Vous êtes sur la page 1sur 7

Algebra Univers.

70 (2013) 403409
DOI 10.1007/s00012-013-0258-8
Published online September 26, 2013
Springer Basel 2013

Algebra Universalis

The product representation theorem for interlaced


pre-bilattices: some historical remarks
Brian A. Davey
Abstract. This note aims to unravel the history of the Product Representation
Theorem for Interlaced Pre-bilattices. We will see that it has its lattice-theoretic
roots in early attempts to solve one of the problems in Birkhos Lattice Theory. The
theorem was presented in its full generality by Czedli, Huhn and Szab
o at a conference
in Szeged, Hungary in 1980 (and published in 1983). This was several years before
Ginsberg introduced bilattices at a conference on articial intelligence in 1986 and in
his foundational paper in 1988.

1. Introduction
This paper is a survey of the history of the Product Representation Theorem for Interlaced Pre-bilattices, and in passing of the closely related result
known the 90-Degree Lemma. It arose when the author was asked to review a
paper on bilattices. After reading the paper, and subsequent reading of related
papers on interlaced pre-bilattices, it became clear that the bilattice community was unaware of some highly relevant papers in the lattice-theory literature
dating back to the 1950s and the 1980s. Unfortunately, the two most relevant
lattice-theory papers from the early 1980s on compatible orders of lattices were
published in conference proceedings and would not have been readily available
to early workers on bilattices. Moreover, some relevant papers from the 1950s
were published in Russian, making them inaccessible to many in the West.
Nevertheless, with the availability of online resources like MathSciNet and the
power of internet search engines, it is surprising that the intimate connection
between interlaced pre-bilattices and compatible lattice orders has not been
discovered before now. The author hopes that the historical comments set out
below will help to set the record straight.
Bilattices were introduced by Ginsberg [14, 15] in 1986/1988 as a framework for inference in articial intelligence. Work on bilattices continued in
the 1990ssee, for example, Romanowska and Trakul [26], Fitting [12], and
Avron [2]. After a period of relative dormancy, there has been a recent spurt
of activity on bilattices. Rivieccios PhD thesis [25], which is available online,
is a very useful source for the basic theory of (pre-)bilattices. The terminology
concerning bilattices and their generalisations is not uniform, so we begin by
Presented by G. Gratzer.
Received May 20, 2013; accepted in nal form May 22, 2013.
2010 Mathematics Subject Classication: Primary: 06B05; Secondary: 03G10.
Key words and phrases: bilattice, compatible ordering of a lattice.

404

B. A. Davey

Algebra Univers.

xing ours. A pre-bilattice L is simply a set carrying two lattice structures, L1


and L2 , and therefore possessing four binary operations; to simplify our discussion, we shall write L = L1 , L2 . A bilattice is a pre-bilattice equipped with
an involution : L L that is a dual lattice automorphism with respect to
L1 and a lattice automorphism with respect to L2 . A pre-bilattice is bounded
if both L1 and L2 are bounded. A pre-bilattice is called distributive if each
of its four operations distributes over the other three. It is easy to see that
in a distributive pre-bilattice, the lattice operations in each of the lattices are
order-preserving with respect to the order of the other lattice. Pre-bilattices
satisfying this weaker condition are said to be interlaced. Equivalently, L is
interlaced if the order on each lattice is compatible with the operations of the
other, that is, the order on each forms a sublattice of the square of the other.
It is this view of interlaced pre-bilattices that links them to lattice-theory
literature from the 1980s.
Every two-factor decomposition of a lattice L1 gives rise to an interlaced
pre-bilattice: if : L1 KM is a lattice isomorphism, we dene L2 to be the
uniquely dened lattice on the underlying set of L1 such that : L2 KM
is a lattice isomorphism. When this happens, we will say that the lattice L2
and the pre-bilattice L = L1 , L2  arise from a two-factor decomposition of L1 .
We intend to trace the history of the following fundamental result.
Product Representation Theorem for Interlaced Pre-bilattices. A
pre-bilattice L = L1 , L2  is interlaced if and only if it arises from a two-factor
decomposition of L1 .
2. The lattice history
We will see that construction of a second lattice structure L2 on the same
base set as a lattice L1 via a two-factor decomposition of L1 has its roots in
the 1950s in attempts to solve a problem from the second edition of Birkhos
Lattice Theory [5]. Problem 8, which is repeated as Problem 6 in the third
edition [6], states: Find a necessary and sucient condition on a lattice L,
in order that every lattice M whose (unoriented ) graph is isomorphic with the
graph of L be lattice-isomorphic with L. Here graph-automorphism of L means
an automorphism of the undirected covering graph G(L) of L. In approaching
this problem, it is natural to consider the following more general problem:
Given a nite lattice L1 , describe all lattices L2 such that G(L2 ) is isomorphic
to G(L1 ). It is trivial that if L2 arises from a two-factor decomposition of L1 ,
then G(L2 ) is identical to G(L1 ). Surprisingly, for a large class of lattices, this
is essentially the only way that such pairs of lattices can arise. This problem
was studied intensively, especially by J. Jakubk, between the mid 1950s and
the mid 1980s.
In 1980, attention turned from considering pairs of lattices with the same
unoriented graph to pairs of lattices with compatible orders, that is, in the

Bilattices: some historical remarks

405

modern parlance, to interlaced pre-bilattices. The full statement and proof


of the Product Representation Theorem for Interlaced Pre-bilattices was presented by Czedli, Huhn and Szab
o at a conference in Szeged, Hungary in 1980
and appeared in print in 1983. They also gave the rst proof of the 90-Degree
Lemma. In the timeline below, statements from the older literature have been
translated into modern bilattice terminology.
1947 Kiss self-published text [20] contains many examples (but no theory) of
multiple lattice structures on a product of lattices obtained by reversing
the order in one or more coordinates. (See also Birkho and Kiss [7] and
Birkhos review of Kiss text [4].) On page 106, Kiss gives the example
of 2 2 with four possible orders. The bilattice on 2 2 is considered in
the context of Boolean algebras on pages 136139. Kiss discussion of a
logic based on the four-element bilattice on pages 300303 is a precursor
to Belnaps four-valued logic [3].
1951 Arnold [1] considered algebras L = L; , ,  such that L; ,  is
a lattice, L;  is a semilattice and all possible distributive laws hold
amongst , and . The Product Representation Theorem for (not
necessarily bounded) distributive pre-bilattices follows immediately from
his Theorem 17. Among his preliminary calculations (Theorem 2) is the
fact that every distributive pre-bilattice is interlaced.
1954 Jakubk and Kolibiar [16] also proved the Product Representation Theorem for (not necessarily bounded) distributive pre-bilattices. In addition, they proved that, for nite distributive lattices L1 and L2 with the
same underlying set, G(L1 ) = G(L2 ) if and only if L := L1 , L2  is a
distributive pre-bilattice.
1954 Jakubk [17] proved that, for nite modular lattices L1 and L2 with the
same underlying set, G(L1 ) = G(L2 ) if and only if L2 arises from a
two-factor decomposition of L1 .
1975 Jakubk [18] extended his 1954 result to discrete modular lattices, that
is, those in which each bounded chain is nite. He proved that if L1 is a
discrete modular lattice and L2 is a discrete lattice such that G(L1 ) =
G(L2 ) then L2 arises from a two-factor decomposition of L1 .
1980 Czedli, Huhn and Szab
o [10] considered compatible lattice orders of
lattices and proved both the 90-degree Lemma, see (c) below, and the
Product Representation Theorem for Interlaced Pre-bilattices. Their
main results are as follows.
(a) A lattice that has a compatible bounded order is itself bounded.
(b) A bounded compatible order of a lattice is a lattice order.
(c) The compatible lattice orders on a bounded lattice are in one-to-one
correspondence with the complemented neutral elements of L, and
consequently every such order arises from a two-factor decomposition of L. This appears to be the rst proof of what is now known
as the 90-degree Lemma (see Jung and Moshier [19, page 31]).

406

B. A. Davey

Algebra Univers.

The proof is straightforward and proceeds via a direct calculation


of mutually inverse maps between the set of complemented neutral
elements of L and the set of compatible lattice orders of L.
(d) Every compatible lattice order of a lattice L arises from a twofactor decomposition of L. The Product Representation Theorem
for Interlaced Pre-bilattices follows immediately from this result,
since there compatibility is assumed in both directions.
The proof uses the compatible order to dene a pair of factor congruences 1 , 2 on L such that L
= L/1 L/2 . The description
of 1 and 2 depends on the fact that the two orders on L have
the same closed intervals, which in turn depends upon the bounded
version of the theorem proved in (c).
1983 Kolibiar [21] proved that compatible orders on a lattice L are in a one-toone correspondence with two-factor subdirect representations of L. The
descriptions of the congruences 1 , 2 on L required for the subdirect
representation is much simpler that the one given by Czedli, Huhn and
Szab
o [10]: if  denotes the natural order on the lattice L and
denotes
a compatible order on L, then


x 1 y (v L) x  v  y & x
v  y ,


x 2 y (u L) x  u  y & x  u
y .
Moreover, he proved that a compatible order of a lattice is itself a lattice
order if and only if it is both up- and down-directed. This gives the
Czedli, Huhn and Szab
o result (b) above as an immediate corollary.
1984 Rosenberg and Schweigert [27] consider compatible orderings of algebras
with a majority term, of semilattices and of lattices. They independently obtain the results of Kolibiar [21] stated above via very similar
proofs. They use their results to give a new proof of Czedli, Huhn and
Szab
os result that every compatible lattice order on a lattice arises from
a two-factor decomposition of L. They also made the observation that
compatibility of lattice orders is symmetric: see (1) below.
1995 Czedli and Szab
o [11] give a particularly straightforward proof that the
lattice Quord(L) of quasiorders on a lattice L is isomorphic to Con2 (L),
the square of the congruence lattice of L. The mutually inverse isomorphisms L : Quord(L) Con2 (L) and L : Con2 (L) Quord(L) are
given by



 (
) ( ), (
) ( ) , and
(1 , 2 )  (1 ) (2 ).
Under this isomorphism compatible orders on L correspond to pairs
(1 , 2 ) Con2 (L) with 1 2 = 0, that is, to subdirect representations of L, and compatible lattice orders on L correspond to pairs
(1 , 2 ) Con2 (L) with 1 2 = 0 and 1 2 = 1, that is, to two-factor
decompositions of L.

Bilattices: some historical remarks

407

We trace the bilattice history of the Product Representation Theorem for


Interlaced Pre-bilattices in the next section. Before doing so, we present several
immediate consequences for pre-bilattices of the fact that every compatible
lattice order of a lattice L arises from a two-factor decomposition of L. Most
of these statements are well known,
(1) If L2 is a compatible lattice order of a lattice L1 , then L1 is a compatible lattice order of L2 (thus the denition of interlaced pre-bilattice can
be weakened). This was proved independently in the bounded case by
Pynko [24].
(2) A pre-bilattice L = L1 , L2  is distributive if and only if it is interlaced
and either L1 or L2 is a distributive lattice.
(3) If L = L1 , L2  is an interlaced pre-bilattice, then L1 satises a lattice
identity if and only if L2 does.
(4) Since lattices have factorizable congruences (see Fraser and Horn [13]) the
congruences on M K are precisely the products of congruences on the
factors, from which it follows at once that an equivalence relation on the
underlying set of an interlaced pre-bilattice L = L1 , L2  is a congruence
on L if and only if it is a congruence on either L1 or L2 .
(5) An interlaced pre-bilattice L = L1 , L2  is subdirectly irreducible if and
only if L1 is a subdirectly irreducible lattice and L2 {L1 , L1 }.
(6) If L2 is a compatible lattice order of a lattice L1 , then L1 and L2 share
the same closed intervals (as sets), and consequently, if L = L1 , L2  is an
interlaced pre-bilattice, then L1 and L2 have the same unoriented covering
graph.

3. The bilattice history


The bilattice community were not aware of the literature on compatible
orderings of lattices and reproved the Product Representation Theorem for
Interlaced Pre-bilattices over a number of years. As Mobasher, Pigozzi, Slutski and Voutsadakis remark, This theorem has a complicated evolution [22,
page 113].
1989 Romanowska and Trakul [26] proved the Product Representation Theorem
in the case of bounded interlaced bilattices (in which case L1
= L2 ), but
the proof is easily modied to yield the proof for bounded interlaced prebilattices. The proof is somewhat indirect and uses Plonka sums.
1990 M. Fitting [12] proved the Product Representation Theorem in the case
of complete distributive interlaced pre-bilattices.
1996 A. Avron [2], independently of Romanowska and Trakul, proved the Product Representation Theorem for bounded interlaced pre-bilattices. The
proof is short and direct and avoids the machinery of Plonka sums used
in [26]. The 90-degree Lemma is proved here.

408

B. A. Davey

Algebra Univers.

2000 Pynko [24] independently obtained the extension to bounded interlaced


pre-bilattices, which he referred to as regular bilattices. He also showed
that, in the bounded case, it was enough to assume only that the order
on L2 is compatible with the order on L1 .
2000 Mobasher, Pigozzi, Slutski and Voutsadakis [22] also independently proved
the extension to bounded interlaced pre-bilattices. In a footnote, they
thank a referee for pointing out the earlier work of Avron [2] and the almost simultaneous work of Pynko [24]. They show that the representation
is in fact the object half of a category equivalence between the category of
bounded interlaced bilattices and the categorical square of the category
of bounded lattices.
2006 Movsisyan, Romanowska and Smith [23] extended the Product Representation Theorem to the not-necessarily bounded case. The proof is based
on the 1989 proof given for bounded interlaced bilattices in [26] and so
uses Plonka sums.
2010 As part of his PhD thesis, Rivieccio [25] also independently extended the
Product Representation Theorem to the not-necessarily bounded case.
The proof is direct and proceeds by proving that
1 := (2 1 ) (2 1 )1

and 2 := (2 1 ) (2 1 )1

are congruences on L that yield the required factorisation. These congruences agree with those dened in Kolibiar [21] and Rosenberg and
Schweigert [27]. The proof was published a year later by Bou and Rivieccio [9].
2011 With the full Product Representation Theorem available to them, Bou,
Jansana, and Rivieccio [8] observe that the category of interlaced prebilattices is equivalent to the the categorical square of the category of
lattices. This extends the corresponding result in the bounded case by
Mobasher, Pigozzi, Slutski and Voutsadakis [22].
Acknowledgement. The author would like to thank Hilary Priestley for
helpful comments on several drafts of this paper.
References
[1] Arnold, B. H.: Distributive lattices with a third operation dened. Pacic J. Math. 1,
3341 (1951)
[2] Avron, A.: The structure of interlaced bilattices. Mathematical Structures in
Computer Science 6, 287299 (1996)
[3] Belnap, N. D.: A useful four-valued logic. Modern uses of multiple-valued logic (Fifth
Internat. Sympos., Indiana Univ., Bloomington, Ind., 1975), Episteme, vol. 2, pp. 537
(1977)
[4] Birkho, G.: Review of S. A. Kiss, Transformations on lattices and structures of
logic. Bull. Amer. Math. Soc. 54, 675676 (1948)
[5] Birkho, G.: Lattice Theory, 2nd (revised) edn. Amer. Math. Soc. Colloq. Publ. vol.
21, Providence (1948)
[6] Birkho, G.: Lattice Theory, 3rd edn. Amer. Math. Soc. Colloq. Publ. vol. 21,
Providence (1967)

Bilattices: some historical remarks

409

[7] Birkho, G., Kiss, S. A.: A ternary operation in distributive lattices. Bull. Amer.
Math. Soc. 53, 749752 (1947)
[8] Bou, F., Jansana, R. and Rivieccio, U.: Varieties of interlaced bilattices. Algebra
Universalis 66, 115141 (2011)
[9] Bou, F., Rivieccio, U.: The logic of distributive bilattices. Logic Journal of the
I.G.P.L. 19, 183216 (2011)
[10] Cz
edli, G., Huhn A. P., Szab
o, L.: On compatible ordering of lattices. In :
Contributions to Lattice Theory (Szeged, 1980), pp. 8799. Colloq. Math. Soc. J
anos
Bolyai 33. North-Holland, Amsterdam (1983)
[11] Cz
edli, G., Szab
o, L.: Quasiorders of lattices versus pairs of congruences. Acta Sci.
Math. (Szeged) 60, 207211 (1995)
[12] Fitting, M.: Bilattices in logic programming. In: Proceedings of the 20th International
Symposium on Multiple-Valued Logic, pp. 238246. The IEEE Computer Society
Press, Charlotte (1990)
[13] Fraser, G. A., Horn, A.: Congruence relations in direct products. Proc. Amer. Math.
Soc. 26, 390394 (1970)
[14] Ginsberg, M. L.: Multi-valued logics, in: Proceedings of AAAI-86, Fifth National
Conference on Articial Intelligence, pp. 243247. Morgan Kaufmann Publishers, Los
Altos (1986)
[15] Ginsberg, M. L.: Multivalued logics: A uniform approach to inference in articial
intelligence. Comput. Intelligence 4, 265316 (1988)
[16] Jakubk, J., Kolibiar, M.: On some properties of a pair of lattices. Czech. Math. J. 4,
127 (1975) (Russian with English summary)
[17] Jakubk, J.: On lattices whose graphs are isomorphic. Czech. Math. J. 4, 131142
(1954) (Russian with English summary)
[18] Jakubk, J.: Unoriented graphs of modular lattices. Czech. Math. J. 25(100),
240246 (1975)
[19] Jung, A., Moshier, M. A.: On the bitopological nature of Stone duality. Technical
Report CSR-06-13, School of Computer Science, University of Birmingham (2006)
[20] Kiss, S. A.: Transformations on Lattices and Structures of Logic. New York (1947)
See http://babel.hathitrust.org/cgi/pt?id=uc1.b4248825;seq=7;view=1up
[21] Kolibiar, M.: Compatible orderings in semilattices. In: Contributions to General
Algebra, 2 (Klagenfurt, 1982), pp. 215220. H
older-Pichler-Tempsky, Vienna (1983)
[22] Mobasher, B., Pigozzi, D., Slutski, V. and Voutsadakis, H.: A duality theory for
bilattices. Algebra Universalis 43, 10925 (2000)
[23] Movsisyan, Y. M., Romanowska, A. B., Smith, J. D. H.: Superproducts,
hyperidentities, and algebraic structures of logic programming. J. Combin. Math.
Combin. Comput. 58, 101111 (2006)
[24] Pynko, A. P.: Regular bilattices. J. Appl. Non-Classical Logics 10, 93111 (2000)
[25] Rivieccio, U.: An Algebraic Study of Bilattice-based Logics. PhD Thesis, University
of Barcelona (2010). See http://arxiv.org/abs/1010.2552
[26] Romanowska, A., Trakul, A.: On the structure of some bilattices. In: Halkowska, K.,
Slawski, B. (eds) Universal and Applied Algebra, pp. 246253. World Scientic (1989)
[27] Rosenberg, I. G.; Schweigert, D.: Compatible orderings and tolerances of lattices. In:
Orders: Description and Roles (LArbresle, 1982), North-Holland Math. Stud., vol. 99,
pp. 119150. North-Holland, Amsterdam (1984)
Brian A. Davey
Mathematics and Statistics, La Trobe University, Victoria 3086, Australia
e-mail: B.Davey@latrobe.edu.au

Vous aimerez peut-être aussi