Cent. Eur. J. Math. • 11(11) • 2013 • 1881-1899 DOI: 10.2478/s11533-013-0302-0
Central European Journal of Mathematics
Quantum B-algebras
Research Article
Wolfgang Rump1∗
1 Institute for Algebra and Number Theory, University of Stuttgart, Pfaffenwaldring 57, 70550 Stuttgart, Germany
Received 29 December 2012; accepted 16 January 2013
Abstract: The concept of quantale was created in 1984 to develop a framework for non-commutative spaces and quantum mechanics with a view toward non-commutative logic. The logic of quantales and its algebraic semantics man- ifests itself in a class of partially ordered algebras with a pair of implicational operations recently introduced as quantum B-algebras. Implicational algebras like pseudo-effect algebras, generalized BL- or MV-algebras, par- tially ordered groups, pseudo-BCK algebras, residuated posets, cone algebras, etc., are quantum B-algebras, and every quantum B-algebra can be recovered from its spectrum which is a quantale. By a two-fold application of the functor “spectrum”, it is shown that quantum B-algebras have a completion which is again a quantale. Every quantale Q is a quantum B-algebra, and its spectrum is a bigger quantale which repairs the deficiency of the inverse residuals of Q. The connected components of a quantum B-algebra are shown to be a group, a fact that applies to normal quantum B-algebras arising in algebraic number theory, as well as to pseudo-BCI algebras and quantum BL-algebras. The logic of quantum B-algebras is shown to be complete.
MSC: 03G12, 03G27, 06F07, 06F15, 06F35, 03F52, 03G25
Keywords: Quantale • Non-commutative logic • Partially ordered group © Versita Sp. z o.o.
Dedicated to B. V. M.
Introduction
As far as implicational algebras are concerned, the algebras of the title may be introduced by inverting a vivid expression of Nguyen Tien Zung [31]: To say that everything is a quantum B-algebra would be a great exaggeration, but to say that everything is contained in a quantum B-algebra would not be too far from the truth.
∗ E-mail: [email protected]
1881 Quantum B-algebras
In [36], we have shown that in a certain sense, every quantum B-algebra X contains a greatest pseudo-BCK subalgebra × I(X), a “greatest” partially ordered subgroup X , and even a greatest GPE-subalgebra E+(X), where GPE abbreviates “generalized pseudo-effect algebra” [9–11]. The algebraic semantics of almost every logic can thus be found among the special classes of quantum B-algebras. To say it from the start, the logic of quantum B-algebras is the logic of quantales. In fact, every quantum B-algebra can be embedded (dually) into a quantale, and every quantale is a quantum B-algebra. Recall that a quantale is a complete partially ordered semigroup (Q; &) where the multiplication distributes over arbitrary joins. Quantales were introduced by Mulvey on a 1984 conference in Taormina (Sicily). The very short title “&” of his seminal paper [27] refers to the non-commutative conjunction, while the infinitary join operation interprets the logical disjunction. Originally, quantales have been invented to formalize the logic of observables in quantum physics. They became useful ∗ to describe the “spaces” of C -algebras [3, 18, 24] and non-commutative geometry [6, 28–30, 34, 35]. Combining both aspects of quantales, we associate a quantalic spectrum U(X) to any quantum B-algebra X in such a way that X embeds into the opposite U(X)op of the quantale U(X) (with inverse partial order) so that X can be recovered from U(X). In general, the opposite of a quantale is not a quantale, nor a quantum B-algebra. This touches a delicate point that ought to be clarified beforehand. By the completeness of a quantale Q, the multiplication (we henceforth write · instead of &) gives rise to a pair of binary operations → and ; satisfying
x · y 6 z ⇐⇒ x 6 y → z ⇐⇒ y 6 x ; z (0)
for all x; y; z ∈ Q. As mentioned above, every quantale is a quantum B-algebra, a partially ordered system (X; →; ;) with two binary operations →; ; and additional properties (see Definition 1.2). For example, the right-hand equivalence of (0) holds for a quantum B-algebra. The role of quantales in this context is twofold. Apart from being a special quantum B-algebra, a quantale U(X) is associated to any quantum B-algebra X as a (non-commutative) spectrum. By definition, U(X) is the set of upper sets of X, so that X can be identified with the set of principal upper sets ↑x in U(X). The delicate point is that the operations →; ; in X do not come from the residuals →; ; of the multiplication in Q = U(X). While the latter are given by
_ _ y → z = {x ∈ Q : x ·y 6 z}; x ; z = {y ∈ Q : x ·y 6 z};
there are dual operations
^ ^ y z = {x ∈ Q : x ·y > z}; x z = {y ∈ Q : x ·y > z}
derived from the multiplication in Q. However these derived operations determine the multiplication in Q. If X is a quantum B-algebra, the inverse operations and of U(Q) induce the operations → and ; in X. Furthermore, there is a one-to-one correspondence X 7→ U(X) between the isomorphism classes of quantum B-algebras and logical quantales (Theorem 1.6). The fact that the quantale U(X) is again a quantum B-algebra allows us to form the iterated embedding X,→ U(X) ,→ U(U(X)). As indicated above, the operations →; ; of U(X) as a quantum B-algebra are induced by the derived operations ; of U(U(X)), while the derived operations ; of U(X) do not fit into an adjunction context (0) with the opposite partial order. However, it turns out that the secondary operations ; of U(X) are the restrictions of the residuals →; ; in U(U(X)). In other words, the pair of operations →; ; and ; in U(X) is just interchanged in U(U(X)). The lack of an adjunction context like (0) for the non-residuals ; is thus repaired in U(U(X)). In particular, the composed embedding X,→ U(X) ,→ U(U(X)) makes X into a subalgebra of a quantale. Therefore, every quantum B-algebra X has a completion Xb, the subquantale of U(U(X)) generated by X, which generalizes the Dedekind–MacNeille completion of an archimedean lattice-ordered group (cf. [2, Théorème 11.3.2]). Secondly, this implies that a quantum B-algebra can be characterized as a subreduct (and subposet) of a quantale (Theorem 2.3).
1882 W. Rump
One of the referees suggested to look at the correspondence between a quantum B-algebra X and its quantalic spectrum Q = U(X) in the opposite way: The quantale Q takes the part of an algebra while X is viewed as its underlying space, similar to the duality between a bounded distributive lattice with operators and its associated relational Priestley space (see [19, Theorems 2.2.3, 2.2.4]). To this end, the infinite join operation as well as the special structure of Q = U(X) as a logical quantale (see Definition 1.5) has to be taken into account. The quantum B-algebra X can then be identified 1 1 with the set of logical [36] quantale morphisms Q → Q( ) onto the endomorphism quantale Q( ) of the poset {0; 1} with 0 < 1. The multiplication of the quantale Q corresponds to a ternary relation in X. Note, however, that the residuals of the multiplication in U(X), as explained above, do not restrict to the binary operations of X. In this context, the relationship with canonical extensions, introduced by Jónsson and Tarski [20, 21] and further developed by Gehrke and Jónsson [13, 14], might be of interest. In [8], the canonical extension is introduced for an arbitrary poset with monotonic operations. Its construction as a MacNeille completion of an “intermediate structure” (see [15]) resembles the above mentioned completion Xb of a quantum B-algebra X. If X is a residuated poset, that is, a partially ordered semigroup with binary operations → and ; satisfying (0), the completion Xb satisfies one and a half of the two characterizing properties of the canonical extension (see Section 2). Every residuated poset is a quantum B-algebra. For example, a partially ordered group is a residuated poset with
− − x → y = yx 1; x ; y = x 1y:
We characterize partially ordered groups as quantum B-algebras X which satisfy
(x → y) ; y = (x ; y) → y = x for all x; y ∈ X (Proposition 3.6). Now assume that X is unital, that is, X contains an element u which makes U(X) × into a unital quantale. By [36, Theorem 3], the whole unit group U(X) of U(X) is contained in X. However, it will turn × × out that the unit group X = U(X) need not be the largest subgroup of X. Namely, we show that every pseudo-BCI algebra [22] is a quantum B-algebra (Proposition 4.4), and that the maximal elements of a pseudo-BCI algebra X form × a subgroup which can happen to be strictly larger than the unit group X . In Section 4, we introduce the connected components of a quantum B-algebra X. We show that they form a group C(X) − even if X does not have a unit element. There is a natural epimorphism c : X C(X) such that the unit fiber c 1(u) is the greatest connected subalgebra of X. In particular, a group can be regarded as a quantum B-algebra whose partial order is trivial. For a unital quantum × B-algebra X, we show that the morphism c : X C(X) admits a section into the unit group X if and only if X is a semidirect product G nY of a group G by a connected unital quantum B-algebra Y (Proposition 4.3). The unital quantum B-algebras satisfying the identity
x → x = x ; x = u arise, e.g., as lattices of fractional ideals in algebraic number fields. We show that every such quantum B-algebra X admits an endomorphism x 7→ x whose image X is a partially ordered group (Theorem 4.7). If X is a pseudo-BCI algebra, × X consists of the maximal elements of X. We show that in this case, X coincides with the unit group X if and only if ∼ X = G nY with a group G and a pseudo-BCK algebra Y (Theorem 4.10). The logic (qB) of quantum B-algebras is briefly discussed in Section 5. It consists of four axioms and three inference rules x; x → y ` y; x → y ` x ; y; x ; y ` x → y:
Models of (qB) are quantum B-algebras X which are locally unital. If X has a unit element u, the true propositions are associated to the upper set X + = ↑u. In the locally unital case, the single unit element u is replaced by local units e ∈ X defined by the equation e → e = e. Every element of the form el(x) = x → x or er(x) = x ; x is a local unit, and the equations u → x = u ; x = x which characterize u in the unital case, are replaced by the equations
el(x) ; x = er(x) → x = x
1883 Quantum B-algebras
for a locally unital quantum B-algebra. Logically, this implies that the true propositions x are characterized by the equivalent inequalities el(x) 6 x and er(x) 6 x. The Lindenbaum–Tarski process works for (qB) and yields the free locally unital quantum B-algebra L with countably many generators. We use L to show that the logic (qB) is complete.
1. Non-commutative logic and quantales
In this section, we briefly review the logical aspects of quantales which led to the concept of quantum B-algebra [36]. Let us recall first that in classical propositional logic, the conjunction A &B and the implication A → B are related by an adjunction A &B 6 C ⇐⇒ A 6 B → C; where 6 stands for the logical entailment. If the commutativity of & is dropped, the implication splits into a left and a right implication, according to the partial maps A 7→ A&B and B 7→ A&B. We keep → as a symbol for the right implication, and use ; for the left implication which is related to → by the equivalence
x 6 y → z ⇐⇒ y 6 x ; z: (1)
In the presence of a conjunction, now denoted by · instead of &, the equivalence (1) extends to
x 6 y → z ⇐⇒ x · y 6 z ⇐⇒ y 6 x ; z; (2)
but even if there is no conjunction, the left and right implication determine each other via (1). Thus, in a wide sense, non- commutative logic can be regarded as the theory of partially ordered algebras (X; →; ;) with two binary operations → and ; satisfying (1). We call X commutative if the operations → and ; coincide. Non-commutative algebras (X; →; ;) of that type have been studied since 1939 by Ward and Dilworth [37] (residuated lattices), Bosbach [4] (pseudo-hoops) and [5] (cone algebras, bricks), Georgescu and Iorgulescu [16, 17] (pseudo-MV algebras, pseudo BCK-algebras), Dvureˇcenskij and Vetterlein [9–11] (GPE-algebras), and Galatos and Tsinakis [12] (GBL-algebras). The case where → and ; are connected with a multiplication, is given by
Definition 1.1. A residuated poset is a partially ordered semigroup (X; ·) with two binary operations → and ; which satisfy (2).
Note that the associativity of · is equivalent to the equational version of (1):
x → (y ; z) = y ; (x → z): (3)
Indeed,
(x · y) · z 6 t ⇐⇒ x · y 6 z → t ⇐⇒ y 6 x ; (z → t) x · (y · z) 6 t ⇐⇒ y · z 6 x ; t ⇐⇒ y 6 z → (x ; t)
holds for all x; y; z; t ∈ X. Note however, that (1) and (3) are not equivalent.
Definition 1.2. A quantum B-algebra [36] is a partially ordered set X with two binary operations → and ; satisfying (1) and (3), and
y 6 z =⇒ x → y 6 x → z (4)
for all x; y; z ∈ X.
1884 W. Rump
Thus every residuated poset is a quantum B-algebra. Equation (4) has a counterpart
y 6 z =⇒ x ; y 6 x ; z (5) which holds for any quantum B-algebra. In fact, (1) gives
x 6 (x → y) ; y; x 6 (x ; y) → y; (6) and thus, y 6 z implies x 6 (x ; y) → y 6 (x ; y) → z by means of (4). Hence (1) yields x ; y 6 x ; z, which proves (5). Furthermore, every quantum B-algebra satisfies
y → z 6 (x → y) → (x → z) (7) y ; z 6 (x ; y) ; (x ; z) (8) which can be also written as
x → y 6 (y → z) ; (x → z) (9) x ; y 6 (y ; z) → (x ; z) (10) by means of (1). By (3), the right-hand side of (9) is equal to x → ((y → z) ; z). Whence (9) follows by (4) and (6). Conversely, it is easily seen [36, Proposition 3] that (3) in Definition 1.2 can be replaced by (7) and (8).
Definition 1.3. W A quantale Q is a partially ordered semigroup (Q; ·) with arbitrary joins A (for A ⊂ Q) so that
! ! _ _ _ _ a· ai = (a·ai); ai ·a = (ai ·a) (11) i∈I i∈I i∈I i∈I holds for all a; ai ∈ Q. A quantale Q is said to be unital if the semigroup (Q; ·) admits a unit element u.
W A quantale Q can be conceived as a non-commutative topology: Elements a ∈ Q stand for the open sets, A for ∗ the union, and a·b for the non-commutative intersection. For example, the spectrum of a C -algebra or the “space” of Penrose tilings can be viewed as a quantale [24, 30]. On the other hand, if the elements a ∈ Q of a quantale Q W are interpreted as logical propositions, the union A stands for the disjunction, while a·b features the sequential conjunction “a and then b”. W W Every quantale Q has a smallest element 0 = ∅, and a greatest element 1 = Q. The residuals are determined by (2) and can be defined as
_ _ a → b = {c ∈ Q : c ·a 6 b}; a ; b = {c ∈ Q : a·c 6 b}: (12)
Since every quantale Q is a complete lattice, the inverse residuals
^ ^ a b = {x ∈ Q : x ·a > b}; a b = {x ∈ Q : a·x > b} (13) are well defined, too.
1885 Quantum B-algebras
Definition 1.4. Let Q be a quantale. We call an element c ∈ Q balanced if it satisfies
! ! ^ ^ ^ ^ c · ai = (c ·ai); ai ·c = (ai ·c): i∈I i∈I i∈I i∈I
Equivalently, c is balanced if and only if
a·c > b ⇐⇒ a > c b; c ·a > b ⇐⇒ a > c b
hold for all a; b ∈ Q. An element c of a quantale Q is said to be supercompact if for any subset A ⊂ Q,
_ c 6 A =⇒ there exists a ∈ A such that c 6 a:
The set of supercompact elements of Q will be denoted by Qsc. The product of balanced elements is balanced. If c is balanced and d supercompact, then c d and c d are supercompact.
Definition 1.5. Q logical Q a ∈ Q We say that aW quantale is if every supercompact element of is balanced and every can be represented as a join a = C with C ⊂ Qsc.
Thus, for a logical quantale Q, the set X = Qsc of supercompact elements is closed under the operations and . For x; y ∈ X, we therefore define x → y = x y; x ; y = x y
to make X into an algebra (X; →; ;). We endow X with the partial ordering induced by the partial order of Qop. Then X becomes a quantum B-algebra. The converse is less obvious: For every quantum B-algebra X, the upper sets A ⊂ X (i.e. the subsets A with a > b ∈ A ⇒ a ∈ A) can be made into a quantale U(X) by defining
A·B = {x ∈ X : there exists b ∈ B such that b → x ∈ A}: (14)
It can be shown [36] that this gives an associative multiplication which distributes over set-theoretic joins. Furthermore, the set of supercompact elements of U(X) coincides with the image of the embedding X,→ U(X) given by x 7→ ↑x, and every balanced element of U(X) is supercompact. So we have
Theorem 1.6. Up to isomorphism, there is a one-to-one correspondence X 7→ U(X) between quantum B-algebras and logical quantales.
With suitable morphisms, this correspondence extends to a categorical duality. Let f : X → Y be a morphism of quantum B-algebras, that is, a monotonic map which satisfies
f(x → y) = f(x) → f(y); f(x ; y) = f(x) ; f(y)
for all x; y ∈ X. Then Y → f(X) ⊂ f(X) ⇐⇒ Y ; f(X) ⊂ f(X); (15)
where Y → f(X) denotes the set of elements y → f(x) with x ∈ X and y ∈ Y , and similarly for Y ; f(X). If these equivalent conditions hold, we call the morphism f spectral. The category of quantum B-algebras with spectral
1886 W. Rump
morphisms will be denoted by qB. Similarly, we call a morphism g: Q → L of quantales logical if g respects arbitrary meets and satisfies g(Q) → L ⊂ g(Q); g(Q) ; L ⊂ g(Q): (In contrast to (15), these two conditions are not equivalent.) By LQuant we denote the category of logical quantales with logical morphisms. Then we have an equivalence of categories [36, Theorem 2]:
U : qBop → LQuant:
For a quantum B-algebra X, we call U(X) the enveloping quantale or the quantalic spectrum of X.
2. The case of residuated posets
Every quantale is a quantum B-algebra, but the quantalic spectrum U(Q) of a quantale Q is usually bigger than Q. The partial order of Q is opposite to that of U(Q). Moreover, U(Q) is a logical quantale. Therefore, Theorem 1.6 shows that the theory of quantales (with spectral morphisms) is part of the theory of logical quantales. One might ask whether it is more natural to redefine the quantalic spectrum in such a way that a quantale Q is associated to itself rather than to the bigger quantale U(Q). Such a direct embedding is well known for residuated posets. In fact, every residuated poset X naturally embeds into a quantale L(X) consisting of the lower sets A ⊂ X (i.e. a 6 b ∈ A ⇒ a ∈ A). Thus L(X) is a complete lattice with respect to inclusion. For A; B ∈ L(X), we set
A·B = {x ∈ X : there exist a ∈ A, b ∈ B such that x 6 a·b}: (16)
The embedding X,→ L(X) is given by x 7→ ↓x, and X can be recovered as the set L(X)sc of supercompact elements (cf. Ono and Komori [32, 33]). Note that the associativity of X carries over to L(X) and thus makes L(X) into a quantale. However, the embedding X,→ L(X) fails to give a quantale if the residuals are not compatible with a product. The following example shows that even in the commutative case, L(X) need not be a quantale.
Example 2.1. Let (X; →; u) be the 4-element BCI-algebra given by the following table (cf. [23, Example 3.8]).
→ u a b c u u a b c a u u b c b u a u c c c c c u
Recall that a BCI-algebra is defined by the axioms
(B) y → z 6 (x → y) → (x → z); (C) x → (y → z) = y → (x → z); (I) x → x = u; together with x 6 y 6 x =⇒ x = y; where the relation 6 (a partial order) is given by
x 6 y ⇐⇒ x → y = u:
Thus every BCI-algebra is a commutative quantum B-algebra. The algebra L(X) consists of 10 elements, L(X) = {0; a; b; c; u; ab; ac; bc; abc; 1}, where 0 = ∅, 1 = {a; b; c; u}, and ab = {a; b}, etc. The lattice structure of L(X) thus looks as follows.
1887 Quantum B-algebras
The equivalences (2) give the following table of products for the elements of X:
· u a b c u u a b c a a a 0 c b b 0 b c c c 0 0 0
which yields the multiplication in L(X) via (16). Note that the multiplication in L(X) is non-commutative although the two residuals coincide. In particular, we have (a·b)·c = 0·c = 0, but a·(b·c) = a·c = c. Thus L(X) is not a quantale.
A close inspection of Example 2.1 shows that the definition of a product in L(X) via (2) is not natural as it makes use V W V of instead of . For x; y ∈ X, we have x ·y = {c ∈ L(X): x 6 y → c}. For a residuated poset X, there are two natural embeddings into a quantale:
1. the direct embedding X,→ L(X); 2. the spectral embedding X,→ U(X):
The direct embedding does not work for pseudo-BCK algebras, although every pseudo-BCK algebra X can be embedded into the (→; ;)-subreduct of a quantale [1, 25]. On the other hand, the latter result extends to arbitrary quantum B-algebras. We obtain this as a consequence of a duality result (Proposition 2.2). Recall that for a quantum B- algebra X, the residuals → and ; of X coincide with the inverse residuals and of U(X), respectively. If X is a quantale, the opposite statement is also true.
Proposition 2.2. Let Q be a quantale. For x; y ∈ Q, the inverse residuals (13) in Q coincide with the residuals (12) in U(Q).
Proof. Let A; B ∈ U(Q) be given. By (14),