF} and for all x\in A, and we prove that \eta=\{\eta_A\} is a natural isomorphism from id_\mathbf{sHF} to \beta\alpha.
\eta_A is well defined: Since x\land x^*=0 and also \mathcal{D}_s(A)\subseteq F it follows that x\lor x^*\in \mathcal{D}_s(A)\subseteq F. Hence, (x,x^*)\in N(A,F).
\eta_A is surjective: If [\![(x,y)]\!]\in\beta\alpha(A,F), then [\![(x,y)]\!]=[\![(x,x^*)]\!]. Indeed, since (x,y)\to(x,x^*)=(x\Rightarrow x, x\land x^*)=(1,0) and (x,x^*)\to(x,y)=(x\Rightarrow x,x\land y)=(1,0), so by definition we have that [\![(x,y)]\!]=[\![(x,x^*)]\!]. Therefore \eta_A(x)=[\![(x,y)]\!].
\eta_A is injective: If [\![(x,x^*)]\!]=[\![(y,y^*)]\!], then (x,x^*)\to(y,y^*)=(x\Rightarrow y,x\land y^*)=(1,0). Therefore, from x\Rightarrow y=1, using properties of semi-Heyting algebras we obtain x\leq y. Analogously, (y,y^*)\to(x,x^*)= (y\Rightarrow x,y\land x^*)=(1,0) implies y\leq x.
\eta_A is a \mathbf{sHF}-morphism:
\eta_A(x)\cup\eta_A(y)=[\![(x,x^*)]\!]\cup[\![(y,y^*)]\!]=[\![(x,x^*)\sqcup(y,y^*)]\!]=[\![(x\lor y,x^*\land y^*)]\!]=[\![(x\lor y, (x\lor y)^*)]\!]=\eta_A(x\lor y).
\eta_A(x)\cap\eta_A(y)=[\![(x,x^*)]\!]\cap[\![(y,y^*)]\!]=[\![(x,x^*)\sqcap(y,y^*)]\!]=[\![(x\land y,x^*\lor y^*)]\!]. Also, \eta_A(x\land y)=[\![(x\land y,(x\land y)^*)]\!]. Since (x\land y,x^*\lor y^*)\to(x\land y,(x\land y)^*)=((x\land y)\Rightarrow(x\land y),(x\land y)\land(x\land y)^*)=(1,0) and (x\land y,(x\land y)^*)\to(x\land y,x^*\lor y^*)=((x\land y)\Rightarrow(x\land y),(x\land y)\land(x^*\lor y^*))=(1,(x\land y\land x^*)\lor(x\land y \land y^*))=(1,0), we have that \eta_A(x)\cap\eta_A(y)=\eta_A(x\land y).
\eta_A(x)\Rightarrow \eta_A(y)=[\![(x,x^*)]\!]\Rightarrow[\![(y,y^*)]\!]=[\![(x,x^*)\Rightarrow(y,y^*)]\!]=[\![(x\Rightarrow y,x\land y^*)]\!]. Also, \eta_A(x\Rightarrow y)=[\![(x\Rightarrow y,(x\Rightarrow y)^*)]\!]. Since (x\Rightarrow y,x\land y^*)\to(x\Rightarrow y,(x\Rightarrow y)^*)=((x\Rightarrow y)\Rightarrow(x\Rightarrow y),(x\Rightarrow y)\land(x\Rightarrow y)^*)=(1,0) and (x\Rightarrow y,(x\Rightarrow y)^*)\to(x\Rightarrow y,x\land y^*)=((x\Rightarrow y)\Rightarrow(x\Rightarrow y),(x\Rightarrow y)\land (x\land y^*))\underset{SH\ref{infimoeimplicaSH})}{=}(1,x\land (y\land y^*))=(1,0), we have that [\![(x\Rightarrow y,x\land y^*)]\!]=[\![(x\Rightarrow y,(x\Rightarrow y)^*)]\!]. Therefore, we conclude that \eta_A(x)\Rightarrow \eta_A(y)=\eta_A(x\Rightarrow y).
\eta_A(F)\subseteq [\![N(A,F)^+]\!]
If y\in\eta_A(F), then y=\eta_A(x), for some x\in F. Hence, y=[\![(x,x^*)]\!], with x\in F.
We observe that for x\in F, we have that (x,0)\in N(A,F). Besides, [\![(x,x^*)]\!]=[\![(x,0)]\!] because (x,0)\to(x,x^*)=(x\Rightarrow x,x\land x^*)=(1,0) and (x,x^*)\to(x,0)=(x\Rightarrow x,x\land 0)=(1,0).
Therefore, we can write y=[\![(x,0)]\!]. Now, since \sim(x,0)=(0,x)\leq(x,0), it follows that (x,0)\in N(A,F)^+, and hence y\in [\![N(A,F)^+]\!].
Finally, we check the naturality of \eta, that is, the following diagram commutes:
Notice that since \eta_A is an isomorphism, \eta_A^{-1}([\![(x,y)]\!])=x for every x\in A. Then we have
Thus f=\eta_{A'}^{-1}\circ \beta\alpha(f)\circ\eta_A, and therefore \eta_{A'}\circ f=\beta\alpha(f)\circ\eta_A.
Now we consider the functor \alpha\beta, which transforms every object \mathbf A of \mathbf{sN} into \mathbf{ N(sH(A),[\![A^+]\!])}. On a morphism f:A\longrightarrow A' it is defined by \alpha\beta(f)([\![x]\!],[\![y]\!])=([\![f(x)]\!],[\![f(y)]\!]). We define for each object \mathbf A of \mathbf{sN}, \delta_A:A\longrightarrow \mathbf{ N(sH(A),[\![A^+]\!])} by \delta_A(x)=([\![x]\!],[\![\sim x]\!]).
\delta_A is well defined: On the one hand, [\![x]\!]\cap[\![\sim x]\!]=[\![x\land\sim x]\!]=[\![0]\!], because (x\land\sim x)\to 0=(x\land\sim x)\to (x\land\sim x\land0)=(x\land\sim x)\mbox{ _N } 0=1 and 0\mbox{ _N } (x\land\sim x)=1 by well known properties of Nelson algebras. On the other hand, using SN?) and De Morgan properties we have \sim(x\lor\sim x)=\sim x\land x\leq x\lor\sim x, so x\lor\sim x\in A^+. Therefore [\![x]\!]\cup[\![\sim x]\!]=[\![x\lor\sim x]\!]\in [\![A^+]\!].
\delta_A is surjective: If ([\![x]\!],[\![y]\!])\in\alpha\beta(A,F), then ([\![x]\!],[\![y]\!])=([\![x]\!],[\![\sim x]\!]). Indeed, this is a consequence of ([\![x]\!],[\![y]\!])\to([\![x]\!],[\![\sim x]\!])=([\![x]\!]\Rightarrow [\![x]\!],[\![x]\!]\cap [\![\sim x]\!])=([\![1]\!],[\![x\land\sim x]\!])=([\![1]\!],[\![0]\!]) and ([\![x]\!],[\![\sim x]\!])\to([\![x]\!],[\![y]\!])=([\![x]\!]\Rightarrow [\![x]\!],[\![x]\!]\cap [\![y]\!])=([\![1]\!],[\![0]\!]). Therefore \delta_A(x)=([\![x]\!],[\![y]\!]).
\delta_A is injective: If ([\![x]\!],[\![\sim x]\!])=([\![y]\!],[\![\sim y]\!]), then [\![x]\!]=[\![y]\!] and [\![\sim x]\!]=[\![\sim y]\!], so we have the following: x\to y=1, y\to x=1, \sim x\to \sim y=1, \sim y\to \sim x=1. By proposition 2.11 (?), we have that x\mbox{ _N } y=1, y\mbox{ _N } x=1, \sim x\mbox{ _N } \sim y=1, \sim y\mbox{ _N } \sim x=1, which implies (see for example [viglizzo99algebras], (1.15)) that x=y.
\delta_A is a \mathbf{sN}-morphism:
\delta_A(x)\sqcup\delta_A(y)=([\![x]\!],[\![\sim x]\!])\sqcup([\![y]\!],[\![\sim y]\!])=([\![x]\!]\cup[\![y]\!],[\![\sim x]\!]\cap[\![\sim y]\!])=([\![x\lor y]\!],[\![\sim x\land\sim y]\!])=([\![x\lor y]\!],[\![\sim (x\lor y)]\!])=\delta_A(x\lor y). In a similar manner we can prove that \delta_A(x)\sqcap\delta_A(y)=\delta_A(x\land y).
\sim\delta_A(x)=\sim([\![x]\!],[\![\sim x]\!])=([\![\sim x]\!],[\![x]\!])=\delta_A(\sim x).
\delta_A(x)\to \delta_A(y)=([\![x]\!],[\![\sim x]\!])\to([\![y]\!],[\![\sim y]\!])=([\![x]\!]\Rightarrow[\![y]\!],[\![x]\!]\cap[\![\sim y]\!])=([\![x\to y]\!],[\![x\land\sim y]\!]).
By SN?) and SN?), we see that [\![x\land\sim y]\!]=[\![\sim(x\to y)]\!]. Hence \delta_A(x)\to \delta_A(y)=([\![x\to y]\!],[\![x\land\sim y]\!])=([\![x\to y]\!],[\![\sim(x\to y)]\!])=\delta_A(x\to y).
Finally, \delta is a natural transformation: to see that this diagram commutes,
we calculate \alpha\beta(f)\circ\delta_A(x)=\alpha\beta(f)([\![x]\!],[\![\sim x]\!])=([\![f(x)]\!],[\![f(\sim x)]\!])=([\![f(x)]\!],[\![\sim f(x)]\!])=\delta_{A'}\circ f(x).
Consider the semi-Nelson algebra \mathbf{A}, with Hasse diagram indicated below, in which the operations \to and \sim are given by the following tables:
1mm
S=\{0,a,d,e,g,1\} is the universe of a subalgebra \mathbf{S} of \mathbf{A}. Both \mathbf{S} and \mathbf{A} have the same quotient semi-Nelson algebra \mathbf{H} indicated below:
1mm
E=\{[\![e]\!],[\![1]\!]\} is an i-filter of \mathbf{H}, and \mathbf{N(H,E)} is isomorphic to \mathbf{S}, while \mathbf{N(H,H)}=\mathbf{V_k(H)} is isomorphic to \mathbf{A}.
Let \mathbf{A}\in\mbox{DSN}. If we consider the equivalence relation \equiv defined in section ?, then that relation is compatible with respect of the operation '. Hence, \equiv is a congruence in \mbox{DSN}.
Let \mathbf{A}=\langle A;\cap,\cup,\Rightarrow,^\dagger,\bot,\top \rangle\in\mbox{DSH}. For (a,b)\in V_k(A) we define the following unary operation on V_k(A):
If \mathbf{A}\in\mbox{DSH}, then the system \mathbf{V_k(A)}=\langle V_k(A); \sqcap,\sqcup,\to,\sim,',\top\rangle is a dually hemimorphic semi-Nelson algebra.
[cornejo19dually] If \mathbf{A}\in\mbox{DSH}, then \mathbf{A} is isomorphic to \mathbf{dsH}(\mathbf{V_k}(\mathbf A)). Also, if \mathbf{A}\in\mbox{DSN}, then \mathbf{A} is isomorphic to a subalgebra of \mathbf{V_k(dsH(A))}.
If \mathbf A\in\mbox{DSN}, then \mathbf A is centered, with center 1'.
We want to prove that \sim 1'=1', or equivalently, the following conditions holds simultaneously:
By Proposition 2.11 (?), (Proof) is equivalent to
which is equivalent to show that
By (DSN?), (DSN?), and proposition ?, we have that \sim 1'\mbox{ _N }(\sim 1\land(1'\to 1))=1, that is (1) \sim 1'\mbox{ _N }0=1. Since in every Nelson algebra 0\mbox{ _N } x=x holds for all x, it follows that (2) 0\mbox{ _N } 1'=1'. By (1), (2) and proposition 2.11 (?), we have that \sim 1'\mbox{ _N }1'=1, which proves (?).
By (DSN?) we have (3) 1=1'\to \sim1=1'\to0=1'\to(1'\land0)=1'\mbox{ _N }0. Using again that 0\mbox{ _N } 1'=1', and proposition 2.11 (?), gives us \sim 1'\mbox{ _N }1=1, which proves (?).
If \mathbf{A}\in\mbox{DSN}, then \mathbf{A} is isomorphic to \mathbf{V_k(dsH(A))}.
Another way of proving Theorem 4.6 is the following: if we consider the structure \mathbf{N(H,E)} as in Lemma 3.5, and take \mathbf{H}\in\mbox{DSH}, we have no guarantee that this set is going to be a closed under the operation ^\dagger. If instead we consider a subalgebra \mathbf S such that \pi_1(S)=H, then the proof indicated in the lemma still holds for the variety \mbox{DSH}, since it does not use the operations ' or ^\dagger. Hence, the sets S and N(H,E) are equal, so N(H,E) inherits the operations from \mathbf S. Indeed, due to Theorem 4.4, we know that the function h:A\to \mathbf{V_k}(\mathbf{dsH}(A)) defined by h(a)=([\![a]\!],[\![\sim a]\!]) is a monomorphism. Hence, A\cong h(A). Let us take then the subalgebra \mathbf S=h(\mathbf A) of \mathbf{V_k}(\mathbf{dsH}(\mathbf A)). It is immediate that \pi_1(S)=\mathbf{dsH}(A). Therefore, by Lemma 3.5, h(\mathbf A)=\mathbf{N(dsH(A),E)} for E=\{[\![x]\!]\in \mathbf{dsH}(A):[\![x]\!]=[\![a]\!]\lor [\![b]\!],([\![a]\!],[\![b]\!])\in S\}.
Since \mathbf S is a subalgebra of \mathbf{V_k}(\mathbf{dsH}(\mathbf A)), we have that ([\![1]\!],[\![0]\!])\in S, so the center ([\![0]\!],[\![0]\!])=([\![1]\!],[\![0]\!])' is an element of S. Then [\![0]\!]=[\![0]\!]\lor[\![0]\!]\in E. Since E is a filter, this shows that E=\mathbf{dsH}(A).
We have proved that S=N(\mathbf{dsH}(A),\mathbf{dsH}(A))=\mathbf{V_k}(\mathbf{dsH}(A)), and therefore h is an isomorphism.
Consider the semi-Nelson algebras \mathbf{A} and its subalgebra \mathbf{S} from Example 3.14. We know by the results above that \mathbf{S} cannot be expanded by adding a unary operation ' in such a way that the system \mathbf{B}=\langle B; \land,\lor,\to,\sim,',1 \rangle becomes a dually hemimorphic semi-Nelson algebra, since \mathbf{S} has no center.
On the other hand, any operation ' that makes \mathbf{sH(A)} into a dually hemimorphic semi-Heyting algebra yields an operation ^\dagger that will turn A into a dually hemimorphic semi-Nelson algebra.
The text on this page is from arXiv:2007.14459v1 by Juan Manuel Cornejo, Andrés Gallardo and Ignacio Viglizzo, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.
The record of each item is in analysis/objects/2007/2007.14459.jsonl.